diff --git a/.idea/.gitignore b/.idea/.gitignore
new file mode 100644
index 0000000..26d3352
--- /dev/null
+++ b/.idea/.gitignore
@@ -0,0 +1,3 @@
+# Default ignored files
+/shelf/
+/workspace.xml
diff --git a/.idea/inspectionProfiles/Project_Default.xml b/.idea/inspectionProfiles/Project_Default.xml
new file mode 100644
index 0000000..56d5f69
--- /dev/null
+++ b/.idea/inspectionProfiles/Project_Default.xml
@@ -0,0 +1,105 @@
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/inspectionProfiles/profiles_settings.xml b/.idea/inspectionProfiles/profiles_settings.xml
new file mode 100644
index 0000000..105ce2d
--- /dev/null
+++ b/.idea/inspectionProfiles/profiles_settings.xml
@@ -0,0 +1,6 @@
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/misc.xml b/.idea/misc.xml
new file mode 100644
index 0000000..03c5b59
--- /dev/null
+++ b/.idea/misc.xml
@@ -0,0 +1,7 @@
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/modules.xml b/.idea/modules.xml
new file mode 100644
index 0000000..235e53d
--- /dev/null
+++ b/.idea/modules.xml
@@ -0,0 +1,8 @@
+
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/synthetics_handwritten_OCR.iml b/.idea/synthetics_handwritten_OCR.iml
new file mode 100644
index 0000000..3fbeb7d
--- /dev/null
+++ b/.idea/synthetics_handwritten_OCR.iml
@@ -0,0 +1,12 @@
+
+
+
+
+
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/vcs.xml b/.idea/vcs.xml
new file mode 100644
index 0000000..35eb1dd
--- /dev/null
+++ b/.idea/vcs.xml
@@ -0,0 +1,6 @@
+
+
+
+
+
+
\ No newline at end of file
diff --git a/OCR_earsing/README.md b/OCR_earsing/README.md
new file mode 100644
index 0000000..7a33470
--- /dev/null
+++ b/OCR_earsing/README.md
@@ -0,0 +1,20 @@
+# OCR Earsing
+This repository contains the code for erasing the OCR text from the images. Code is written based on
+[OCR_SAM]("https://github.com/yeungchenwa/OCR-SAM")
+
+Have 2 method to earse the text from the image:
+- Traditional method: using the bounding box of the text to earse the text, and use function OpenCV to earse the text.
+- Deep learning method: using the bounding box of the text, segment the text based on [SAM]("https://github.com/facebookresearch/segment-anything") (Segment Anything) and use diffuse model to earse the text.
+
+In my code, I use the deep learning method to earse the text from the image.
+
+## Installation
+```shell
+pip install -r requirements.txt
+```
+
+## Run
+```shell
+python ocr_eraser.py
+```
+
diff --git a/OCR_earsing/__init__.py b/OCR_earsing/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/__init__.py b/OCR_earsing/latent_diffusion/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/inpainting_big/config.yaml b/OCR_earsing/latent_diffusion/inpainting_big/config.yaml
new file mode 100644
index 0000000..da5fd5e
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/inpainting_big/config.yaml
@@ -0,0 +1,67 @@
+model:
+ base_learning_rate: 1.0e-06
+ target: ldm.models.diffusion.ddpm.LatentDiffusion
+ params:
+ linear_start: 0.0015
+ linear_end: 0.0205
+ log_every_t: 100
+ timesteps: 1000
+ loss_type: l1
+ first_stage_key: image
+ cond_stage_key: masked_image
+ image_size: 64
+ channels: 3
+ concat_mode: true
+ monitor: val/loss
+ scheduler_config:
+ target: ldm.lr_scheduler.LambdaWarmUpCosineScheduler
+ params:
+ verbosity_interval: 0
+ warm_up_steps: 1000
+ max_decay_steps: 50000
+ lr_start: 0.001
+ lr_max: 0.1
+ lr_min: 0.0001
+ unet_config:
+ target: ldm.modules.diffusionmodules.openaimodel.UNetModel
+ params:
+ image_size: 64
+ in_channels: 7
+ out_channels: 3
+ model_channels: 256
+ attention_resolutions:
+ - 8
+ - 4
+ - 2
+ num_res_blocks: 2
+ channel_mult:
+ - 1
+ - 2
+ - 3
+ - 4
+ num_heads: 8
+ resblock_updown: true
+ first_stage_config:
+ target: ldm.models.autoencoder.VQModelInterface
+ params:
+ embed_dim: 3
+ n_embed: 8192
+ monitor: val/rec_loss
+ ddconfig:
+ attn_type: none
+ double_z: false
+ z_channels: 3
+ resolution: 256
+ in_channels: 3
+ out_ch: 3
+ ch: 128
+ ch_mult:
+ - 1
+ - 2
+ - 4
+ num_res_blocks: 2
+ attn_resolutions: []
+ dropout: 0.0
+ lossconfig:
+ target: ldm.modules.losses.contperceptual.DummyLoss
+ cond_stage_config: __is_first_stage__
diff --git a/OCR_earsing/latent_diffusion/ldm/__init__.py b/OCR_earsing/latent_diffusion/ldm/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/ldm/models/__init__.py b/OCR_earsing/latent_diffusion/ldm/models/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/ldm/models/autoencoder.py b/OCR_earsing/latent_diffusion/ldm/models/autoencoder.py
new file mode 100644
index 0000000..6a9c4f4
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/models/autoencoder.py
@@ -0,0 +1,443 @@
+import torch
+import pytorch_lightning as pl
+import torch.nn.functional as F
+from contextlib import contextmanager
+
+from taming.modules.vqvae.quantize import VectorQuantizer2 as VectorQuantizer
+
+from ldm.modules.diffusionmodules.model import Encoder, Decoder
+from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
+
+from ldm.util import instantiate_from_config
+
+
+class VQModel(pl.LightningModule):
+ def __init__(self,
+ ddconfig,
+ lossconfig,
+ n_embed,
+ embed_dim,
+ ckpt_path=None,
+ ignore_keys=[],
+ image_key="image",
+ colorize_nlabels=None,
+ monitor=None,
+ batch_resize_range=None,
+ scheduler_config=None,
+ lr_g_factor=1.0,
+ remap=None,
+ sane_index_shape=False, # tell vector quantizer to return indices as bhw
+ use_ema=False
+ ):
+ super().__init__()
+ self.embed_dim = embed_dim
+ self.n_embed = n_embed
+ self.image_key = image_key
+ self.encoder = Encoder(**ddconfig)
+ self.decoder = Decoder(**ddconfig)
+ self.loss = instantiate_from_config(lossconfig)
+ self.quantize = VectorQuantizer(n_embed, embed_dim, beta=0.25,
+ remap=remap,
+ sane_index_shape=sane_index_shape)
+ self.quant_conv = torch.nn.Conv2d(ddconfig["z_channels"], embed_dim, 1)
+ self.post_quant_conv = torch.nn.Conv2d(embed_dim, ddconfig["z_channels"], 1)
+ if colorize_nlabels is not None:
+ assert type(colorize_nlabels)==int
+ self.register_buffer("colorize", torch.randn(3, colorize_nlabels, 1, 1))
+ if monitor is not None:
+ self.monitor = monitor
+ self.batch_resize_range = batch_resize_range
+ if self.batch_resize_range is not None:
+ print(f"{self.__class__.__name__}: Using per-batch resizing in range {batch_resize_range}.")
+
+ self.use_ema = use_ema
+ if self.use_ema:
+ self.model_ema = LitEma(self)
+ print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.")
+
+ if ckpt_path is not None:
+ self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys)
+ self.scheduler_config = scheduler_config
+ self.lr_g_factor = lr_g_factor
+
+ @contextmanager
+ def ema_scope(self, context=None):
+ if self.use_ema:
+ self.model_ema.store(self.parameters())
+ self.model_ema.copy_to(self)
+ if context is not None:
+ print(f"{context}: Switched to EMA weights")
+ try:
+ yield None
+ finally:
+ if self.use_ema:
+ self.model_ema.restore(self.parameters())
+ if context is not None:
+ print(f"{context}: Restored training weights")
+
+ def init_from_ckpt(self, path, ignore_keys=list()):
+ sd = torch.load(path, map_location="cpu")["state_dict"]
+ keys = list(sd.keys())
+ for k in keys:
+ for ik in ignore_keys:
+ if k.startswith(ik):
+ print("Deleting key {} from state_dict.".format(k))
+ del sd[k]
+ missing, unexpected = self.load_state_dict(sd, strict=False)
+ print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+ if len(missing) > 0:
+ print(f"Missing Keys: {missing}")
+ print(f"Unexpected Keys: {unexpected}")
+
+ def on_train_batch_end(self, *args, **kwargs):
+ if self.use_ema:
+ self.model_ema(self)
+
+ def encode(self, x):
+ h = self.encoder(x)
+ h = self.quant_conv(h)
+ quant, emb_loss, info = self.quantize(h)
+ return quant, emb_loss, info
+
+ def encode_to_prequant(self, x):
+ h = self.encoder(x)
+ h = self.quant_conv(h)
+ return h
+
+ def decode(self, quant):
+ quant = self.post_quant_conv(quant)
+ dec = self.decoder(quant)
+ return dec
+
+ def decode_code(self, code_b):
+ quant_b = self.quantize.embed_code(code_b)
+ dec = self.decode(quant_b)
+ return dec
+
+ def forward(self, input, return_pred_indices=False):
+ quant, diff, (_,_,ind) = self.encode(input)
+ dec = self.decode(quant)
+ if return_pred_indices:
+ return dec, diff, ind
+ return dec, diff
+
+ def get_input(self, batch, k):
+ x = batch[k]
+ if len(x.shape) == 3:
+ x = x[..., None]
+ x = x.permute(0, 3, 1, 2).to(memory_format=torch.contiguous_format).float()
+ if self.batch_resize_range is not None:
+ lower_size = self.batch_resize_range[0]
+ upper_size = self.batch_resize_range[1]
+ if self.global_step <= 4:
+ # do the first few batches with max size to avoid later oom
+ new_resize = upper_size
+ else:
+ new_resize = np.random.choice(np.arange(lower_size, upper_size+16, 16))
+ if new_resize != x.shape[2]:
+ x = F.interpolate(x, size=new_resize, mode="bicubic")
+ x = x.detach()
+ return x
+
+ def training_step(self, batch, batch_idx, optimizer_idx):
+ # https://github.com/pytorch/pytorch/issues/37142
+ # try not to fool the heuristics
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss, ind = self(x, return_pred_indices=True)
+
+ if optimizer_idx == 0:
+ # autoencode
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train",
+ predicted_indices=ind)
+
+ self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return aeloss
+
+ if optimizer_idx == 1:
+ # discriminator
+ discloss, log_dict_disc = self.loss(qloss, x, xrec, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train")
+ self.log_dict(log_dict_disc, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return discloss
+
+ def validation_step(self, batch, batch_idx):
+ log_dict = self._validation_step(batch, batch_idx)
+ with self.ema_scope():
+ log_dict_ema = self._validation_step(batch, batch_idx, suffix="_ema")
+ return log_dict
+
+ def _validation_step(self, batch, batch_idx, suffix=""):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss, ind = self(x, return_pred_indices=True)
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, 0,
+ self.global_step,
+ last_layer=self.get_last_layer(),
+ split="val"+suffix,
+ predicted_indices=ind
+ )
+
+ discloss, log_dict_disc = self.loss(qloss, x, xrec, 1,
+ self.global_step,
+ last_layer=self.get_last_layer(),
+ split="val"+suffix,
+ predicted_indices=ind
+ )
+ rec_loss = log_dict_ae[f"val{suffix}/rec_loss"]
+ self.log(f"val{suffix}/rec_loss", rec_loss,
+ prog_bar=True, logger=True, on_step=False, on_epoch=True, sync_dist=True)
+ self.log(f"val{suffix}/aeloss", aeloss,
+ prog_bar=True, logger=True, on_step=False, on_epoch=True, sync_dist=True)
+ if version.parse(pl.__version__) >= version.parse('1.4.0'):
+ del log_dict_ae[f"val{suffix}/rec_loss"]
+ self.log_dict(log_dict_ae)
+ self.log_dict(log_dict_disc)
+ return self.log_dict
+
+ def configure_optimizers(self):
+ lr_d = self.learning_rate
+ lr_g = self.lr_g_factor*self.learning_rate
+ print("lr_d", lr_d)
+ print("lr_g", lr_g)
+ opt_ae = torch.optim.Adam(list(self.encoder.parameters())+
+ list(self.decoder.parameters())+
+ list(self.quantize.parameters())+
+ list(self.quant_conv.parameters())+
+ list(self.post_quant_conv.parameters()),
+ lr=lr_g, betas=(0.5, 0.9))
+ opt_disc = torch.optim.Adam(self.loss.discriminator.parameters(),
+ lr=lr_d, betas=(0.5, 0.9))
+
+ if self.scheduler_config is not None:
+ scheduler = instantiate_from_config(self.scheduler_config)
+
+ print("Setting up LambdaLR scheduler...")
+ scheduler = [
+ {
+ 'scheduler': LambdaLR(opt_ae, lr_lambda=scheduler.schedule),
+ 'interval': 'step',
+ 'frequency': 1
+ },
+ {
+ 'scheduler': LambdaLR(opt_disc, lr_lambda=scheduler.schedule),
+ 'interval': 'step',
+ 'frequency': 1
+ },
+ ]
+ return [opt_ae, opt_disc], scheduler
+ return [opt_ae, opt_disc], []
+
+ def get_last_layer(self):
+ return self.decoder.conv_out.weight
+
+ def log_images(self, batch, only_inputs=False, plot_ema=False, **kwargs):
+ log = dict()
+ x = self.get_input(batch, self.image_key)
+ x = x.to(self.device)
+ if only_inputs:
+ log["inputs"] = x
+ return log
+ xrec, _ = self(x)
+ if x.shape[1] > 3:
+ # colorize with random projection
+ assert xrec.shape[1] > 3
+ x = self.to_rgb(x)
+ xrec = self.to_rgb(xrec)
+ log["inputs"] = x
+ log["reconstructions"] = xrec
+ if plot_ema:
+ with self.ema_scope():
+ xrec_ema, _ = self(x)
+ if x.shape[1] > 3: xrec_ema = self.to_rgb(xrec_ema)
+ log["reconstructions_ema"] = xrec_ema
+ return log
+
+ def to_rgb(self, x):
+ assert self.image_key == "segmentation"
+ if not hasattr(self, "colorize"):
+ self.register_buffer("colorize", torch.randn(3, x.shape[1], 1, 1).to(x))
+ x = F.conv2d(x, weight=self.colorize)
+ x = 2.*(x-x.min())/(x.max()-x.min()) - 1.
+ return x
+
+
+class VQModelInterface(VQModel):
+ def __init__(self, embed_dim, *args, **kwargs):
+ super().__init__(embed_dim=embed_dim, *args, **kwargs)
+ self.embed_dim = embed_dim
+
+ def encode(self, x):
+ h = self.encoder(x)
+ h = self.quant_conv(h)
+ return h
+
+ def decode(self, h, force_not_quantize=False):
+ # also go through quantization layer
+ if not force_not_quantize:
+ quant, emb_loss, info = self.quantize(h)
+ else:
+ quant = h
+ quant = self.post_quant_conv(quant)
+ dec = self.decoder(quant)
+ return dec
+
+
+class AutoencoderKL(pl.LightningModule):
+ def __init__(self,
+ ddconfig,
+ lossconfig,
+ embed_dim,
+ ckpt_path=None,
+ ignore_keys=[],
+ image_key="image",
+ colorize_nlabels=None,
+ monitor=None,
+ ):
+ super().__init__()
+ self.image_key = image_key
+ self.encoder = Encoder(**ddconfig)
+ self.decoder = Decoder(**ddconfig)
+ self.loss = instantiate_from_config(lossconfig)
+ assert ddconfig["double_z"]
+ self.quant_conv = torch.nn.Conv2d(2*ddconfig["z_channels"], 2*embed_dim, 1)
+ self.post_quant_conv = torch.nn.Conv2d(embed_dim, ddconfig["z_channels"], 1)
+ self.embed_dim = embed_dim
+ if colorize_nlabels is not None:
+ assert type(colorize_nlabels)==int
+ self.register_buffer("colorize", torch.randn(3, colorize_nlabels, 1, 1))
+ if monitor is not None:
+ self.monitor = monitor
+ if ckpt_path is not None:
+ self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys)
+
+ def init_from_ckpt(self, path, ignore_keys=list()):
+ sd = torch.load(path, map_location="cpu")["state_dict"]
+ keys = list(sd.keys())
+ for k in keys:
+ for ik in ignore_keys:
+ if k.startswith(ik):
+ print("Deleting key {} from state_dict.".format(k))
+ del sd[k]
+ self.load_state_dict(sd, strict=False)
+ print(f"Restored from {path}")
+
+ def encode(self, x):
+ h = self.encoder(x)
+ moments = self.quant_conv(h)
+ posterior = DiagonalGaussianDistribution(moments)
+ return posterior
+
+ def decode(self, z):
+ z = self.post_quant_conv(z)
+ dec = self.decoder(z)
+ return dec
+
+ def forward(self, input, sample_posterior=True):
+ posterior = self.encode(input)
+ if sample_posterior:
+ z = posterior.sample()
+ else:
+ z = posterior.mode()
+ dec = self.decode(z)
+ return dec, posterior
+
+ def get_input(self, batch, k):
+ x = batch[k]
+ if len(x.shape) == 3:
+ x = x[..., None]
+ x = x.permute(0, 3, 1, 2).to(memory_format=torch.contiguous_format).float()
+ return x
+
+ def training_step(self, batch, batch_idx, optimizer_idx):
+ inputs = self.get_input(batch, self.image_key)
+ reconstructions, posterior = self(inputs)
+
+ if optimizer_idx == 0:
+ # train encoder+decoder+logvar
+ aeloss, log_dict_ae = self.loss(inputs, reconstructions, posterior, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train")
+ self.log("aeloss", aeloss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=False)
+ return aeloss
+
+ if optimizer_idx == 1:
+ # train the discriminator
+ discloss, log_dict_disc = self.loss(inputs, reconstructions, posterior, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train")
+
+ self.log("discloss", discloss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ self.log_dict(log_dict_disc, prog_bar=False, logger=True, on_step=True, on_epoch=False)
+ return discloss
+
+ def validation_step(self, batch, batch_idx):
+ inputs = self.get_input(batch, self.image_key)
+ reconstructions, posterior = self(inputs)
+ aeloss, log_dict_ae = self.loss(inputs, reconstructions, posterior, 0, self.global_step,
+ last_layer=self.get_last_layer(), split="val")
+
+ discloss, log_dict_disc = self.loss(inputs, reconstructions, posterior, 1, self.global_step,
+ last_layer=self.get_last_layer(), split="val")
+
+ self.log("val/rec_loss", log_dict_ae["val/rec_loss"])
+ self.log_dict(log_dict_ae)
+ self.log_dict(log_dict_disc)
+ return self.log_dict
+
+ def configure_optimizers(self):
+ lr = self.learning_rate
+ opt_ae = torch.optim.Adam(list(self.encoder.parameters())+
+ list(self.decoder.parameters())+
+ list(self.quant_conv.parameters())+
+ list(self.post_quant_conv.parameters()),
+ lr=lr, betas=(0.5, 0.9))
+ opt_disc = torch.optim.Adam(self.loss.discriminator.parameters(),
+ lr=lr, betas=(0.5, 0.9))
+ return [opt_ae, opt_disc], []
+
+ def get_last_layer(self):
+ return self.decoder.conv_out.weight
+
+ @torch.no_grad()
+ def log_images(self, batch, only_inputs=False, **kwargs):
+ log = dict()
+ x = self.get_input(batch, self.image_key)
+ x = x.to(self.device)
+ if not only_inputs:
+ xrec, posterior = self(x)
+ if x.shape[1] > 3:
+ # colorize with random projection
+ assert xrec.shape[1] > 3
+ x = self.to_rgb(x)
+ xrec = self.to_rgb(xrec)
+ log["samples"] = self.decode(torch.randn_like(posterior.sample()))
+ log["reconstructions"] = xrec
+ log["inputs"] = x
+ return log
+
+ def to_rgb(self, x):
+ assert self.image_key == "segmentation"
+ if not hasattr(self, "colorize"):
+ self.register_buffer("colorize", torch.randn(3, x.shape[1], 1, 1).to(x))
+ x = F.conv2d(x, weight=self.colorize)
+ x = 2.*(x-x.min())/(x.max()-x.min()) - 1.
+ return x
+
+
+class IdentityFirstStage(torch.nn.Module):
+ def __init__(self, *args, vq_interface=False, **kwargs):
+ self.vq_interface = vq_interface # TODO: Should be true by default but check to not break older stuff
+ super().__init__()
+
+ def encode(self, x, *args, **kwargs):
+ return x
+
+ def decode(self, x, *args, **kwargs):
+ return x
+
+ def quantize(self, x, *args, **kwargs):
+ if self.vq_interface:
+ return x, None, [None, None, None]
+ return x
+
+ def forward(self, x, *args, **kwargs):
+ return x
diff --git a/OCR_earsing/latent_diffusion/ldm/models/diffusion/__init__.py b/OCR_earsing/latent_diffusion/ldm/models/diffusion/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/ldm/models/diffusion/ddim.py b/OCR_earsing/latent_diffusion/ldm/models/diffusion/ddim.py
new file mode 100644
index 0000000..edf1eaf
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/models/diffusion/ddim.py
@@ -0,0 +1,203 @@
+"""SAMPLING ONLY."""
+
+import torch
+import numpy as np
+from tqdm import tqdm
+from functools import partial
+
+from ldm.modules.diffusionmodules.util import make_ddim_sampling_parameters, make_ddim_timesteps, noise_like
+
+
+class DDIMSampler(object):
+ def __init__(self, model, schedule="linear", **kwargs):
+ super().__init__()
+ self.model = model
+ self.ddpm_num_timesteps = model.num_timesteps
+ self.schedule = schedule
+
+ def register_buffer(self, name, attr):
+ if type(attr) == torch.Tensor:
+ if attr.device != torch.device("cuda"):
+ attr = attr.to(torch.device("cuda"))
+ setattr(self, name, attr)
+
+ def make_schedule(self, ddim_num_steps, ddim_discretize="uniform", ddim_eta=0., verbose=True):
+ self.ddim_timesteps = make_ddim_timesteps(ddim_discr_method=ddim_discretize, num_ddim_timesteps=ddim_num_steps,
+ num_ddpm_timesteps=self.ddpm_num_timesteps,verbose=verbose)
+ alphas_cumprod = self.model.alphas_cumprod
+ assert alphas_cumprod.shape[0] == self.ddpm_num_timesteps, 'alphas have to be defined for each timestep'
+ to_torch = lambda x: x.clone().detach().to(torch.float32).to(self.model.device)
+
+ self.register_buffer('betas', to_torch(self.model.betas))
+ self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+ self.register_buffer('alphas_cumprod_prev', to_torch(self.model.alphas_cumprod_prev))
+
+ # calculations for diffusion q(x_t | x_{t-1}) and others
+ self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod.cpu())))
+ self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod.cpu())))
+ self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod.cpu())))
+ self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod.cpu())))
+ self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod.cpu() - 1)))
+
+ # ddim sampling parameters
+ ddim_sigmas, ddim_alphas, ddim_alphas_prev = make_ddim_sampling_parameters(alphacums=alphas_cumprod.cpu(),
+ ddim_timesteps=self.ddim_timesteps,
+ eta=ddim_eta,verbose=verbose)
+ self.register_buffer('ddim_sigmas', ddim_sigmas)
+ self.register_buffer('ddim_alphas', ddim_alphas)
+ self.register_buffer('ddim_alphas_prev', ddim_alphas_prev)
+ self.register_buffer('ddim_sqrt_one_minus_alphas', np.sqrt(1. - ddim_alphas))
+ sigmas_for_original_sampling_steps = ddim_eta * torch.sqrt(
+ (1 - self.alphas_cumprod_prev) / (1 - self.alphas_cumprod) * (
+ 1 - self.alphas_cumprod / self.alphas_cumprod_prev))
+ self.register_buffer('ddim_sigmas_for_original_num_steps', sigmas_for_original_sampling_steps)
+
+ @torch.no_grad()
+ def sample(self,
+ S,
+ batch_size,
+ shape,
+ conditioning=None,
+ callback=None,
+ normals_sequence=None,
+ img_callback=None,
+ quantize_x0=False,
+ eta=0.,
+ mask=None,
+ x0=None,
+ temperature=1.,
+ noise_dropout=0.,
+ score_corrector=None,
+ corrector_kwargs=None,
+ verbose=True,
+ x_T=None,
+ log_every_t=100,
+ unconditional_guidance_scale=1.,
+ unconditional_conditioning=None,
+ # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ...
+ **kwargs
+ ):
+ if conditioning is not None:
+ if isinstance(conditioning, dict):
+ cbs = conditioning[list(conditioning.keys())[0]].shape[0]
+ if cbs != batch_size:
+ print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}")
+ else:
+ if conditioning.shape[0] != batch_size:
+ print(f"Warning: Got {conditioning.shape[0]} conditionings but batch-size is {batch_size}")
+
+ self.make_schedule(ddim_num_steps=S, ddim_eta=eta, verbose=verbose)
+ # sampling
+ C, H, W = shape
+ size = (batch_size, C, H, W)
+ print(f'Data shape for DDIM sampling is {size}, eta {eta}')
+
+ samples, intermediates = self.ddim_sampling(conditioning, size,
+ callback=callback,
+ img_callback=img_callback,
+ quantize_denoised=quantize_x0,
+ mask=mask, x0=x0,
+ ddim_use_original_steps=False,
+ noise_dropout=noise_dropout,
+ temperature=temperature,
+ score_corrector=score_corrector,
+ corrector_kwargs=corrector_kwargs,
+ x_T=x_T,
+ log_every_t=log_every_t,
+ unconditional_guidance_scale=unconditional_guidance_scale,
+ unconditional_conditioning=unconditional_conditioning,
+ )
+ return samples, intermediates
+
+ @torch.no_grad()
+ def ddim_sampling(self, cond, shape,
+ x_T=None, ddim_use_original_steps=False,
+ callback=None, timesteps=None, quantize_denoised=False,
+ mask=None, x0=None, img_callback=None, log_every_t=100,
+ temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None,
+ unconditional_guidance_scale=1., unconditional_conditioning=None,):
+ device = self.model.betas.device
+ b = shape[0]
+ if x_T is None:
+ img = torch.randn(shape, device=device)
+ else:
+ img = x_T
+
+ if timesteps is None:
+ timesteps = self.ddpm_num_timesteps if ddim_use_original_steps else self.ddim_timesteps
+ elif timesteps is not None and not ddim_use_original_steps:
+ subset_end = int(min(timesteps / self.ddim_timesteps.shape[0], 1) * self.ddim_timesteps.shape[0]) - 1
+ timesteps = self.ddim_timesteps[:subset_end]
+
+ intermediates = {'x_inter': [img], 'pred_x0': [img]}
+ time_range = reversed(range(0,timesteps)) if ddim_use_original_steps else np.flip(timesteps)
+ total_steps = timesteps if ddim_use_original_steps else timesteps.shape[0]
+ print(f"Running DDIM Sampling with {total_steps} timesteps")
+
+ iterator = tqdm(time_range, desc='DDIM Sampler', total=total_steps)
+
+ for i, step in enumerate(iterator):
+ index = total_steps - i - 1
+ ts = torch.full((b,), step, device=device, dtype=torch.long)
+
+ if mask is not None:
+ assert x0 is not None
+ img_orig = self.model.q_sample(x0, ts) # TODO: deterministic forward pass?
+ img = img_orig * mask + (1. - mask) * img
+
+ outs = self.p_sample_ddim(img, cond, ts, index=index, use_original_steps=ddim_use_original_steps,
+ quantize_denoised=quantize_denoised, temperature=temperature,
+ noise_dropout=noise_dropout, score_corrector=score_corrector,
+ corrector_kwargs=corrector_kwargs,
+ unconditional_guidance_scale=unconditional_guidance_scale,
+ unconditional_conditioning=unconditional_conditioning)
+ img, pred_x0 = outs
+ if callback: callback(i)
+ if img_callback: img_callback(pred_x0, i)
+
+ if index % log_every_t == 0 or index == total_steps - 1:
+ intermediates['x_inter'].append(img)
+ intermediates['pred_x0'].append(pred_x0)
+
+ return img, intermediates
+
+ @torch.no_grad()
+ def p_sample_ddim(self, x, c, t, index, repeat_noise=False, use_original_steps=False, quantize_denoised=False,
+ temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None,
+ unconditional_guidance_scale=1., unconditional_conditioning=None):
+ b, *_, device = *x.shape, x.device
+
+ if unconditional_conditioning is None or unconditional_guidance_scale == 1.:
+ e_t = self.model.apply_model(x, t, c)
+ else:
+ x_in = torch.cat([x] * 2)
+ t_in = torch.cat([t] * 2)
+ c_in = torch.cat([unconditional_conditioning, c])
+ e_t_uncond, e_t = self.model.apply_model(x_in, t_in, c_in).chunk(2)
+ e_t = e_t_uncond + unconditional_guidance_scale * (e_t - e_t_uncond)
+
+ if score_corrector is not None:
+ assert self.model.parameterization == "eps"
+ e_t = score_corrector.modify_score(self.model, e_t, x, t, c, **corrector_kwargs)
+
+ alphas = self.model.alphas_cumprod if use_original_steps else self.ddim_alphas
+ alphas_prev = self.model.alphas_cumprod_prev if use_original_steps else self.ddim_alphas_prev
+ sqrt_one_minus_alphas = self.model.sqrt_one_minus_alphas_cumprod if use_original_steps else self.ddim_sqrt_one_minus_alphas
+ sigmas = self.model.ddim_sigmas_for_original_num_steps if use_original_steps else self.ddim_sigmas
+ # select parameters corresponding to the currently considered timestep
+ a_t = torch.full((b, 1, 1, 1), alphas[index], device=device)
+ a_prev = torch.full((b, 1, 1, 1), alphas_prev[index], device=device)
+ sigma_t = torch.full((b, 1, 1, 1), sigmas[index], device=device)
+ sqrt_one_minus_at = torch.full((b, 1, 1, 1), sqrt_one_minus_alphas[index],device=device)
+
+ # current prediction for x_0
+ pred_x0 = (x - sqrt_one_minus_at * e_t) / a_t.sqrt()
+ if quantize_denoised:
+ pred_x0, _, *_ = self.model.first_stage_model.quantize(pred_x0)
+ # direction pointing to x_t
+ dir_xt = (1. - a_prev - sigma_t**2).sqrt() * e_t
+ noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature
+ if noise_dropout > 0.:
+ noise = torch.nn.functional.dropout(noise, p=noise_dropout)
+ x_prev = a_prev.sqrt() * pred_x0 + dir_xt + noise
+ return x_prev, pred_x0
diff --git a/OCR_earsing/latent_diffusion/ldm/models/diffusion/ddpm.py b/OCR_earsing/latent_diffusion/ldm/models/diffusion/ddpm.py
new file mode 100644
index 0000000..ef0990e
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/models/diffusion/ddpm.py
@@ -0,0 +1,1445 @@
+"""
+wild mixture of
+https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+https://github.com/openai/improved-diffusion/blob/e94489283bb876ac1477d5dd7709bbbd2d9902ce/improved_diffusion/gaussian_diffusion.py
+https://github.com/CompVis/taming-transformers
+-- merci
+"""
+
+import torch
+import torch.nn as nn
+import numpy as np
+import pytorch_lightning as pl
+from torch.optim.lr_scheduler import LambdaLR
+from einops import rearrange, repeat
+from contextlib import contextmanager
+from functools import partial
+from tqdm import tqdm
+from torchvision.utils import make_grid
+from pytorch_lightning.utilities.rank_zero import rank_zero_only
+
+from ldm.util import log_txt_as_img, exists, default, ismap, isimage, mean_flat, count_params, instantiate_from_config
+from ldm.modules.ema import LitEma
+from ldm.modules.distributions.distributions import normal_kl, DiagonalGaussianDistribution
+from ldm.models.autoencoder import VQModelInterface, IdentityFirstStage, AutoencoderKL
+from ldm.modules.diffusionmodules.util import make_beta_schedule, extract_into_tensor, noise_like
+from ldm.models.diffusion.ddim import DDIMSampler
+
+
+__conditioning_keys__ = {'concat': 'c_concat',
+ 'crossattn': 'c_crossattn',
+ 'adm': 'y'}
+
+
+def disabled_train(self, mode=True):
+ """Overwrite model.train with this function to make sure train/eval mode
+ does not change anymore."""
+ return self
+
+
+def uniform_on_device(r1, r2, shape, device):
+ return (r1 - r2) * torch.rand(*shape, device=device) + r2
+
+
+class DDPM(pl.LightningModule):
+ # classic DDPM with Gaussian diffusion, in image space
+ def __init__(self,
+ unet_config,
+ timesteps=1000,
+ beta_schedule="linear",
+ loss_type="l2",
+ ckpt_path=None,
+ ignore_keys=[],
+ load_only_unet=False,
+ monitor="val/loss",
+ use_ema=True,
+ first_stage_key="image",
+ image_size=256,
+ channels=3,
+ log_every_t=100,
+ clip_denoised=True,
+ linear_start=1e-4,
+ linear_end=2e-2,
+ cosine_s=8e-3,
+ given_betas=None,
+ original_elbo_weight=0.,
+ v_posterior=0., # weight for choosing posterior variance as sigma = (1-v) * beta_tilde + v * beta
+ l_simple_weight=1.,
+ conditioning_key=None,
+ parameterization="eps", # all assuming fixed variance schedules
+ scheduler_config=None,
+ use_positional_encodings=False,
+ learn_logvar=False,
+ logvar_init=0.,
+ ):
+ super().__init__()
+ assert parameterization in ["eps", "x0"], 'currently only supporting "eps" and "x0"'
+ self.parameterization = parameterization
+ print(f"{self.__class__.__name__}: Running in {self.parameterization}-prediction mode")
+ self.cond_stage_model = None
+ self.clip_denoised = clip_denoised
+ self.log_every_t = log_every_t
+ self.first_stage_key = first_stage_key
+ self.image_size = image_size # try conv?
+ self.channels = channels
+ self.use_positional_encodings = use_positional_encodings
+ self.model = DiffusionWrapper(unet_config, conditioning_key)
+ count_params(self.model, verbose=True)
+ self.use_ema = use_ema
+ if self.use_ema:
+ self.model_ema = LitEma(self.model)
+ print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.")
+
+ self.use_scheduler = scheduler_config is not None
+ if self.use_scheduler:
+ self.scheduler_config = scheduler_config
+
+ self.v_posterior = v_posterior
+ self.original_elbo_weight = original_elbo_weight
+ self.l_simple_weight = l_simple_weight
+
+ if monitor is not None:
+ self.monitor = monitor
+ if ckpt_path is not None:
+ self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys, only_model=load_only_unet)
+
+ self.register_schedule(given_betas=given_betas, beta_schedule=beta_schedule, timesteps=timesteps,
+ linear_start=linear_start, linear_end=linear_end, cosine_s=cosine_s)
+
+ self.loss_type = loss_type
+
+ self.learn_logvar = learn_logvar
+ self.logvar = torch.full(fill_value=logvar_init, size=(self.num_timesteps,))
+ if self.learn_logvar:
+ self.logvar = nn.Parameter(self.logvar, requires_grad=True)
+
+
+ def register_schedule(self, given_betas=None, beta_schedule="linear", timesteps=1000,
+ linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+ if exists(given_betas):
+ betas = given_betas
+ else:
+ betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end,
+ cosine_s=cosine_s)
+ alphas = 1. - betas
+ alphas_cumprod = np.cumprod(alphas, axis=0)
+ alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
+
+ timesteps, = betas.shape
+ self.num_timesteps = int(timesteps)
+ self.linear_start = linear_start
+ self.linear_end = linear_end
+ assert alphas_cumprod.shape[0] == self.num_timesteps, 'alphas have to be defined for each timestep'
+
+ to_torch = partial(torch.tensor, dtype=torch.float32)
+
+ self.register_buffer('betas', to_torch(betas))
+ self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+ self.register_buffer('alphas_cumprod_prev', to_torch(alphas_cumprod_prev))
+
+ # calculations for diffusion q(x_t | x_{t-1}) and others
+ self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+ self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+ self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod)))
+ self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod)))
+ self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod - 1)))
+
+ # calculations for posterior q(x_{t-1} | x_t, x_0)
+ posterior_variance = (1 - self.v_posterior) * betas * (1. - alphas_cumprod_prev) / (
+ 1. - alphas_cumprod) + self.v_posterior * betas
+ # above: equal to 1. / (1. / (1. - alpha_cumprod_tm1) + alpha_t / beta_t)
+ self.register_buffer('posterior_variance', to_torch(posterior_variance))
+ # below: log calculation clipped because the posterior variance is 0 at the beginning of the diffusion chain
+ self.register_buffer('posterior_log_variance_clipped', to_torch(np.log(np.maximum(posterior_variance, 1e-20))))
+ self.register_buffer('posterior_mean_coef1', to_torch(
+ betas * np.sqrt(alphas_cumprod_prev) / (1. - alphas_cumprod)))
+ self.register_buffer('posterior_mean_coef2', to_torch(
+ (1. - alphas_cumprod_prev) * np.sqrt(alphas) / (1. - alphas_cumprod)))
+
+ if self.parameterization == "eps":
+ lvlb_weights = self.betas ** 2 / (
+ 2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod))
+ elif self.parameterization == "x0":
+ lvlb_weights = 0.5 * np.sqrt(torch.Tensor(alphas_cumprod)) / (2. * 1 - torch.Tensor(alphas_cumprod))
+ else:
+ raise NotImplementedError("mu not supported")
+ # TODO how to choose this term
+ lvlb_weights[0] = lvlb_weights[1]
+ self.register_buffer('lvlb_weights', lvlb_weights, persistent=False)
+ assert not torch.isnan(self.lvlb_weights).all()
+
+ @contextmanager
+ def ema_scope(self, context=None):
+ if self.use_ema:
+ self.model_ema.store(self.model.parameters())
+ self.model_ema.copy_to(self.model)
+ if context is not None:
+ print(f"{context}: Switched to EMA weights")
+ try:
+ yield None
+ finally:
+ if self.use_ema:
+ self.model_ema.restore(self.model.parameters())
+ if context is not None:
+ print(f"{context}: Restored training weights")
+
+ def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+ sd = torch.load(path, map_location="cpu")
+ if "state_dict" in list(sd.keys()):
+ sd = sd["state_dict"]
+ keys = list(sd.keys())
+ for k in keys:
+ for ik in ignore_keys:
+ if k.startswith(ik):
+ print("Deleting key {} from state_dict.".format(k))
+ del sd[k]
+ missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict(
+ sd, strict=False)
+ print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+ if len(missing) > 0:
+ print(f"Missing Keys: {missing}")
+ if len(unexpected) > 0:
+ print(f"Unexpected Keys: {unexpected}")
+
+ def q_mean_variance(self, x_start, t):
+ """
+ Get the distribution q(x_t | x_0).
+ :param x_start: the [N x C x ...] tensor of noiseless inputs.
+ :param t: the number of diffusion steps (minus 1). Here, 0 means one step.
+ :return: A tuple (mean, variance, log_variance), all of x_start's shape.
+ """
+ mean = (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start)
+ variance = extract_into_tensor(1.0 - self.alphas_cumprod, t, x_start.shape)
+ log_variance = extract_into_tensor(self.log_one_minus_alphas_cumprod, t, x_start.shape)
+ return mean, variance, log_variance
+
+ def predict_start_from_noise(self, x_t, t, noise):
+ return (
+ extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t -
+ extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * noise
+ )
+
+ def q_posterior(self, x_start, x_t, t):
+ posterior_mean = (
+ extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start +
+ extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t
+ )
+ posterior_variance = extract_into_tensor(self.posterior_variance, t, x_t.shape)
+ posterior_log_variance_clipped = extract_into_tensor(self.posterior_log_variance_clipped, t, x_t.shape)
+ return posterior_mean, posterior_variance, posterior_log_variance_clipped
+
+ def p_mean_variance(self, x, t, clip_denoised: bool):
+ model_out = self.model(x, t)
+ if self.parameterization == "eps":
+ x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+ elif self.parameterization == "x0":
+ x_recon = model_out
+ if clip_denoised:
+ x_recon.clamp_(-1., 1.)
+
+ model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+ return model_mean, posterior_variance, posterior_log_variance
+
+ @torch.no_grad()
+ def p_sample(self, x, t, clip_denoised=True, repeat_noise=False):
+ b, *_, device = *x.shape, x.device
+ model_mean, _, model_log_variance = self.p_mean_variance(x=x, t=t, clip_denoised=clip_denoised)
+ noise = noise_like(x.shape, device, repeat_noise)
+ # no noise when t == 0
+ nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+ return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+ @torch.no_grad()
+ def p_sample_loop(self, shape, return_intermediates=False):
+ device = self.betas.device
+ b = shape[0]
+ img = torch.randn(shape, device=device)
+ intermediates = [img]
+ for i in tqdm(reversed(range(0, self.num_timesteps)), desc='Sampling t', total=self.num_timesteps):
+ img = self.p_sample(img, torch.full((b,), i, device=device, dtype=torch.long),
+ clip_denoised=self.clip_denoised)
+ if i % self.log_every_t == 0 or i == self.num_timesteps - 1:
+ intermediates.append(img)
+ if return_intermediates:
+ return img, intermediates
+ return img
+
+ @torch.no_grad()
+ def sample(self, batch_size=16, return_intermediates=False):
+ image_size = self.image_size
+ channels = self.channels
+ return self.p_sample_loop((batch_size, channels, image_size, image_size),
+ return_intermediates=return_intermediates)
+
+ def q_sample(self, x_start, t, noise=None):
+ noise = default(noise, lambda: torch.randn_like(x_start))
+ return (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start +
+ extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise)
+
+ def get_loss(self, pred, target, mean=True):
+ if self.loss_type == 'l1':
+ loss = (target - pred).abs()
+ if mean:
+ loss = loss.mean()
+ elif self.loss_type == 'l2':
+ if mean:
+ loss = torch.nn.functional.mse_loss(target, pred)
+ else:
+ loss = torch.nn.functional.mse_loss(target, pred, reduction='none')
+ else:
+ raise NotImplementedError("unknown loss type '{loss_type}'")
+
+ return loss
+
+ def p_losses(self, x_start, t, noise=None):
+ noise = default(noise, lambda: torch.randn_like(x_start))
+ x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+ model_out = self.model(x_noisy, t)
+
+ loss_dict = {}
+ if self.parameterization == "eps":
+ target = noise
+ elif self.parameterization == "x0":
+ target = x_start
+ else:
+ raise NotImplementedError(f"Paramterization {self.parameterization} not yet supported")
+
+ loss = self.get_loss(model_out, target, mean=False).mean(dim=[1, 2, 3])
+
+ log_prefix = 'train' if self.training else 'val'
+
+ loss_dict.update({f'{log_prefix}/loss_simple': loss.mean()})
+ loss_simple = loss.mean() * self.l_simple_weight
+
+ loss_vlb = (self.lvlb_weights[t] * loss).mean()
+ loss_dict.update({f'{log_prefix}/loss_vlb': loss_vlb})
+
+ loss = loss_simple + self.original_elbo_weight * loss_vlb
+
+ loss_dict.update({f'{log_prefix}/loss': loss})
+
+ return loss, loss_dict
+
+ def forward(self, x, *args, **kwargs):
+ # b, c, h, w, device, img_size, = *x.shape, x.device, self.image_size
+ # assert h == img_size and w == img_size, f'height and width of image must be {img_size}'
+ t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long()
+ return self.p_losses(x, t, *args, **kwargs)
+
+ def get_input(self, batch, k):
+ x = batch[k]
+ if len(x.shape) == 3:
+ x = x[..., None]
+ x = rearrange(x, 'b h w c -> b c h w')
+ x = x.to(memory_format=torch.contiguous_format).float()
+ return x
+
+ def shared_step(self, batch):
+ x = self.get_input(batch, self.first_stage_key)
+ loss, loss_dict = self(x)
+ return loss, loss_dict
+
+ def training_step(self, batch, batch_idx):
+ loss, loss_dict = self.shared_step(batch)
+
+ self.log_dict(loss_dict, prog_bar=True,
+ logger=True, on_step=True, on_epoch=True)
+
+ self.log("global_step", self.global_step,
+ prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+ if self.use_scheduler:
+ lr = self.optimizers().param_groups[0]['lr']
+ self.log('lr_abs', lr, prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+ return loss
+
+ @torch.no_grad()
+ def validation_step(self, batch, batch_idx):
+ _, loss_dict_no_ema = self.shared_step(batch)
+ with self.ema_scope():
+ _, loss_dict_ema = self.shared_step(batch)
+ loss_dict_ema = {key + '_ema': loss_dict_ema[key] for key in loss_dict_ema}
+ self.log_dict(loss_dict_no_ema, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+ self.log_dict(loss_dict_ema, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+
+ def on_train_batch_end(self, *args, **kwargs):
+ if self.use_ema:
+ self.model_ema(self.model)
+
+ def _get_rows_from_list(self, samples):
+ n_imgs_per_row = len(samples)
+ denoise_grid = rearrange(samples, 'n b c h w -> b n c h w')
+ denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+ denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+ return denoise_grid
+
+ @torch.no_grad()
+ def log_images(self, batch, N=8, n_row=2, sample=True, return_keys=None, **kwargs):
+ log = dict()
+ x = self.get_input(batch, self.first_stage_key)
+ N = min(x.shape[0], N)
+ n_row = min(x.shape[0], n_row)
+ x = x.to(self.device)[:N]
+ log["inputs"] = x
+
+ # get diffusion row
+ diffusion_row = list()
+ x_start = x[:n_row]
+
+ for t in range(self.num_timesteps):
+ if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+ t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+ t = t.to(self.device).long()
+ noise = torch.randn_like(x_start)
+ x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+ diffusion_row.append(x_noisy)
+
+ log["diffusion_row"] = self._get_rows_from_list(diffusion_row)
+
+ if sample:
+ # get denoise row
+ with self.ema_scope("Plotting"):
+ samples, denoise_row = self.sample(batch_size=N, return_intermediates=True)
+
+ log["samples"] = samples
+ log["denoise_row"] = self._get_rows_from_list(denoise_row)
+
+ if return_keys:
+ if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+ return log
+ else:
+ return {key: log[key] for key in return_keys}
+ return log
+
+ def configure_optimizers(self):
+ lr = self.learning_rate
+ params = list(self.model.parameters())
+ if self.learn_logvar:
+ params = params + [self.logvar]
+ opt = torch.optim.AdamW(params, lr=lr)
+ return opt
+
+
+class LatentDiffusion(DDPM):
+ """main class"""
+ def __init__(self,
+ first_stage_config,
+ cond_stage_config,
+ num_timesteps_cond=None,
+ cond_stage_key="image",
+ cond_stage_trainable=False,
+ concat_mode=True,
+ cond_stage_forward=None,
+ conditioning_key=None,
+ scale_factor=1.0,
+ scale_by_std=False,
+ *args, **kwargs):
+ self.num_timesteps_cond = default(num_timesteps_cond, 1)
+ self.scale_by_std = scale_by_std
+ assert self.num_timesteps_cond <= kwargs['timesteps']
+ # for backwards compatibility after implementation of DiffusionWrapper
+ if conditioning_key is None:
+ conditioning_key = 'concat' if concat_mode else 'crossattn'
+ if cond_stage_config == '__is_unconditional__':
+ conditioning_key = None
+ ckpt_path = kwargs.pop("ckpt_path", None)
+ ignore_keys = kwargs.pop("ignore_keys", [])
+ super().__init__(conditioning_key=conditioning_key, *args, **kwargs)
+ self.concat_mode = concat_mode
+ self.cond_stage_trainable = cond_stage_trainable
+ self.cond_stage_key = cond_stage_key
+ try:
+ self.num_downs = len(first_stage_config.params.ddconfig.ch_mult) - 1
+ except:
+ self.num_downs = 0
+ if not scale_by_std:
+ self.scale_factor = scale_factor
+ else:
+ self.register_buffer('scale_factor', torch.tensor(scale_factor))
+ self.instantiate_first_stage(first_stage_config)
+ self.instantiate_cond_stage(cond_stage_config)
+ self.cond_stage_forward = cond_stage_forward
+ self.clip_denoised = False
+ self.bbox_tokenizer = None
+
+ self.restarted_from_ckpt = False
+ if ckpt_path is not None:
+ self.init_from_ckpt(ckpt_path, ignore_keys)
+ self.restarted_from_ckpt = True
+
+ def make_cond_schedule(self, ):
+ self.cond_ids = torch.full(size=(self.num_timesteps,), fill_value=self.num_timesteps - 1, dtype=torch.long)
+ ids = torch.round(torch.linspace(0, self.num_timesteps - 1, self.num_timesteps_cond)).long()
+ self.cond_ids[:self.num_timesteps_cond] = ids
+
+ @rank_zero_only
+ @torch.no_grad()
+ def on_train_batch_start(self, batch, batch_idx, dataloader_idx):
+ # only for very first batch
+ if self.scale_by_std and self.current_epoch == 0 and self.global_step == 0 and batch_idx == 0 and not self.restarted_from_ckpt:
+ assert self.scale_factor == 1., 'rather not use custom rescaling and std-rescaling simultaneously'
+ # set rescale weight to 1./std of encodings
+ print("### USING STD-RESCALING ###")
+ x = super().get_input(batch, self.first_stage_key)
+ x = x.to(self.device)
+ encoder_posterior = self.encode_first_stage(x)
+ z = self.get_first_stage_encoding(encoder_posterior).detach()
+ del self.scale_factor
+ self.register_buffer('scale_factor', 1. / z.flatten().std())
+ print(f"setting self.scale_factor to {self.scale_factor}")
+ print("### USING STD-RESCALING ###")
+
+ def register_schedule(self,
+ given_betas=None, beta_schedule="linear", timesteps=1000,
+ linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+ super().register_schedule(given_betas, beta_schedule, timesteps, linear_start, linear_end, cosine_s)
+
+ self.shorten_cond_schedule = self.num_timesteps_cond > 1
+ if self.shorten_cond_schedule:
+ self.make_cond_schedule()
+
+ def instantiate_first_stage(self, config):
+ model = instantiate_from_config(config)
+ self.first_stage_model = model.eval()
+ self.first_stage_model.train = disabled_train
+ for param in self.first_stage_model.parameters():
+ param.requires_grad = False
+
+ def instantiate_cond_stage(self, config):
+ if not self.cond_stage_trainable:
+ if config == "__is_first_stage__":
+ print("Using first stage also as cond stage.")
+ self.cond_stage_model = self.first_stage_model
+ elif config == "__is_unconditional__":
+ print(f"Training {self.__class__.__name__} as an unconditional model.")
+ self.cond_stage_model = None
+ # self.be_unconditional = True
+ else:
+ model = instantiate_from_config(config)
+ self.cond_stage_model = model.eval()
+ self.cond_stage_model.train = disabled_train
+ for param in self.cond_stage_model.parameters():
+ param.requires_grad = False
+ else:
+ assert config != '__is_first_stage__'
+ assert config != '__is_unconditional__'
+ model = instantiate_from_config(config)
+ self.cond_stage_model = model
+
+ def _get_denoise_row_from_list(self, samples, desc='', force_no_decoder_quantization=False):
+ denoise_row = []
+ for zd in tqdm(samples, desc=desc):
+ denoise_row.append(self.decode_first_stage(zd.to(self.device),
+ force_not_quantize=force_no_decoder_quantization))
+ n_imgs_per_row = len(denoise_row)
+ denoise_row = torch.stack(denoise_row) # n_log_step, n_row, C, H, W
+ denoise_grid = rearrange(denoise_row, 'n b c h w -> b n c h w')
+ denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+ denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+ return denoise_grid
+
+ def get_first_stage_encoding(self, encoder_posterior):
+ if isinstance(encoder_posterior, DiagonalGaussianDistribution):
+ z = encoder_posterior.sample()
+ elif isinstance(encoder_posterior, torch.Tensor):
+ z = encoder_posterior
+ else:
+ raise NotImplementedError(f"encoder_posterior of type '{type(encoder_posterior)}' not yet implemented")
+ return self.scale_factor * z
+
+ def get_learned_conditioning(self, c):
+ if self.cond_stage_forward is None:
+ if hasattr(self.cond_stage_model, 'encode') and callable(self.cond_stage_model.encode):
+ c = self.cond_stage_model.encode(c)
+ if isinstance(c, DiagonalGaussianDistribution):
+ c = c.mode()
+ else:
+ c = self.cond_stage_model(c)
+ else:
+ assert hasattr(self.cond_stage_model, self.cond_stage_forward)
+ c = getattr(self.cond_stage_model, self.cond_stage_forward)(c)
+ return c
+
+ def meshgrid(self, h, w):
+ y = torch.arange(0, h).view(h, 1, 1).repeat(1, w, 1)
+ x = torch.arange(0, w).view(1, w, 1).repeat(h, 1, 1)
+
+ arr = torch.cat([y, x], dim=-1)
+ return arr
+
+ def delta_border(self, h, w):
+ """
+ :param h: height
+ :param w: width
+ :return: normalized distance to image border,
+ wtith min distance = 0 at border and max dist = 0.5 at image center
+ """
+ lower_right_corner = torch.tensor([h - 1, w - 1]).view(1, 1, 2)
+ arr = self.meshgrid(h, w) / lower_right_corner
+ dist_left_up = torch.min(arr, dim=-1, keepdims=True)[0]
+ dist_right_down = torch.min(1 - arr, dim=-1, keepdims=True)[0]
+ edge_dist = torch.min(torch.cat([dist_left_up, dist_right_down], dim=-1), dim=-1)[0]
+ return edge_dist
+
+ def get_weighting(self, h, w, Ly, Lx, device):
+ weighting = self.delta_border(h, w)
+ weighting = torch.clip(weighting, self.split_input_params["clip_min_weight"],
+ self.split_input_params["clip_max_weight"], )
+ weighting = weighting.view(1, h * w, 1).repeat(1, 1, Ly * Lx).to(device)
+
+ if self.split_input_params["tie_braker"]:
+ L_weighting = self.delta_border(Ly, Lx)
+ L_weighting = torch.clip(L_weighting,
+ self.split_input_params["clip_min_tie_weight"],
+ self.split_input_params["clip_max_tie_weight"])
+
+ L_weighting = L_weighting.view(1, 1, Ly * Lx).to(device)
+ weighting = weighting * L_weighting
+ return weighting
+
+ def get_fold_unfold(self, x, kernel_size, stride, uf=1, df=1): # todo load once not every time, shorten code
+ """
+ :param x: img of size (bs, c, h, w)
+ :return: n img crops of size (n, bs, c, kernel_size[0], kernel_size[1])
+ """
+ bs, nc, h, w = x.shape
+
+ # number of crops in image
+ Ly = (h - kernel_size[0]) // stride[0] + 1
+ Lx = (w - kernel_size[1]) // stride[1] + 1
+
+ if uf == 1 and df == 1:
+ fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+ unfold = torch.nn.Unfold(**fold_params)
+
+ fold = torch.nn.Fold(output_size=x.shape[2:], **fold_params)
+
+ weighting = self.get_weighting(kernel_size[0], kernel_size[1], Ly, Lx, x.device).to(x.dtype)
+ normalization = fold(weighting).view(1, 1, h, w) # normalizes the overlap
+ weighting = weighting.view((1, 1, kernel_size[0], kernel_size[1], Ly * Lx))
+
+ elif uf > 1 and df == 1:
+ fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+ unfold = torch.nn.Unfold(**fold_params)
+
+ fold_params2 = dict(kernel_size=(kernel_size[0] * uf, kernel_size[0] * uf),
+ dilation=1, padding=0,
+ stride=(stride[0] * uf, stride[1] * uf))
+ fold = torch.nn.Fold(output_size=(x.shape[2] * uf, x.shape[3] * uf), **fold_params2)
+
+ weighting = self.get_weighting(kernel_size[0] * uf, kernel_size[1] * uf, Ly, Lx, x.device).to(x.dtype)
+ normalization = fold(weighting).view(1, 1, h * uf, w * uf) # normalizes the overlap
+ weighting = weighting.view((1, 1, kernel_size[0] * uf, kernel_size[1] * uf, Ly * Lx))
+
+ elif df > 1 and uf == 1:
+ fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+ unfold = torch.nn.Unfold(**fold_params)
+
+ fold_params2 = dict(kernel_size=(kernel_size[0] // df, kernel_size[0] // df),
+ dilation=1, padding=0,
+ stride=(stride[0] // df, stride[1] // df))
+ fold = torch.nn.Fold(output_size=(x.shape[2] // df, x.shape[3] // df), **fold_params2)
+
+ weighting = self.get_weighting(kernel_size[0] // df, kernel_size[1] // df, Ly, Lx, x.device).to(x.dtype)
+ normalization = fold(weighting).view(1, 1, h // df, w // df) # normalizes the overlap
+ weighting = weighting.view((1, 1, kernel_size[0] // df, kernel_size[1] // df, Ly * Lx))
+
+ else:
+ raise NotImplementedError
+
+ return fold, unfold, normalization, weighting
+
+ @torch.no_grad()
+ def get_input(self, batch, k, return_first_stage_outputs=False, force_c_encode=False,
+ cond_key=None, return_original_cond=False, bs=None):
+ x = super().get_input(batch, k)
+ if bs is not None:
+ x = x[:bs]
+ x = x.to(self.device)
+ encoder_posterior = self.encode_first_stage(x)
+ z = self.get_first_stage_encoding(encoder_posterior).detach()
+
+ if self.model.conditioning_key is not None:
+ if cond_key is None:
+ cond_key = self.cond_stage_key
+ if cond_key != self.first_stage_key:
+ if cond_key in ['caption', 'coordinates_bbox']:
+ xc = batch[cond_key]
+ elif cond_key == 'class_label':
+ xc = batch
+ else:
+ xc = super().get_input(batch, cond_key).to(self.device)
+ else:
+ xc = x
+ if not self.cond_stage_trainable or force_c_encode:
+ if isinstance(xc, dict) or isinstance(xc, list):
+ # import pudb; pudb.set_trace()
+ c = self.get_learned_conditioning(xc)
+ else:
+ c = self.get_learned_conditioning(xc.to(self.device))
+ else:
+ c = xc
+ if bs is not None:
+ c = c[:bs]
+
+ if self.use_positional_encodings:
+ pos_x, pos_y = self.compute_latent_shifts(batch)
+ ckey = __conditioning_keys__[self.model.conditioning_key]
+ c = {ckey: c, 'pos_x': pos_x, 'pos_y': pos_y}
+
+ else:
+ c = None
+ xc = None
+ if self.use_positional_encodings:
+ pos_x, pos_y = self.compute_latent_shifts(batch)
+ c = {'pos_x': pos_x, 'pos_y': pos_y}
+ out = [z, c]
+ if return_first_stage_outputs:
+ xrec = self.decode_first_stage(z)
+ out.extend([x, xrec])
+ if return_original_cond:
+ out.append(xc)
+ return out
+
+ @torch.no_grad()
+ def decode_first_stage(self, z, predict_cids=False, force_not_quantize=False):
+ if predict_cids:
+ if z.dim() == 4:
+ z = torch.argmax(z.exp(), dim=1).long()
+ z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+ z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+ z = 1. / self.scale_factor * z
+
+ if hasattr(self, "split_input_params"):
+ if self.split_input_params["patch_distributed_vq"]:
+ ks = self.split_input_params["ks"] # eg. (128, 128)
+ stride = self.split_input_params["stride"] # eg. (64, 64)
+ uf = self.split_input_params["vqf"]
+ bs, nc, h, w = z.shape
+ if ks[0] > h or ks[1] > w:
+ ks = (min(ks[0], h), min(ks[1], w))
+ print("reducing Kernel")
+
+ if stride[0] > h or stride[1] > w:
+ stride = (min(stride[0], h), min(stride[1], w))
+ print("reducing stride")
+
+ fold, unfold, normalization, weighting = self.get_fold_unfold(z, ks, stride, uf=uf)
+
+ z = unfold(z) # (bn, nc * prod(**ks), L)
+ # 1. Reshape to img shape
+ z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L )
+
+ # 2. apply model loop over last dim
+ if isinstance(self.first_stage_model, VQModelInterface):
+ output_list = [self.first_stage_model.decode(z[:, :, :, :, i],
+ force_not_quantize=predict_cids or force_not_quantize)
+ for i in range(z.shape[-1])]
+ else:
+
+ output_list = [self.first_stage_model.decode(z[:, :, :, :, i])
+ for i in range(z.shape[-1])]
+
+ o = torch.stack(output_list, axis=-1) # # (bn, nc, ks[0], ks[1], L)
+ o = o * weighting
+ # Reverse 1. reshape to img shape
+ o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L)
+ # stitch crops together
+ decoded = fold(o)
+ decoded = decoded / normalization # norm is shape (1, 1, h, w)
+ return decoded
+ else:
+ if isinstance(self.first_stage_model, VQModelInterface):
+ return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize)
+ else:
+ return self.first_stage_model.decode(z)
+
+ else:
+ if isinstance(self.first_stage_model, VQModelInterface):
+ return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize)
+ else:
+ return self.first_stage_model.decode(z)
+
+ # same as above but without decorator
+ def differentiable_decode_first_stage(self, z, predict_cids=False, force_not_quantize=False):
+ if predict_cids:
+ if z.dim() == 4:
+ z = torch.argmax(z.exp(), dim=1).long()
+ z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+ z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+ z = 1. / self.scale_factor * z
+
+ if hasattr(self, "split_input_params"):
+ if self.split_input_params["patch_distributed_vq"]:
+ ks = self.split_input_params["ks"] # eg. (128, 128)
+ stride = self.split_input_params["stride"] # eg. (64, 64)
+ uf = self.split_input_params["vqf"]
+ bs, nc, h, w = z.shape
+ if ks[0] > h or ks[1] > w:
+ ks = (min(ks[0], h), min(ks[1], w))
+ print("reducing Kernel")
+
+ if stride[0] > h or stride[1] > w:
+ stride = (min(stride[0], h), min(stride[1], w))
+ print("reducing stride")
+
+ fold, unfold, normalization, weighting = self.get_fold_unfold(z, ks, stride, uf=uf)
+
+ z = unfold(z) # (bn, nc * prod(**ks), L)
+ # 1. Reshape to img shape
+ z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L )
+
+ # 2. apply model loop over last dim
+ if isinstance(self.first_stage_model, VQModelInterface):
+ output_list = [self.first_stage_model.decode(z[:, :, :, :, i],
+ force_not_quantize=predict_cids or force_not_quantize)
+ for i in range(z.shape[-1])]
+ else:
+
+ output_list = [self.first_stage_model.decode(z[:, :, :, :, i])
+ for i in range(z.shape[-1])]
+
+ o = torch.stack(output_list, axis=-1) # # (bn, nc, ks[0], ks[1], L)
+ o = o * weighting
+ # Reverse 1. reshape to img shape
+ o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L)
+ # stitch crops together
+ decoded = fold(o)
+ decoded = decoded / normalization # norm is shape (1, 1, h, w)
+ return decoded
+ else:
+ if isinstance(self.first_stage_model, VQModelInterface):
+ return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize)
+ else:
+ return self.first_stage_model.decode(z)
+
+ else:
+ if isinstance(self.first_stage_model, VQModelInterface):
+ return self.first_stage_model.decode(z, force_not_quantize=predict_cids or force_not_quantize)
+ else:
+ return self.first_stage_model.decode(z)
+
+ @torch.no_grad()
+ def encode_first_stage(self, x):
+ if hasattr(self, "split_input_params"):
+ if self.split_input_params["patch_distributed_vq"]:
+ ks = self.split_input_params["ks"] # eg. (128, 128)
+ stride = self.split_input_params["stride"] # eg. (64, 64)
+ df = self.split_input_params["vqf"]
+ self.split_input_params['original_image_size'] = x.shape[-2:]
+ bs, nc, h, w = x.shape
+ if ks[0] > h or ks[1] > w:
+ ks = (min(ks[0], h), min(ks[1], w))
+ print("reducing Kernel")
+
+ if stride[0] > h or stride[1] > w:
+ stride = (min(stride[0], h), min(stride[1], w))
+ print("reducing stride")
+
+ fold, unfold, normalization, weighting = self.get_fold_unfold(x, ks, stride, df=df)
+ z = unfold(x) # (bn, nc * prod(**ks), L)
+ # Reshape to img shape
+ z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L )
+
+ output_list = [self.first_stage_model.encode(z[:, :, :, :, i])
+ for i in range(z.shape[-1])]
+
+ o = torch.stack(output_list, axis=-1)
+ o = o * weighting
+
+ # Reverse reshape to img shape
+ o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L)
+ # stitch crops together
+ decoded = fold(o)
+ decoded = decoded / normalization
+ return decoded
+
+ else:
+ return self.first_stage_model.encode(x)
+ else:
+ return self.first_stage_model.encode(x)
+
+ def shared_step(self, batch, **kwargs):
+ x, c = self.get_input(batch, self.first_stage_key)
+ loss = self(x, c)
+ return loss
+
+ def forward(self, x, c, *args, **kwargs):
+ t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long()
+ if self.model.conditioning_key is not None:
+ assert c is not None
+ if self.cond_stage_trainable:
+ c = self.get_learned_conditioning(c)
+ if self.shorten_cond_schedule: # TODO: drop this option
+ tc = self.cond_ids[t].to(self.device)
+ c = self.q_sample(x_start=c, t=tc, noise=torch.randn_like(c.float()))
+ return self.p_losses(x, c, t, *args, **kwargs)
+
+ def _rescale_annotations(self, bboxes, crop_coordinates): # TODO: move to dataset
+ def rescale_bbox(bbox):
+ x0 = clamp((bbox[0] - crop_coordinates[0]) / crop_coordinates[2])
+ y0 = clamp((bbox[1] - crop_coordinates[1]) / crop_coordinates[3])
+ w = min(bbox[2] / crop_coordinates[2], 1 - x0)
+ h = min(bbox[3] / crop_coordinates[3], 1 - y0)
+ return x0, y0, w, h
+
+ return [rescale_bbox(b) for b in bboxes]
+
+ def apply_model(self, x_noisy, t, cond, return_ids=False):
+
+ if isinstance(cond, dict):
+ # hybrid case, cond is exptected to be a dict
+ pass
+ else:
+ if not isinstance(cond, list):
+ cond = [cond]
+ key = 'c_concat' if self.model.conditioning_key == 'concat' else 'c_crossattn'
+ cond = {key: cond}
+
+ if hasattr(self, "split_input_params"):
+ assert len(cond) == 1 # todo can only deal with one conditioning atm
+ assert not return_ids
+ ks = self.split_input_params["ks"] # eg. (128, 128)
+ stride = self.split_input_params["stride"] # eg. (64, 64)
+
+ h, w = x_noisy.shape[-2:]
+
+ fold, unfold, normalization, weighting = self.get_fold_unfold(x_noisy, ks, stride)
+
+ z = unfold(x_noisy) # (bn, nc * prod(**ks), L)
+ # Reshape to img shape
+ z = z.view((z.shape[0], -1, ks[0], ks[1], z.shape[-1])) # (bn, nc, ks[0], ks[1], L )
+ z_list = [z[:, :, :, :, i] for i in range(z.shape[-1])]
+
+ if self.cond_stage_key in ["image", "LR_image", "segmentation",
+ 'bbox_img'] and self.model.conditioning_key: # todo check for completeness
+ c_key = next(iter(cond.keys())) # get key
+ c = next(iter(cond.values())) # get value
+ assert (len(c) == 1) # todo extend to list with more than one elem
+ c = c[0] # get element
+
+ c = unfold(c)
+ c = c.view((c.shape[0], -1, ks[0], ks[1], c.shape[-1])) # (bn, nc, ks[0], ks[1], L )
+
+ cond_list = [{c_key: [c[:, :, :, :, i]]} for i in range(c.shape[-1])]
+
+ elif self.cond_stage_key == 'coordinates_bbox':
+ assert 'original_image_size' in self.split_input_params, 'BoudingBoxRescaling is missing original_image_size'
+
+ # assuming padding of unfold is always 0 and its dilation is always 1
+ n_patches_per_row = int((w - ks[0]) / stride[0] + 1)
+ full_img_h, full_img_w = self.split_input_params['original_image_size']
+ # as we are operating on latents, we need the factor from the original image size to the
+ # spatial latent size to properly rescale the crops for regenerating the bbox annotations
+ num_downs = self.first_stage_model.encoder.num_resolutions - 1
+ rescale_latent = 2 ** (num_downs)
+
+ # get top left postions of patches as conforming for the bbbox tokenizer, therefore we
+ # need to rescale the tl patch coordinates to be in between (0,1)
+ tl_patch_coordinates = [(rescale_latent * stride[0] * (patch_nr % n_patches_per_row) / full_img_w,
+ rescale_latent * stride[1] * (patch_nr // n_patches_per_row) / full_img_h)
+ for patch_nr in range(z.shape[-1])]
+
+ # patch_limits are tl_coord, width and height coordinates as (x_tl, y_tl, h, w)
+ patch_limits = [(x_tl, y_tl,
+ rescale_latent * ks[0] / full_img_w,
+ rescale_latent * ks[1] / full_img_h) for x_tl, y_tl in tl_patch_coordinates]
+ # patch_values = [(np.arange(x_tl,min(x_tl+ks, 1.)),np.arange(y_tl,min(y_tl+ks, 1.))) for x_tl, y_tl in tl_patch_coordinates]
+
+ # tokenize crop coordinates for the bounding boxes of the respective patches
+ patch_limits_tknzd = [torch.LongTensor(self.bbox_tokenizer._crop_encoder(bbox))[None].to(self.device)
+ for bbox in patch_limits] # list of length l with tensors of shape (1, 2)
+ print(patch_limits_tknzd[0].shape)
+ # cut tknzd crop position from conditioning
+ assert isinstance(cond, dict), 'cond must be dict to be fed into model'
+ cut_cond = cond['c_crossattn'][0][..., :-2].to(self.device)
+ print(cut_cond.shape)
+
+ adapted_cond = torch.stack([torch.cat([cut_cond, p], dim=1) for p in patch_limits_tknzd])
+ adapted_cond = rearrange(adapted_cond, 'l b n -> (l b) n')
+ print(adapted_cond.shape)
+ adapted_cond = self.get_learned_conditioning(adapted_cond)
+ print(adapted_cond.shape)
+ adapted_cond = rearrange(adapted_cond, '(l b) n d -> l b n d', l=z.shape[-1])
+ print(adapted_cond.shape)
+
+ cond_list = [{'c_crossattn': [e]} for e in adapted_cond]
+
+ else:
+ cond_list = [cond for i in range(z.shape[-1])] # Todo make this more efficient
+
+ # apply model by loop over crops
+ output_list = [self.model(z_list[i], t, **cond_list[i]) for i in range(z.shape[-1])]
+ assert not isinstance(output_list[0],
+ tuple) # todo cant deal with multiple model outputs check this never happens
+
+ o = torch.stack(output_list, axis=-1)
+ o = o * weighting
+ # Reverse reshape to img shape
+ o = o.view((o.shape[0], -1, o.shape[-1])) # (bn, nc * ks[0] * ks[1], L)
+ # stitch crops together
+ x_recon = fold(o) / normalization
+
+ else:
+ x_recon = self.model(x_noisy, t, **cond)
+
+ if isinstance(x_recon, tuple) and not return_ids:
+ return x_recon[0]
+ else:
+ return x_recon
+
+ def _predict_eps_from_xstart(self, x_t, t, pred_xstart):
+ return (extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t - pred_xstart) / \
+ extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape)
+
+ def _prior_bpd(self, x_start):
+ """
+ Get the prior KL term for the variational lower-bound, measured in
+ bits-per-dim.
+ This term can't be optimized, as it only depends on the encoder.
+ :param x_start: the [N x C x ...] tensor of inputs.
+ :return: a batch of [N] KL values (in bits), one per batch element.
+ """
+ batch_size = x_start.shape[0]
+ t = torch.tensor([self.num_timesteps - 1] * batch_size, device=x_start.device)
+ qt_mean, _, qt_log_variance = self.q_mean_variance(x_start, t)
+ kl_prior = normal_kl(mean1=qt_mean, logvar1=qt_log_variance, mean2=0.0, logvar2=0.0)
+ return mean_flat(kl_prior) / np.log(2.0)
+
+ def p_losses(self, x_start, cond, t, noise=None):
+ noise = default(noise, lambda: torch.randn_like(x_start))
+ x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+ model_output = self.apply_model(x_noisy, t, cond)
+
+ loss_dict = {}
+ prefix = 'train' if self.training else 'val'
+
+ if self.parameterization == "x0":
+ target = x_start
+ elif self.parameterization == "eps":
+ target = noise
+ else:
+ raise NotImplementedError()
+
+ loss_simple = self.get_loss(model_output, target, mean=False).mean([1, 2, 3])
+ loss_dict.update({f'{prefix}/loss_simple': loss_simple.mean()})
+
+ logvar_t = self.logvar[t].to(self.device)
+ loss = loss_simple / torch.exp(logvar_t) + logvar_t
+ # loss = loss_simple / torch.exp(self.logvar) + self.logvar
+ if self.learn_logvar:
+ loss_dict.update({f'{prefix}/loss_gamma': loss.mean()})
+ loss_dict.update({'logvar': self.logvar.data.mean()})
+
+ loss = self.l_simple_weight * loss.mean()
+
+ loss_vlb = self.get_loss(model_output, target, mean=False).mean(dim=(1, 2, 3))
+ loss_vlb = (self.lvlb_weights[t] * loss_vlb).mean()
+ loss_dict.update({f'{prefix}/loss_vlb': loss_vlb})
+ loss += (self.original_elbo_weight * loss_vlb)
+ loss_dict.update({f'{prefix}/loss': loss})
+
+ return loss, loss_dict
+
+ def p_mean_variance(self, x, c, t, clip_denoised: bool, return_codebook_ids=False, quantize_denoised=False,
+ return_x0=False, score_corrector=None, corrector_kwargs=None):
+ t_in = t
+ model_out = self.apply_model(x, t_in, c, return_ids=return_codebook_ids)
+
+ if score_corrector is not None:
+ assert self.parameterization == "eps"
+ model_out = score_corrector.modify_score(self, model_out, x, t, c, **corrector_kwargs)
+
+ if return_codebook_ids:
+ model_out, logits = model_out
+
+ if self.parameterization == "eps":
+ x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+ elif self.parameterization == "x0":
+ x_recon = model_out
+ else:
+ raise NotImplementedError()
+
+ if clip_denoised:
+ x_recon.clamp_(-1., 1.)
+ if quantize_denoised:
+ x_recon, _, [_, _, indices] = self.first_stage_model.quantize(x_recon)
+ model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+ if return_codebook_ids:
+ return model_mean, posterior_variance, posterior_log_variance, logits
+ elif return_x0:
+ return model_mean, posterior_variance, posterior_log_variance, x_recon
+ else:
+ return model_mean, posterior_variance, posterior_log_variance
+
+ @torch.no_grad()
+ def p_sample(self, x, c, t, clip_denoised=False, repeat_noise=False,
+ return_codebook_ids=False, quantize_denoised=False, return_x0=False,
+ temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None):
+ b, *_, device = *x.shape, x.device
+ outputs = self.p_mean_variance(x=x, c=c, t=t, clip_denoised=clip_denoised,
+ return_codebook_ids=return_codebook_ids,
+ quantize_denoised=quantize_denoised,
+ return_x0=return_x0,
+ score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+ if return_codebook_ids:
+ raise DeprecationWarning("Support dropped.")
+ model_mean, _, model_log_variance, logits = outputs
+ elif return_x0:
+ model_mean, _, model_log_variance, x0 = outputs
+ else:
+ model_mean, _, model_log_variance = outputs
+
+ noise = noise_like(x.shape, device, repeat_noise) * temperature
+ if noise_dropout > 0.:
+ noise = torch.nn.functional.dropout(noise, p=noise_dropout)
+ # no noise when t == 0
+ nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+
+ if return_codebook_ids:
+ return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, logits.argmax(dim=1)
+ if return_x0:
+ return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, x0
+ else:
+ return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+ @torch.no_grad()
+ def progressive_denoising(self, cond, shape, verbose=True, callback=None, quantize_denoised=False,
+ img_callback=None, mask=None, x0=None, temperature=1., noise_dropout=0.,
+ score_corrector=None, corrector_kwargs=None, batch_size=None, x_T=None, start_T=None,
+ log_every_t=None):
+ if not log_every_t:
+ log_every_t = self.log_every_t
+ timesteps = self.num_timesteps
+ if batch_size is not None:
+ b = batch_size if batch_size is not None else shape[0]
+ shape = [batch_size] + list(shape)
+ else:
+ b = batch_size = shape[0]
+ if x_T is None:
+ img = torch.randn(shape, device=self.device)
+ else:
+ img = x_T
+ intermediates = []
+ if cond is not None:
+ if isinstance(cond, dict):
+ cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+ list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+ else:
+ cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+
+ if start_T is not None:
+ timesteps = min(timesteps, start_T)
+ iterator = tqdm(reversed(range(0, timesteps)), desc='Progressive Generation',
+ total=timesteps) if verbose else reversed(
+ range(0, timesteps))
+ if type(temperature) == float:
+ temperature = [temperature] * timesteps
+
+ for i in iterator:
+ ts = torch.full((b,), i, device=self.device, dtype=torch.long)
+ if self.shorten_cond_schedule:
+ assert self.model.conditioning_key != 'hybrid'
+ tc = self.cond_ids[ts].to(cond.device)
+ cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+ img, x0_partial = self.p_sample(img, cond, ts,
+ clip_denoised=self.clip_denoised,
+ quantize_denoised=quantize_denoised, return_x0=True,
+ temperature=temperature[i], noise_dropout=noise_dropout,
+ score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+ if mask is not None:
+ assert x0 is not None
+ img_orig = self.q_sample(x0, ts)
+ img = img_orig * mask + (1. - mask) * img
+
+ if i % log_every_t == 0 or i == timesteps - 1:
+ intermediates.append(x0_partial)
+ if callback: callback(i)
+ if img_callback: img_callback(img, i)
+ return img, intermediates
+
+ @torch.no_grad()
+ def p_sample_loop(self, cond, shape, return_intermediates=False,
+ x_T=None, verbose=True, callback=None, timesteps=None, quantize_denoised=False,
+ mask=None, x0=None, img_callback=None, start_T=None,
+ log_every_t=None):
+
+ if not log_every_t:
+ log_every_t = self.log_every_t
+ device = self.betas.device
+ b = shape[0]
+ if x_T is None:
+ img = torch.randn(shape, device=device)
+ else:
+ img = x_T
+
+ intermediates = [img]
+ if timesteps is None:
+ timesteps = self.num_timesteps
+
+ if start_T is not None:
+ timesteps = min(timesteps, start_T)
+ iterator = tqdm(reversed(range(0, timesteps)), desc='Sampling t', total=timesteps) if verbose else reversed(
+ range(0, timesteps))
+
+ if mask is not None:
+ assert x0 is not None
+ assert x0.shape[2:3] == mask.shape[2:3] # spatial size has to match
+
+ for i in iterator:
+ ts = torch.full((b,), i, device=device, dtype=torch.long)
+ if self.shorten_cond_schedule:
+ assert self.model.conditioning_key != 'hybrid'
+ tc = self.cond_ids[ts].to(cond.device)
+ cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+ img = self.p_sample(img, cond, ts,
+ clip_denoised=self.clip_denoised,
+ quantize_denoised=quantize_denoised)
+ if mask is not None:
+ img_orig = self.q_sample(x0, ts)
+ img = img_orig * mask + (1. - mask) * img
+
+ if i % log_every_t == 0 or i == timesteps - 1:
+ intermediates.append(img)
+ if callback: callback(i)
+ if img_callback: img_callback(img, i)
+
+ if return_intermediates:
+ return img, intermediates
+ return img
+
+ @torch.no_grad()
+ def sample(self, cond, batch_size=16, return_intermediates=False, x_T=None,
+ verbose=True, timesteps=None, quantize_denoised=False,
+ mask=None, x0=None, shape=None,**kwargs):
+ if shape is None:
+ shape = (batch_size, self.channels, self.image_size, self.image_size)
+ if cond is not None:
+ if isinstance(cond, dict):
+ cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+ list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+ else:
+ cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+ return self.p_sample_loop(cond,
+ shape,
+ return_intermediates=return_intermediates, x_T=x_T,
+ verbose=verbose, timesteps=timesteps, quantize_denoised=quantize_denoised,
+ mask=mask, x0=x0)
+
+ @torch.no_grad()
+ def sample_log(self,cond,batch_size,ddim, ddim_steps,**kwargs):
+
+ if ddim:
+ ddim_sampler = DDIMSampler(self)
+ shape = (self.channels, self.image_size, self.image_size)
+ samples, intermediates =ddim_sampler.sample(ddim_steps,batch_size,
+ shape,cond,verbose=False,**kwargs)
+
+ else:
+ samples, intermediates = self.sample(cond=cond, batch_size=batch_size,
+ return_intermediates=True,**kwargs)
+
+ return samples, intermediates
+
+
+ @torch.no_grad()
+ def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None,
+ quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+ plot_diffusion_rows=True, **kwargs):
+
+ use_ddim = ddim_steps is not None
+
+ log = dict()
+ z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key,
+ return_first_stage_outputs=True,
+ force_c_encode=True,
+ return_original_cond=True,
+ bs=N)
+ N = min(x.shape[0], N)
+ n_row = min(x.shape[0], n_row)
+ log["inputs"] = x
+ log["reconstruction"] = xrec
+ if self.model.conditioning_key is not None:
+ if hasattr(self.cond_stage_model, "decode"):
+ xc = self.cond_stage_model.decode(c)
+ log["conditioning"] = xc
+ elif self.cond_stage_key in ["caption"]:
+ xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["caption"])
+ log["conditioning"] = xc
+ elif self.cond_stage_key == 'class_label':
+ xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"])
+ log['conditioning'] = xc
+ elif isimage(xc):
+ log["conditioning"] = xc
+ if ismap(xc):
+ log["original_conditioning"] = self.to_rgb(xc)
+
+ if plot_diffusion_rows:
+ # get diffusion row
+ diffusion_row = list()
+ z_start = z[:n_row]
+ for t in range(self.num_timesteps):
+ if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+ t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+ t = t.to(self.device).long()
+ noise = torch.randn_like(z_start)
+ z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+ diffusion_row.append(self.decode_first_stage(z_noisy))
+
+ diffusion_row = torch.stack(diffusion_row) # n_log_step, n_row, C, H, W
+ diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+ diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+ diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+ log["diffusion_row"] = diffusion_grid
+
+ if sample:
+ # get denoise row
+ with self.ema_scope("Plotting"):
+ samples, z_denoise_row = self.sample_log(cond=c,batch_size=N,ddim=use_ddim,
+ ddim_steps=ddim_steps,eta=ddim_eta)
+ # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+ x_samples = self.decode_first_stage(samples)
+ log["samples"] = x_samples
+ if plot_denoise_rows:
+ denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+ log["denoise_row"] = denoise_grid
+
+ if quantize_denoised and not isinstance(self.first_stage_model, AutoencoderKL) and not isinstance(
+ self.first_stage_model, IdentityFirstStage):
+ # also display when quantizing x0 while sampling
+ with self.ema_scope("Plotting Quantized Denoised"):
+ samples, z_denoise_row = self.sample_log(cond=c,batch_size=N,ddim=use_ddim,
+ ddim_steps=ddim_steps,eta=ddim_eta,
+ quantize_denoised=True)
+ # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True,
+ # quantize_denoised=True)
+ x_samples = self.decode_first_stage(samples.to(self.device))
+ log["samples_x0_quantized"] = x_samples
+
+ if inpaint:
+ # make a simple center square
+ b, h, w = z.shape[0], z.shape[2], z.shape[3]
+ mask = torch.ones(N, h, w).to(self.device)
+ # zeros will be filled in
+ mask[:, h // 4:3 * h // 4, w // 4:3 * w // 4] = 0.
+ mask = mask[:, None, ...]
+ with self.ema_scope("Plotting Inpaint"):
+
+ samples, _ = self.sample_log(cond=c,batch_size=N,ddim=use_ddim, eta=ddim_eta,
+ ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+ x_samples = self.decode_first_stage(samples.to(self.device))
+ log["samples_inpainting"] = x_samples
+ log["mask"] = mask
+
+ # outpaint
+ with self.ema_scope("Plotting Outpaint"):
+ samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,eta=ddim_eta,
+ ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+ x_samples = self.decode_first_stage(samples.to(self.device))
+ log["samples_outpainting"] = x_samples
+
+ if plot_progressive_rows:
+ with self.ema_scope("Plotting Progressives"):
+ img, progressives = self.progressive_denoising(c,
+ shape=(self.channels, self.image_size, self.image_size),
+ batch_size=N)
+ prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation")
+ log["progressive_row"] = prog_row
+
+ if return_keys:
+ if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+ return log
+ else:
+ return {key: log[key] for key in return_keys}
+ return log
+
+ def configure_optimizers(self):
+ lr = self.learning_rate
+ params = list(self.model.parameters())
+ if self.cond_stage_trainable:
+ print(f"{self.__class__.__name__}: Also optimizing conditioner params!")
+ params = params + list(self.cond_stage_model.parameters())
+ if self.learn_logvar:
+ print('Diffusion model optimizing logvar')
+ params.append(self.logvar)
+ opt = torch.optim.AdamW(params, lr=lr)
+ if self.use_scheduler:
+ assert 'target' in self.scheduler_config
+ scheduler = instantiate_from_config(self.scheduler_config)
+
+ print("Setting up LambdaLR scheduler...")
+ scheduler = [
+ {
+ 'scheduler': LambdaLR(opt, lr_lambda=scheduler.schedule),
+ 'interval': 'step',
+ 'frequency': 1
+ }]
+ return [opt], scheduler
+ return opt
+
+ @torch.no_grad()
+ def to_rgb(self, x):
+ x = x.float()
+ if not hasattr(self, "colorize"):
+ self.colorize = torch.randn(3, x.shape[1], 1, 1).to(x)
+ x = nn.functional.conv2d(x, weight=self.colorize)
+ x = 2. * (x - x.min()) / (x.max() - x.min()) - 1.
+ return x
+
+
+class DiffusionWrapper(pl.LightningModule):
+ def __init__(self, diff_model_config, conditioning_key):
+ super().__init__()
+ self.diffusion_model = instantiate_from_config(diff_model_config)
+ self.conditioning_key = conditioning_key
+ assert self.conditioning_key in [None, 'concat', 'crossattn', 'hybrid', 'adm']
+
+ def forward(self, x, t, c_concat: list = None, c_crossattn: list = None):
+ if self.conditioning_key is None:
+ out = self.diffusion_model(x, t)
+ elif self.conditioning_key == 'concat':
+ xc = torch.cat([x] + c_concat, dim=1)
+ out = self.diffusion_model(xc, t)
+ elif self.conditioning_key == 'crossattn':
+ cc = torch.cat(c_crossattn, 1)
+ out = self.diffusion_model(x, t, context=cc)
+ elif self.conditioning_key == 'hybrid':
+ xc = torch.cat([x] + c_concat, dim=1)
+ cc = torch.cat(c_crossattn, 1)
+ out = self.diffusion_model(xc, t, context=cc)
+ elif self.conditioning_key == 'adm':
+ cc = c_crossattn[0]
+ out = self.diffusion_model(x, t, y=cc)
+ else:
+ raise NotImplementedError()
+
+ return out
+
+
+class Layout2ImgDiffusion(LatentDiffusion):
+ # TODO: move all layout-specific hacks to this class
+ def __init__(self, cond_stage_key, *args, **kwargs):
+ assert cond_stage_key == 'coordinates_bbox', 'Layout2ImgDiffusion only for cond_stage_key="coordinates_bbox"'
+ super().__init__(cond_stage_key=cond_stage_key, *args, **kwargs)
+
+ def log_images(self, batch, N=8, *args, **kwargs):
+ logs = super().log_images(batch=batch, N=N, *args, **kwargs)
+
+ key = 'train' if self.training else 'validation'
+ dset = self.trainer.datamodule.datasets[key]
+ mapper = dset.conditional_builders[self.cond_stage_key]
+
+ bbox_imgs = []
+ map_fn = lambda catno: dset.get_textual_label(dset.get_category_id(catno))
+ for tknzd_bbox in batch[self.cond_stage_key][:N]:
+ bboximg = mapper.plot(tknzd_bbox.detach().cpu(), map_fn, (256, 256))
+ bbox_imgs.append(bboximg)
+
+ cond_img = torch.stack(bbox_imgs, dim=0)
+ logs['bbox_image'] = cond_img
+ return logs
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/__init__.py b/OCR_earsing/latent_diffusion/ldm/modules/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/attention.py b/OCR_earsing/latent_diffusion/ldm/modules/attention.py
new file mode 100644
index 0000000..f4eff39
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/attention.py
@@ -0,0 +1,261 @@
+from inspect import isfunction
+import math
+import torch
+import torch.nn.functional as F
+from torch import nn, einsum
+from einops import rearrange, repeat
+
+from ldm.modules.diffusionmodules.util import checkpoint
+
+
+def exists(val):
+ return val is not None
+
+
+def uniq(arr):
+ return{el: True for el in arr}.keys()
+
+
+def default(val, d):
+ if exists(val):
+ return val
+ return d() if isfunction(d) else d
+
+
+def max_neg_value(t):
+ return -torch.finfo(t.dtype).max
+
+
+def init_(tensor):
+ dim = tensor.shape[-1]
+ std = 1 / math.sqrt(dim)
+ tensor.uniform_(-std, std)
+ return tensor
+
+
+# feedforward
+class GEGLU(nn.Module):
+ def __init__(self, dim_in, dim_out):
+ super().__init__()
+ self.proj = nn.Linear(dim_in, dim_out * 2)
+
+ def forward(self, x):
+ x, gate = self.proj(x).chunk(2, dim=-1)
+ return x * F.gelu(gate)
+
+
+class FeedForward(nn.Module):
+ def __init__(self, dim, dim_out=None, mult=4, glu=False, dropout=0.):
+ super().__init__()
+ inner_dim = int(dim * mult)
+ dim_out = default(dim_out, dim)
+ project_in = nn.Sequential(
+ nn.Linear(dim, inner_dim),
+ nn.GELU()
+ ) if not glu else GEGLU(dim, inner_dim)
+
+ self.net = nn.Sequential(
+ project_in,
+ nn.Dropout(dropout),
+ nn.Linear(inner_dim, dim_out)
+ )
+
+ def forward(self, x):
+ return self.net(x)
+
+
+def zero_module(module):
+ """
+ Zero out the parameters of a module and return it.
+ """
+ for p in module.parameters():
+ p.detach().zero_()
+ return module
+
+
+def Normalize(in_channels):
+ return torch.nn.GroupNorm(num_groups=32, num_channels=in_channels, eps=1e-6, affine=True)
+
+
+class LinearAttention(nn.Module):
+ def __init__(self, dim, heads=4, dim_head=32):
+ super().__init__()
+ self.heads = heads
+ hidden_dim = dim_head * heads
+ self.to_qkv = nn.Conv2d(dim, hidden_dim * 3, 1, bias = False)
+ self.to_out = nn.Conv2d(hidden_dim, dim, 1)
+
+ def forward(self, x):
+ b, c, h, w = x.shape
+ qkv = self.to_qkv(x)
+ q, k, v = rearrange(qkv, 'b (qkv heads c) h w -> qkv b heads c (h w)', heads = self.heads, qkv=3)
+ k = k.softmax(dim=-1)
+ context = torch.einsum('bhdn,bhen->bhde', k, v)
+ out = torch.einsum('bhde,bhdn->bhen', context, q)
+ out = rearrange(out, 'b heads c (h w) -> b (heads c) h w', heads=self.heads, h=h, w=w)
+ return self.to_out(out)
+
+
+class SpatialSelfAttention(nn.Module):
+ def __init__(self, in_channels):
+ super().__init__()
+ self.in_channels = in_channels
+
+ self.norm = Normalize(in_channels)
+ self.q = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.k = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.v = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.proj_out = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+
+ def forward(self, x):
+ h_ = x
+ h_ = self.norm(h_)
+ q = self.q(h_)
+ k = self.k(h_)
+ v = self.v(h_)
+
+ # compute attention
+ b,c,h,w = q.shape
+ q = rearrange(q, 'b c h w -> b (h w) c')
+ k = rearrange(k, 'b c h w -> b c (h w)')
+ w_ = torch.einsum('bij,bjk->bik', q, k)
+
+ w_ = w_ * (int(c)**(-0.5))
+ w_ = torch.nn.functional.softmax(w_, dim=2)
+
+ # attend to values
+ v = rearrange(v, 'b c h w -> b c (h w)')
+ w_ = rearrange(w_, 'b i j -> b j i')
+ h_ = torch.einsum('bij,bjk->bik', v, w_)
+ h_ = rearrange(h_, 'b c (h w) -> b c h w', h=h)
+ h_ = self.proj_out(h_)
+
+ return x+h_
+
+
+class CrossAttention(nn.Module):
+ def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.):
+ super().__init__()
+ inner_dim = dim_head * heads
+ context_dim = default(context_dim, query_dim)
+
+ self.scale = dim_head ** -0.5
+ self.heads = heads
+
+ self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
+ self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
+ self.to_v = nn.Linear(context_dim, inner_dim, bias=False)
+
+ self.to_out = nn.Sequential(
+ nn.Linear(inner_dim, query_dim),
+ nn.Dropout(dropout)
+ )
+
+ def forward(self, x, context=None, mask=None):
+ h = self.heads
+
+ q = self.to_q(x)
+ context = default(context, x)
+ k = self.to_k(context)
+ v = self.to_v(context)
+
+ q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))
+
+ sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+
+ if exists(mask):
+ mask = rearrange(mask, 'b ... -> b (...)')
+ max_neg_value = -torch.finfo(sim.dtype).max
+ mask = repeat(mask, 'b j -> (b h) () j', h=h)
+ sim.masked_fill_(~mask, max_neg_value)
+
+ # attention, what we cannot get enough of
+ attn = sim.softmax(dim=-1)
+
+ out = einsum('b i j, b j d -> b i d', attn, v)
+ out = rearrange(out, '(b h) n d -> b n (h d)', h=h)
+ return self.to_out(out)
+
+
+class BasicTransformerBlock(nn.Module):
+ def __init__(self, dim, n_heads, d_head, dropout=0., context_dim=None, gated_ff=True, checkpoint=True):
+ super().__init__()
+ self.attn1 = CrossAttention(query_dim=dim, heads=n_heads, dim_head=d_head, dropout=dropout) # is a self-attention
+ self.ff = FeedForward(dim, dropout=dropout, glu=gated_ff)
+ self.attn2 = CrossAttention(query_dim=dim, context_dim=context_dim,
+ heads=n_heads, dim_head=d_head, dropout=dropout) # is self-attn if context is none
+ self.norm1 = nn.LayerNorm(dim)
+ self.norm2 = nn.LayerNorm(dim)
+ self.norm3 = nn.LayerNorm(dim)
+ self.checkpoint = checkpoint
+
+ def forward(self, x, context=None):
+ return checkpoint(self._forward, (x, context), self.parameters(), self.checkpoint)
+
+ def _forward(self, x, context=None):
+ x = self.attn1(self.norm1(x)) + x
+ x = self.attn2(self.norm2(x), context=context) + x
+ x = self.ff(self.norm3(x)) + x
+ return x
+
+
+class SpatialTransformer(nn.Module):
+ """
+ Transformer block for image-like data.
+ First, project the input (aka embedding)
+ and reshape to b, t, d.
+ Then apply standard transformer action.
+ Finally, reshape to image
+ """
+ def __init__(self, in_channels, n_heads, d_head,
+ depth=1, dropout=0., context_dim=None):
+ super().__init__()
+ self.in_channels = in_channels
+ inner_dim = n_heads * d_head
+ self.norm = Normalize(in_channels)
+
+ self.proj_in = nn.Conv2d(in_channels,
+ inner_dim,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+
+ self.transformer_blocks = nn.ModuleList(
+ [BasicTransformerBlock(inner_dim, n_heads, d_head, dropout=dropout, context_dim=context_dim)
+ for d in range(depth)]
+ )
+
+ self.proj_out = zero_module(nn.Conv2d(inner_dim,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0))
+
+ def forward(self, x, context=None):
+ # note: if no context is given, cross-attention defaults to self-attention
+ b, c, h, w = x.shape
+ x_in = x
+ x = self.norm(x)
+ x = self.proj_in(x)
+ x = rearrange(x, 'b c h w -> b (h w) c')
+ for block in self.transformer_blocks:
+ x = block(x, context=context)
+ x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w)
+ x = self.proj_out(x)
+ return x + x_in
\ No newline at end of file
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/__init__.py b/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/model.py b/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/model.py
new file mode 100644
index 0000000..533e589
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/model.py
@@ -0,0 +1,835 @@
+# pytorch_diffusion + derived encoder decoder
+import math
+import torch
+import torch.nn as nn
+import numpy as np
+from einops import rearrange
+
+from ldm.util import instantiate_from_config
+from ldm.modules.attention import LinearAttention
+
+
+def get_timestep_embedding(timesteps, embedding_dim):
+ """
+ This matches the implementation in Denoising Diffusion Probabilistic Models:
+ From Fairseq.
+ Build sinusoidal embeddings.
+ This matches the implementation in tensor2tensor, but differs slightly
+ from the description in Section 3.5 of "Attention Is All You Need".
+ """
+ assert len(timesteps.shape) == 1
+
+ half_dim = embedding_dim // 2
+ emb = math.log(10000) / (half_dim - 1)
+ emb = torch.exp(torch.arange(half_dim, dtype=torch.float32) * -emb)
+ emb = emb.to(device=timesteps.device)
+ emb = timesteps.float()[:, None] * emb[None, :]
+ emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
+ if embedding_dim % 2 == 1: # zero pad
+ emb = torch.nn.functional.pad(emb, (0,1,0,0))
+ return emb
+
+
+def nonlinearity(x):
+ # swish
+ return x*torch.sigmoid(x)
+
+
+def Normalize(in_channels, num_groups=32):
+ return torch.nn.GroupNorm(num_groups=num_groups, num_channels=in_channels, eps=1e-6, affine=True)
+
+
+class Upsample(nn.Module):
+ def __init__(self, in_channels, with_conv):
+ super().__init__()
+ self.with_conv = with_conv
+ if self.with_conv:
+ self.conv = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x):
+ x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest")
+ if self.with_conv:
+ x = self.conv(x)
+ return x
+
+
+class Downsample(nn.Module):
+ def __init__(self, in_channels, with_conv):
+ super().__init__()
+ self.with_conv = with_conv
+ if self.with_conv:
+ # no asymmetric padding in torch conv, must do it ourselves
+ self.conv = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=3,
+ stride=2,
+ padding=0)
+
+ def forward(self, x):
+ if self.with_conv:
+ pad = (0,1,0,1)
+ x = torch.nn.functional.pad(x, pad, mode="constant", value=0)
+ x = self.conv(x)
+ else:
+ x = torch.nn.functional.avg_pool2d(x, kernel_size=2, stride=2)
+ return x
+
+
+class ResnetBlock(nn.Module):
+ def __init__(self, *, in_channels, out_channels=None, conv_shortcut=False,
+ dropout, temb_channels=512):
+ super().__init__()
+ self.in_channels = in_channels
+ out_channels = in_channels if out_channels is None else out_channels
+ self.out_channels = out_channels
+ self.use_conv_shortcut = conv_shortcut
+
+ self.norm1 = Normalize(in_channels)
+ self.conv1 = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+ if temb_channels > 0:
+ self.temb_proj = torch.nn.Linear(temb_channels,
+ out_channels)
+ self.norm2 = Normalize(out_channels)
+ self.dropout = torch.nn.Dropout(dropout)
+ self.conv2 = torch.nn.Conv2d(out_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+ if self.in_channels != self.out_channels:
+ if self.use_conv_shortcut:
+ self.conv_shortcut = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+ else:
+ self.nin_shortcut = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+
+ def forward(self, x, temb):
+ h = x
+ h = self.norm1(h)
+ h = nonlinearity(h)
+ h = self.conv1(h)
+
+ if temb is not None:
+ h = h + self.temb_proj(nonlinearity(temb))[:,:,None,None]
+
+ h = self.norm2(h)
+ h = nonlinearity(h)
+ h = self.dropout(h)
+ h = self.conv2(h)
+
+ if self.in_channels != self.out_channels:
+ if self.use_conv_shortcut:
+ x = self.conv_shortcut(x)
+ else:
+ x = self.nin_shortcut(x)
+
+ return x+h
+
+
+class LinAttnBlock(LinearAttention):
+ """to match AttnBlock usage"""
+ def __init__(self, in_channels):
+ super().__init__(dim=in_channels, heads=1, dim_head=in_channels)
+
+
+class AttnBlock(nn.Module):
+ def __init__(self, in_channels):
+ super().__init__()
+ self.in_channels = in_channels
+
+ self.norm = Normalize(in_channels)
+ self.q = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.k = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.v = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.proj_out = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+
+
+ def forward(self, x):
+ h_ = x
+ h_ = self.norm(h_)
+ q = self.q(h_)
+ k = self.k(h_)
+ v = self.v(h_)
+
+ # compute attention
+ b,c,h,w = q.shape
+ q = q.reshape(b,c,h*w)
+ q = q.permute(0,2,1) # b,hw,c
+ k = k.reshape(b,c,h*w) # b,c,hw
+ w_ = torch.bmm(q,k) # b,hw,hw w[b,i,j]=sum_c q[b,i,c]k[b,c,j]
+ w_ = w_ * (int(c)**(-0.5))
+ w_ = torch.nn.functional.softmax(w_, dim=2)
+
+ # attend to values
+ v = v.reshape(b,c,h*w)
+ w_ = w_.permute(0,2,1) # b,hw,hw (first hw of k, second of q)
+ h_ = torch.bmm(v,w_) # b, c,hw (hw of q) h_[b,c,j] = sum_i v[b,c,i] w_[b,i,j]
+ h_ = h_.reshape(b,c,h,w)
+
+ h_ = self.proj_out(h_)
+
+ return x+h_
+
+
+def make_attn(in_channels, attn_type="vanilla"):
+ assert attn_type in ["vanilla", "linear", "none"], f'attn_type {attn_type} unknown'
+ print(f"making attention of type '{attn_type}' with {in_channels} in_channels")
+ if attn_type == "vanilla":
+ return AttnBlock(in_channels)
+ elif attn_type == "none":
+ return nn.Identity(in_channels)
+ else:
+ return LinAttnBlock(in_channels)
+
+
+class Model(nn.Module):
+ def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+ resolution, use_timestep=True, use_linear_attn=False, attn_type="vanilla"):
+ super().__init__()
+ if use_linear_attn: attn_type = "linear"
+ self.ch = ch
+ self.temb_ch = self.ch*4
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ self.resolution = resolution
+ self.in_channels = in_channels
+
+ self.use_timestep = use_timestep
+ if self.use_timestep:
+ # timestep embedding
+ self.temb = nn.Module()
+ self.temb.dense = nn.ModuleList([
+ torch.nn.Linear(self.ch,
+ self.temb_ch),
+ torch.nn.Linear(self.temb_ch,
+ self.temb_ch),
+ ])
+
+ # downsampling
+ self.conv_in = torch.nn.Conv2d(in_channels,
+ self.ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ curr_res = resolution
+ in_ch_mult = (1,)+tuple(ch_mult)
+ self.down = nn.ModuleList()
+ for i_level in range(self.num_resolutions):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_in = ch*in_ch_mult[i_level]
+ block_out = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks):
+ block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(make_attn(block_in, attn_type=attn_type))
+ down = nn.Module()
+ down.block = block
+ down.attn = attn
+ if i_level != self.num_resolutions-1:
+ down.downsample = Downsample(block_in, resamp_with_conv)
+ curr_res = curr_res // 2
+ self.down.append(down)
+
+ # middle
+ self.mid = nn.Module()
+ self.mid.block_1 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+ self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
+ self.mid.block_2 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+
+ # upsampling
+ self.up = nn.ModuleList()
+ for i_level in reversed(range(self.num_resolutions)):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_out = ch*ch_mult[i_level]
+ skip_in = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks+1):
+ if i_block == self.num_res_blocks:
+ skip_in = ch*in_ch_mult[i_level]
+ block.append(ResnetBlock(in_channels=block_in+skip_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(make_attn(block_in, attn_type=attn_type))
+ up = nn.Module()
+ up.block = block
+ up.attn = attn
+ if i_level != 0:
+ up.upsample = Upsample(block_in, resamp_with_conv)
+ curr_res = curr_res * 2
+ self.up.insert(0, up) # prepend to get consistent order
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ out_ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x, t=None, context=None):
+ #assert x.shape[2] == x.shape[3] == self.resolution
+ if context is not None:
+ # assume aligned context, cat along channel axis
+ x = torch.cat((x, context), dim=1)
+ if self.use_timestep:
+ # timestep embedding
+ assert t is not None
+ temb = get_timestep_embedding(t, self.ch)
+ temb = self.temb.dense[0](temb)
+ temb = nonlinearity(temb)
+ temb = self.temb.dense[1](temb)
+ else:
+ temb = None
+
+ # downsampling
+ hs = [self.conv_in(x)]
+ for i_level in range(self.num_resolutions):
+ for i_block in range(self.num_res_blocks):
+ h = self.down[i_level].block[i_block](hs[-1], temb)
+ if len(self.down[i_level].attn) > 0:
+ h = self.down[i_level].attn[i_block](h)
+ hs.append(h)
+ if i_level != self.num_resolutions-1:
+ hs.append(self.down[i_level].downsample(hs[-1]))
+
+ # middle
+ h = hs[-1]
+ h = self.mid.block_1(h, temb)
+ h = self.mid.attn_1(h)
+ h = self.mid.block_2(h, temb)
+
+ # upsampling
+ for i_level in reversed(range(self.num_resolutions)):
+ for i_block in range(self.num_res_blocks+1):
+ h = self.up[i_level].block[i_block](
+ torch.cat([h, hs.pop()], dim=1), temb)
+ if len(self.up[i_level].attn) > 0:
+ h = self.up[i_level].attn[i_block](h)
+ if i_level != 0:
+ h = self.up[i_level].upsample(h)
+
+ # end
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
+ def get_last_layer(self):
+ return self.conv_out.weight
+
+
+class Encoder(nn.Module):
+ def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+ resolution, z_channels, double_z=True, use_linear_attn=False, attn_type="vanilla",
+ **ignore_kwargs):
+ super().__init__()
+ if use_linear_attn: attn_type = "linear"
+ self.ch = ch
+ self.temb_ch = 0
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ self.resolution = resolution
+ self.in_channels = in_channels
+
+ # downsampling
+ self.conv_in = torch.nn.Conv2d(in_channels,
+ self.ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ curr_res = resolution
+ in_ch_mult = (1,)+tuple(ch_mult)
+ self.in_ch_mult = in_ch_mult
+ self.down = nn.ModuleList()
+ for i_level in range(self.num_resolutions):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_in = ch*in_ch_mult[i_level]
+ block_out = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks):
+ block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(make_attn(block_in, attn_type=attn_type))
+ down = nn.Module()
+ down.block = block
+ down.attn = attn
+ if i_level != self.num_resolutions-1:
+ down.downsample = Downsample(block_in, resamp_with_conv)
+ curr_res = curr_res // 2
+ self.down.append(down)
+
+ # middle
+ self.mid = nn.Module()
+ self.mid.block_1 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+ self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
+ self.mid.block_2 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ 2*z_channels if double_z else z_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x):
+ # timestep embedding
+ temb = None
+
+ # downsampling
+ hs = [self.conv_in(x)]
+ for i_level in range(self.num_resolutions):
+ for i_block in range(self.num_res_blocks):
+ h = self.down[i_level].block[i_block](hs[-1], temb)
+ if len(self.down[i_level].attn) > 0:
+ h = self.down[i_level].attn[i_block](h)
+ hs.append(h)
+ if i_level != self.num_resolutions-1:
+ hs.append(self.down[i_level].downsample(hs[-1]))
+
+ # middle
+ h = hs[-1]
+ h = self.mid.block_1(h, temb)
+ h = self.mid.attn_1(h)
+ h = self.mid.block_2(h, temb)
+
+ # end
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
+
+class Decoder(nn.Module):
+ def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+ resolution, z_channels, give_pre_end=False, tanh_out=False, use_linear_attn=False,
+ attn_type="vanilla", **ignorekwargs):
+ super().__init__()
+ if use_linear_attn: attn_type = "linear"
+ self.ch = ch
+ self.temb_ch = 0
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ self.resolution = resolution
+ self.in_channels = in_channels
+ self.give_pre_end = give_pre_end
+ self.tanh_out = tanh_out
+
+ # compute in_ch_mult, block_in and curr_res at lowest res
+ in_ch_mult = (1,)+tuple(ch_mult)
+ block_in = ch*ch_mult[self.num_resolutions-1]
+ curr_res = resolution // 2**(self.num_resolutions-1)
+ self.z_shape = (1,z_channels,curr_res,curr_res)
+ print("Working with z of shape {} = {} dimensions.".format(
+ self.z_shape, np.prod(self.z_shape)))
+
+ # z to block_in
+ self.conv_in = torch.nn.Conv2d(z_channels,
+ block_in,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ # middle
+ self.mid = nn.Module()
+ self.mid.block_1 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+ self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
+ self.mid.block_2 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+
+ # upsampling
+ self.up = nn.ModuleList()
+ for i_level in reversed(range(self.num_resolutions)):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_out = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks+1):
+ block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(make_attn(block_in, attn_type=attn_type))
+ up = nn.Module()
+ up.block = block
+ up.attn = attn
+ if i_level != 0:
+ up.upsample = Upsample(block_in, resamp_with_conv)
+ curr_res = curr_res * 2
+ self.up.insert(0, up) # prepend to get consistent order
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ out_ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, z):
+ #assert z.shape[1:] == self.z_shape[1:]
+ self.last_z_shape = z.shape
+
+ # timestep embedding
+ temb = None
+
+ # z to block_in
+ h = self.conv_in(z)
+
+ # middle
+ h = self.mid.block_1(h, temb)
+ h = self.mid.attn_1(h)
+ h = self.mid.block_2(h, temb)
+
+ # upsampling
+ for i_level in reversed(range(self.num_resolutions)):
+ for i_block in range(self.num_res_blocks+1):
+ h = self.up[i_level].block[i_block](h, temb)
+ if len(self.up[i_level].attn) > 0:
+ h = self.up[i_level].attn[i_block](h)
+ if i_level != 0:
+ h = self.up[i_level].upsample(h)
+
+ # end
+ if self.give_pre_end:
+ return h
+
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ if self.tanh_out:
+ h = torch.tanh(h)
+ return h
+
+
+class SimpleDecoder(nn.Module):
+ def __init__(self, in_channels, out_channels, *args, **kwargs):
+ super().__init__()
+ self.model = nn.ModuleList([nn.Conv2d(in_channels, in_channels, 1),
+ ResnetBlock(in_channels=in_channels,
+ out_channels=2 * in_channels,
+ temb_channels=0, dropout=0.0),
+ ResnetBlock(in_channels=2 * in_channels,
+ out_channels=4 * in_channels,
+ temb_channels=0, dropout=0.0),
+ ResnetBlock(in_channels=4 * in_channels,
+ out_channels=2 * in_channels,
+ temb_channels=0, dropout=0.0),
+ nn.Conv2d(2*in_channels, in_channels, 1),
+ Upsample(in_channels, with_conv=True)])
+ # end
+ self.norm_out = Normalize(in_channels)
+ self.conv_out = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x):
+ for i, layer in enumerate(self.model):
+ if i in [1,2,3]:
+ x = layer(x, None)
+ else:
+ x = layer(x)
+
+ h = self.norm_out(x)
+ h = nonlinearity(h)
+ x = self.conv_out(h)
+ return x
+
+
+class UpsampleDecoder(nn.Module):
+ def __init__(self, in_channels, out_channels, ch, num_res_blocks, resolution,
+ ch_mult=(2,2), dropout=0.0):
+ super().__init__()
+ # upsampling
+ self.temb_ch = 0
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ block_in = in_channels
+ curr_res = resolution // 2 ** (self.num_resolutions - 1)
+ self.res_blocks = nn.ModuleList()
+ self.upsample_blocks = nn.ModuleList()
+ for i_level in range(self.num_resolutions):
+ res_block = []
+ block_out = ch * ch_mult[i_level]
+ for i_block in range(self.num_res_blocks + 1):
+ res_block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ self.res_blocks.append(nn.ModuleList(res_block))
+ if i_level != self.num_resolutions - 1:
+ self.upsample_blocks.append(Upsample(block_in, True))
+ curr_res = curr_res * 2
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x):
+ # upsampling
+ h = x
+ for k, i_level in enumerate(range(self.num_resolutions)):
+ for i_block in range(self.num_res_blocks + 1):
+ h = self.res_blocks[i_level][i_block](h, None)
+ if i_level != self.num_resolutions - 1:
+ h = self.upsample_blocks[k](h)
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
+
+class LatentRescaler(nn.Module):
+ def __init__(self, factor, in_channels, mid_channels, out_channels, depth=2):
+ super().__init__()
+ # residual block, interpolate, residual block
+ self.factor = factor
+ self.conv_in = nn.Conv2d(in_channels,
+ mid_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+ self.res_block1 = nn.ModuleList([ResnetBlock(in_channels=mid_channels,
+ out_channels=mid_channels,
+ temb_channels=0,
+ dropout=0.0) for _ in range(depth)])
+ self.attn = AttnBlock(mid_channels)
+ self.res_block2 = nn.ModuleList([ResnetBlock(in_channels=mid_channels,
+ out_channels=mid_channels,
+ temb_channels=0,
+ dropout=0.0) for _ in range(depth)])
+
+ self.conv_out = nn.Conv2d(mid_channels,
+ out_channels,
+ kernel_size=1,
+ )
+
+ def forward(self, x):
+ x = self.conv_in(x)
+ for block in self.res_block1:
+ x = block(x, None)
+ x = torch.nn.functional.interpolate(x, size=(int(round(x.shape[2]*self.factor)), int(round(x.shape[3]*self.factor))))
+ x = self.attn(x)
+ for block in self.res_block2:
+ x = block(x, None)
+ x = self.conv_out(x)
+ return x
+
+
+class MergedRescaleEncoder(nn.Module):
+ def __init__(self, in_channels, ch, resolution, out_ch, num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True,
+ ch_mult=(1,2,4,8), rescale_factor=1.0, rescale_module_depth=1):
+ super().__init__()
+ intermediate_chn = ch * ch_mult[-1]
+ self.encoder = Encoder(in_channels=in_channels, num_res_blocks=num_res_blocks, ch=ch, ch_mult=ch_mult,
+ z_channels=intermediate_chn, double_z=False, resolution=resolution,
+ attn_resolutions=attn_resolutions, dropout=dropout, resamp_with_conv=resamp_with_conv,
+ out_ch=None)
+ self.rescaler = LatentRescaler(factor=rescale_factor, in_channels=intermediate_chn,
+ mid_channels=intermediate_chn, out_channels=out_ch, depth=rescale_module_depth)
+
+ def forward(self, x):
+ x = self.encoder(x)
+ x = self.rescaler(x)
+ return x
+
+
+class MergedRescaleDecoder(nn.Module):
+ def __init__(self, z_channels, out_ch, resolution, num_res_blocks, attn_resolutions, ch, ch_mult=(1,2,4,8),
+ dropout=0.0, resamp_with_conv=True, rescale_factor=1.0, rescale_module_depth=1):
+ super().__init__()
+ tmp_chn = z_channels*ch_mult[-1]
+ self.decoder = Decoder(out_ch=out_ch, z_channels=tmp_chn, attn_resolutions=attn_resolutions, dropout=dropout,
+ resamp_with_conv=resamp_with_conv, in_channels=None, num_res_blocks=num_res_blocks,
+ ch_mult=ch_mult, resolution=resolution, ch=ch)
+ self.rescaler = LatentRescaler(factor=rescale_factor, in_channels=z_channels, mid_channels=tmp_chn,
+ out_channels=tmp_chn, depth=rescale_module_depth)
+
+ def forward(self, x):
+ x = self.rescaler(x)
+ x = self.decoder(x)
+ return x
+
+
+class Upsampler(nn.Module):
+ def __init__(self, in_size, out_size, in_channels, out_channels, ch_mult=2):
+ super().__init__()
+ assert out_size >= in_size
+ num_blocks = int(np.log2(out_size//in_size))+1
+ factor_up = 1.+ (out_size % in_size)
+ print(f"Building {self.__class__.__name__} with in_size: {in_size} --> out_size {out_size} and factor {factor_up}")
+ self.rescaler = LatentRescaler(factor=factor_up, in_channels=in_channels, mid_channels=2*in_channels,
+ out_channels=in_channels)
+ self.decoder = Decoder(out_ch=out_channels, resolution=out_size, z_channels=in_channels, num_res_blocks=2,
+ attn_resolutions=[], in_channels=None, ch=in_channels,
+ ch_mult=[ch_mult for _ in range(num_blocks)])
+
+ def forward(self, x):
+ x = self.rescaler(x)
+ x = self.decoder(x)
+ return x
+
+
+class Resize(nn.Module):
+ def __init__(self, in_channels=None, learned=False, mode="bilinear"):
+ super().__init__()
+ self.with_conv = learned
+ self.mode = mode
+ if self.with_conv:
+ print(f"Note: {self.__class__.__name} uses learned downsampling and will ignore the fixed {mode} mode")
+ raise NotImplementedError()
+ assert in_channels is not None
+ # no asymmetric padding in torch conv, must do it ourselves
+ self.conv = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=4,
+ stride=2,
+ padding=1)
+
+ def forward(self, x, scale_factor=1.0):
+ if scale_factor==1.0:
+ return x
+ else:
+ x = torch.nn.functional.interpolate(x, mode=self.mode, align_corners=False, scale_factor=scale_factor)
+ return x
+
+class FirstStagePostProcessor(nn.Module):
+
+ def __init__(self, ch_mult:list, in_channels,
+ pretrained_model:nn.Module=None,
+ reshape=False,
+ n_channels=None,
+ dropout=0.,
+ pretrained_config=None):
+ super().__init__()
+ if pretrained_config is None:
+ assert pretrained_model is not None, 'Either "pretrained_model" or "pretrained_config" must not be None'
+ self.pretrained_model = pretrained_model
+ else:
+ assert pretrained_config is not None, 'Either "pretrained_model" or "pretrained_config" must not be None'
+ self.instantiate_pretrained(pretrained_config)
+
+ self.do_reshape = reshape
+
+ if n_channels is None:
+ n_channels = self.pretrained_model.encoder.ch
+
+ self.proj_norm = Normalize(in_channels,num_groups=in_channels//2)
+ self.proj = nn.Conv2d(in_channels,n_channels,kernel_size=3,
+ stride=1,padding=1)
+
+ blocks = []
+ downs = []
+ ch_in = n_channels
+ for m in ch_mult:
+ blocks.append(ResnetBlock(in_channels=ch_in,out_channels=m*n_channels,dropout=dropout))
+ ch_in = m * n_channels
+ downs.append(Downsample(ch_in, with_conv=False))
+
+ self.model = nn.ModuleList(blocks)
+ self.downsampler = nn.ModuleList(downs)
+
+
+ def instantiate_pretrained(self, config):
+ model = instantiate_from_config(config)
+ self.pretrained_model = model.eval()
+ # self.pretrained_model.train = False
+ for param in self.pretrained_model.parameters():
+ param.requires_grad = False
+
+
+ @torch.no_grad()
+ def encode_with_pretrained(self,x):
+ c = self.pretrained_model.encode(x)
+ if isinstance(c, DiagonalGaussianDistribution):
+ c = c.mode()
+ return c
+
+ def forward(self,x):
+ z_fs = self.encode_with_pretrained(x)
+ z = self.proj_norm(z_fs)
+ z = self.proj(z)
+ z = nonlinearity(z)
+
+ for submodel, downmodel in zip(self.model,self.downsampler):
+ z = submodel(z,temb=None)
+ z = downmodel(z)
+
+ if self.do_reshape:
+ z = rearrange(z,'b c h w -> b (h w) c')
+ return z
+
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/openaimodel.py b/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/openaimodel.py
new file mode 100644
index 0000000..fcf95d1
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/openaimodel.py
@@ -0,0 +1,961 @@
+from abc import abstractmethod
+from functools import partial
+import math
+from typing import Iterable
+
+import numpy as np
+import torch as th
+import torch.nn as nn
+import torch.nn.functional as F
+
+from ldm.modules.diffusionmodules.util import (
+ checkpoint,
+ conv_nd,
+ linear,
+ avg_pool_nd,
+ zero_module,
+ normalization,
+ timestep_embedding,
+)
+from ldm.modules.attention import SpatialTransformer
+
+
+# dummy replace
+def convert_module_to_f16(x):
+ pass
+
+def convert_module_to_f32(x):
+ pass
+
+
+## go
+class AttentionPool2d(nn.Module):
+ """
+ Adapted from CLIP: https://github.com/openai/CLIP/blob/main/clip/model.py
+ """
+
+ def __init__(
+ self,
+ spacial_dim: int,
+ embed_dim: int,
+ num_heads_channels: int,
+ output_dim: int = None,
+ ):
+ super().__init__()
+ self.positional_embedding = nn.Parameter(th.randn(embed_dim, spacial_dim ** 2 + 1) / embed_dim ** 0.5)
+ self.qkv_proj = conv_nd(1, embed_dim, 3 * embed_dim, 1)
+ self.c_proj = conv_nd(1, embed_dim, output_dim or embed_dim, 1)
+ self.num_heads = embed_dim // num_heads_channels
+ self.attention = QKVAttention(self.num_heads)
+
+ def forward(self, x):
+ b, c, *_spatial = x.shape
+ x = x.reshape(b, c, -1) # NC(HW)
+ x = th.cat([x.mean(dim=-1, keepdim=True), x], dim=-1) # NC(HW+1)
+ x = x + self.positional_embedding[None, :, :].to(x.dtype) # NC(HW+1)
+ x = self.qkv_proj(x)
+ x = self.attention(x)
+ x = self.c_proj(x)
+ return x[:, :, 0]
+
+
+class TimestepBlock(nn.Module):
+ """
+ Any module where forward() takes timestep embeddings as a second argument.
+ """
+
+ @abstractmethod
+ def forward(self, x, emb):
+ """
+ Apply the module to `x` given `emb` timestep embeddings.
+ """
+
+
+class TimestepEmbedSequential(nn.Sequential, TimestepBlock):
+ """
+ A sequential module that passes timestep embeddings to the children that
+ support it as an extra input.
+ """
+
+ def forward(self, x, emb, context=None):
+ for layer in self:
+ if isinstance(layer, TimestepBlock):
+ x = layer(x, emb)
+ elif isinstance(layer, SpatialTransformer):
+ x = layer(x, context)
+ else:
+ x = layer(x)
+ return x
+
+
+class Upsample(nn.Module):
+ """
+ An upsampling layer with an optional convolution.
+ :param channels: channels in the inputs and outputs.
+ :param use_conv: a bool determining if a convolution is applied.
+ :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+ upsampling occurs in the inner-two dimensions.
+ """
+
+ def __init__(self, channels, use_conv, dims=2, out_channels=None, padding=1):
+ super().__init__()
+ self.channels = channels
+ self.out_channels = out_channels or channels
+ self.use_conv = use_conv
+ self.dims = dims
+ if use_conv:
+ self.conv = conv_nd(dims, self.channels, self.out_channels, 3, padding=padding)
+
+ def forward(self, x):
+ assert x.shape[1] == self.channels
+ if self.dims == 3:
+ x = F.interpolate(
+ x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest"
+ )
+ else:
+ x = F.interpolate(x, scale_factor=2, mode="nearest")
+ if self.use_conv:
+ x = self.conv(x)
+ return x
+
+class TransposedUpsample(nn.Module):
+ 'Learned 2x upsampling without padding'
+ def __init__(self, channels, out_channels=None, ks=5):
+ super().__init__()
+ self.channels = channels
+ self.out_channels = out_channels or channels
+
+ self.up = nn.ConvTranspose2d(self.channels,self.out_channels,kernel_size=ks,stride=2)
+
+ def forward(self,x):
+ return self.up(x)
+
+
+class Downsample(nn.Module):
+ """
+ A downsampling layer with an optional convolution.
+ :param channels: channels in the inputs and outputs.
+ :param use_conv: a bool determining if a convolution is applied.
+ :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+ downsampling occurs in the inner-two dimensions.
+ """
+
+ def __init__(self, channels, use_conv, dims=2, out_channels=None,padding=1):
+ super().__init__()
+ self.channels = channels
+ self.out_channels = out_channels or channels
+ self.use_conv = use_conv
+ self.dims = dims
+ stride = 2 if dims != 3 else (1, 2, 2)
+ if use_conv:
+ self.op = conv_nd(
+ dims, self.channels, self.out_channels, 3, stride=stride, padding=padding
+ )
+ else:
+ assert self.channels == self.out_channels
+ self.op = avg_pool_nd(dims, kernel_size=stride, stride=stride)
+
+ def forward(self, x):
+ assert x.shape[1] == self.channels
+ return self.op(x)
+
+
+class ResBlock(TimestepBlock):
+ """
+ A residual block that can optionally change the number of channels.
+ :param channels: the number of input channels.
+ :param emb_channels: the number of timestep embedding channels.
+ :param dropout: the rate of dropout.
+ :param out_channels: if specified, the number of out channels.
+ :param use_conv: if True and out_channels is specified, use a spatial
+ convolution instead of a smaller 1x1 convolution to change the
+ channels in the skip connection.
+ :param dims: determines if the signal is 1D, 2D, or 3D.
+ :param use_checkpoint: if True, use gradient checkpointing on this module.
+ :param up: if True, use this block for upsampling.
+ :param down: if True, use this block for downsampling.
+ """
+
+ def __init__(
+ self,
+ channels,
+ emb_channels,
+ dropout,
+ out_channels=None,
+ use_conv=False,
+ use_scale_shift_norm=False,
+ dims=2,
+ use_checkpoint=False,
+ up=False,
+ down=False,
+ ):
+ super().__init__()
+ self.channels = channels
+ self.emb_channels = emb_channels
+ self.dropout = dropout
+ self.out_channels = out_channels or channels
+ self.use_conv = use_conv
+ self.use_checkpoint = use_checkpoint
+ self.use_scale_shift_norm = use_scale_shift_norm
+
+ self.in_layers = nn.Sequential(
+ normalization(channels),
+ nn.SiLU(),
+ conv_nd(dims, channels, self.out_channels, 3, padding=1),
+ )
+
+ self.updown = up or down
+
+ if up:
+ self.h_upd = Upsample(channels, False, dims)
+ self.x_upd = Upsample(channels, False, dims)
+ elif down:
+ self.h_upd = Downsample(channels, False, dims)
+ self.x_upd = Downsample(channels, False, dims)
+ else:
+ self.h_upd = self.x_upd = nn.Identity()
+
+ self.emb_layers = nn.Sequential(
+ nn.SiLU(),
+ linear(
+ emb_channels,
+ 2 * self.out_channels if use_scale_shift_norm else self.out_channels,
+ ),
+ )
+ self.out_layers = nn.Sequential(
+ normalization(self.out_channels),
+ nn.SiLU(),
+ nn.Dropout(p=dropout),
+ zero_module(
+ conv_nd(dims, self.out_channels, self.out_channels, 3, padding=1)
+ ),
+ )
+
+ if self.out_channels == channels:
+ self.skip_connection = nn.Identity()
+ elif use_conv:
+ self.skip_connection = conv_nd(
+ dims, channels, self.out_channels, 3, padding=1
+ )
+ else:
+ self.skip_connection = conv_nd(dims, channels, self.out_channels, 1)
+
+ def forward(self, x, emb):
+ """
+ Apply the block to a Tensor, conditioned on a timestep embedding.
+ :param x: an [N x C x ...] Tensor of features.
+ :param emb: an [N x emb_channels] Tensor of timestep embeddings.
+ :return: an [N x C x ...] Tensor of outputs.
+ """
+ return checkpoint(
+ self._forward, (x, emb), self.parameters(), self.use_checkpoint
+ )
+
+
+ def _forward(self, x, emb):
+ if self.updown:
+ in_rest, in_conv = self.in_layers[:-1], self.in_layers[-1]
+ h = in_rest(x)
+ h = self.h_upd(h)
+ x = self.x_upd(x)
+ h = in_conv(h)
+ else:
+ h = self.in_layers(x)
+ emb_out = self.emb_layers(emb).type(h.dtype)
+ while len(emb_out.shape) < len(h.shape):
+ emb_out = emb_out[..., None]
+ if self.use_scale_shift_norm:
+ out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
+ scale, shift = th.chunk(emb_out, 2, dim=1)
+ h = out_norm(h) * (1 + scale) + shift
+ h = out_rest(h)
+ else:
+ h = h + emb_out
+ h = self.out_layers(h)
+ return self.skip_connection(x) + h
+
+
+class AttentionBlock(nn.Module):
+ """
+ An attention block that allows spatial positions to attend to each other.
+ Originally ported from here, but adapted to the N-d case.
+ https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/models/unet.py#L66.
+ """
+
+ def __init__(
+ self,
+ channels,
+ num_heads=1,
+ num_head_channels=-1,
+ use_checkpoint=False,
+ use_new_attention_order=False,
+ ):
+ super().__init__()
+ self.channels = channels
+ if num_head_channels == -1:
+ self.num_heads = num_heads
+ else:
+ assert (
+ channels % num_head_channels == 0
+ ), f"q,k,v channels {channels} is not divisible by num_head_channels {num_head_channels}"
+ self.num_heads = channels // num_head_channels
+ self.use_checkpoint = use_checkpoint
+ self.norm = normalization(channels)
+ self.qkv = conv_nd(1, channels, channels * 3, 1)
+ if use_new_attention_order:
+ # split qkv before split heads
+ self.attention = QKVAttention(self.num_heads)
+ else:
+ # split heads before split qkv
+ self.attention = QKVAttentionLegacy(self.num_heads)
+
+ self.proj_out = zero_module(conv_nd(1, channels, channels, 1))
+
+ def forward(self, x):
+ return checkpoint(self._forward, (x,), self.parameters(), True) # TODO: check checkpoint usage, is True # TODO: fix the .half call!!!
+ #return pt_checkpoint(self._forward, x) # pytorch
+
+ def _forward(self, x):
+ b, c, *spatial = x.shape
+ x = x.reshape(b, c, -1)
+ qkv = self.qkv(self.norm(x))
+ h = self.attention(qkv)
+ h = self.proj_out(h)
+ return (x + h).reshape(b, c, *spatial)
+
+
+def count_flops_attn(model, _x, y):
+ """
+ A counter for the `thop` package to count the operations in an
+ attention operation.
+ Meant to be used like:
+ macs, params = thop.profile(
+ model,
+ inputs=(inputs, timestamps),
+ custom_ops={QKVAttention: QKVAttention.count_flops},
+ )
+ """
+ b, c, *spatial = y[0].shape
+ num_spatial = int(np.prod(spatial))
+ # We perform two matmuls with the same number of ops.
+ # The first computes the weight matrix, the second computes
+ # the combination of the value vectors.
+ matmul_ops = 2 * b * (num_spatial ** 2) * c
+ model.total_ops += th.DoubleTensor([matmul_ops])
+
+
+class QKVAttentionLegacy(nn.Module):
+ """
+ A module which performs QKV attention. Matches legacy QKVAttention + input/ouput heads shaping
+ """
+
+ def __init__(self, n_heads):
+ super().__init__()
+ self.n_heads = n_heads
+
+ def forward(self, qkv):
+ """
+ Apply QKV attention.
+ :param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs.
+ :return: an [N x (H * C) x T] tensor after attention.
+ """
+ bs, width, length = qkv.shape
+ assert width % (3 * self.n_heads) == 0
+ ch = width // (3 * self.n_heads)
+ q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1)
+ scale = 1 / math.sqrt(math.sqrt(ch))
+ weight = th.einsum(
+ "bct,bcs->bts", q * scale, k * scale
+ ) # More stable with f16 than dividing afterwards
+ weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
+ a = th.einsum("bts,bcs->bct", weight, v)
+ return a.reshape(bs, -1, length)
+
+ @staticmethod
+ def count_flops(model, _x, y):
+ return count_flops_attn(model, _x, y)
+
+
+class QKVAttention(nn.Module):
+ """
+ A module which performs QKV attention and splits in a different order.
+ """
+
+ def __init__(self, n_heads):
+ super().__init__()
+ self.n_heads = n_heads
+
+ def forward(self, qkv):
+ """
+ Apply QKV attention.
+ :param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs.
+ :return: an [N x (H * C) x T] tensor after attention.
+ """
+ bs, width, length = qkv.shape
+ assert width % (3 * self.n_heads) == 0
+ ch = width // (3 * self.n_heads)
+ q, k, v = qkv.chunk(3, dim=1)
+ scale = 1 / math.sqrt(math.sqrt(ch))
+ weight = th.einsum(
+ "bct,bcs->bts",
+ (q * scale).view(bs * self.n_heads, ch, length),
+ (k * scale).view(bs * self.n_heads, ch, length),
+ ) # More stable with f16 than dividing afterwards
+ weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
+ a = th.einsum("bts,bcs->bct", weight, v.reshape(bs * self.n_heads, ch, length))
+ return a.reshape(bs, -1, length)
+
+ @staticmethod
+ def count_flops(model, _x, y):
+ return count_flops_attn(model, _x, y)
+
+
+class UNetModel(nn.Module):
+ """
+ The full UNet model with attention and timestep embedding.
+ :param in_channels: channels in the input Tensor.
+ :param model_channels: base channel count for the model.
+ :param out_channels: channels in the output Tensor.
+ :param num_res_blocks: number of residual blocks per downsample.
+ :param attention_resolutions: a collection of downsample rates at which
+ attention will take place. May be a set, list, or tuple.
+ For example, if this contains 4, then at 4x downsampling, attention
+ will be used.
+ :param dropout: the dropout probability.
+ :param channel_mult: channel multiplier for each level of the UNet.
+ :param conv_resample: if True, use learned convolutions for upsampling and
+ downsampling.
+ :param dims: determines if the signal is 1D, 2D, or 3D.
+ :param num_classes: if specified (as an int), then this model will be
+ class-conditional with `num_classes` classes.
+ :param use_checkpoint: use gradient checkpointing to reduce memory usage.
+ :param num_heads: the number of attention heads in each attention layer.
+ :param num_heads_channels: if specified, ignore num_heads and instead use
+ a fixed channel width per attention head.
+ :param num_heads_upsample: works with num_heads to set a different number
+ of heads for upsampling. Deprecated.
+ :param use_scale_shift_norm: use a FiLM-like conditioning mechanism.
+ :param resblock_updown: use residual blocks for up/downsampling.
+ :param use_new_attention_order: use a different attention pattern for potentially
+ increased efficiency.
+ """
+
+ def __init__(
+ self,
+ image_size,
+ in_channels,
+ model_channels,
+ out_channels,
+ num_res_blocks,
+ attention_resolutions,
+ dropout=0,
+ channel_mult=(1, 2, 4, 8),
+ conv_resample=True,
+ dims=2,
+ num_classes=None,
+ use_checkpoint=False,
+ use_fp16=False,
+ num_heads=-1,
+ num_head_channels=-1,
+ num_heads_upsample=-1,
+ use_scale_shift_norm=False,
+ resblock_updown=False,
+ use_new_attention_order=False,
+ use_spatial_transformer=False, # custom transformer support
+ transformer_depth=1, # custom transformer support
+ context_dim=None, # custom transformer support
+ n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model
+ legacy=True,
+ ):
+ super().__init__()
+ if use_spatial_transformer:
+ assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...'
+
+ if context_dim is not None:
+ assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...'
+ from omegaconf.listconfig import ListConfig
+ if type(context_dim) == ListConfig:
+ context_dim = list(context_dim)
+
+ if num_heads_upsample == -1:
+ num_heads_upsample = num_heads
+
+ if num_heads == -1:
+ assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set'
+
+ if num_head_channels == -1:
+ assert num_heads != -1, 'Either num_heads or num_head_channels has to be set'
+
+ self.image_size = image_size
+ self.in_channels = in_channels
+ self.model_channels = model_channels
+ self.out_channels = out_channels
+ self.num_res_blocks = num_res_blocks
+ self.attention_resolutions = attention_resolutions
+ self.dropout = dropout
+ self.channel_mult = channel_mult
+ self.conv_resample = conv_resample
+ self.num_classes = num_classes
+ self.use_checkpoint = use_checkpoint
+ self.dtype = th.float16 if use_fp16 else th.float32
+ self.num_heads = num_heads
+ self.num_head_channels = num_head_channels
+ self.num_heads_upsample = num_heads_upsample
+ self.predict_codebook_ids = n_embed is not None
+
+ time_embed_dim = model_channels * 4
+ self.time_embed = nn.Sequential(
+ linear(model_channels, time_embed_dim),
+ nn.SiLU(),
+ linear(time_embed_dim, time_embed_dim),
+ )
+
+ if self.num_classes is not None:
+ self.label_emb = nn.Embedding(num_classes, time_embed_dim)
+
+ self.input_blocks = nn.ModuleList(
+ [
+ TimestepEmbedSequential(
+ conv_nd(dims, in_channels, model_channels, 3, padding=1)
+ )
+ ]
+ )
+ self._feature_size = model_channels
+ input_block_chans = [model_channels]
+ ch = model_channels
+ ds = 1
+ for level, mult in enumerate(channel_mult):
+ for _ in range(num_res_blocks):
+ layers = [
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ out_channels=mult * model_channels,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ )
+ ]
+ ch = mult * model_channels
+ if ds in attention_resolutions:
+ if num_head_channels == -1:
+ dim_head = ch // num_heads
+ else:
+ num_heads = ch // num_head_channels
+ dim_head = num_head_channels
+ if legacy:
+ #num_heads = 1
+ dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+ layers.append(
+ AttentionBlock(
+ ch,
+ use_checkpoint=use_checkpoint,
+ num_heads=num_heads,
+ num_head_channels=dim_head,
+ use_new_attention_order=use_new_attention_order,
+ ) if not use_spatial_transformer else SpatialTransformer(
+ ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim
+ )
+ )
+ self.input_blocks.append(TimestepEmbedSequential(*layers))
+ self._feature_size += ch
+ input_block_chans.append(ch)
+ if level != len(channel_mult) - 1:
+ out_ch = ch
+ self.input_blocks.append(
+ TimestepEmbedSequential(
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ out_channels=out_ch,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ down=True,
+ )
+ if resblock_updown
+ else Downsample(
+ ch, conv_resample, dims=dims, out_channels=out_ch
+ )
+ )
+ )
+ ch = out_ch
+ input_block_chans.append(ch)
+ ds *= 2
+ self._feature_size += ch
+
+ if num_head_channels == -1:
+ dim_head = ch // num_heads
+ else:
+ num_heads = ch // num_head_channels
+ dim_head = num_head_channels
+ if legacy:
+ #num_heads = 1
+ dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+ self.middle_block = TimestepEmbedSequential(
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ ),
+ AttentionBlock(
+ ch,
+ use_checkpoint=use_checkpoint,
+ num_heads=num_heads,
+ num_head_channels=dim_head,
+ use_new_attention_order=use_new_attention_order,
+ ) if not use_spatial_transformer else SpatialTransformer(
+ ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim
+ ),
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ ),
+ )
+ self._feature_size += ch
+
+ self.output_blocks = nn.ModuleList([])
+ for level, mult in list(enumerate(channel_mult))[::-1]:
+ for i in range(num_res_blocks + 1):
+ ich = input_block_chans.pop()
+ layers = [
+ ResBlock(
+ ch + ich,
+ time_embed_dim,
+ dropout,
+ out_channels=model_channels * mult,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ )
+ ]
+ ch = model_channels * mult
+ if ds in attention_resolutions:
+ if num_head_channels == -1:
+ dim_head = ch // num_heads
+ else:
+ num_heads = ch // num_head_channels
+ dim_head = num_head_channels
+ if legacy:
+ #num_heads = 1
+ dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+ layers.append(
+ AttentionBlock(
+ ch,
+ use_checkpoint=use_checkpoint,
+ num_heads=num_heads_upsample,
+ num_head_channels=dim_head,
+ use_new_attention_order=use_new_attention_order,
+ ) if not use_spatial_transformer else SpatialTransformer(
+ ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim
+ )
+ )
+ if level and i == num_res_blocks:
+ out_ch = ch
+ layers.append(
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ out_channels=out_ch,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ up=True,
+ )
+ if resblock_updown
+ else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch)
+ )
+ ds //= 2
+ self.output_blocks.append(TimestepEmbedSequential(*layers))
+ self._feature_size += ch
+
+ self.out = nn.Sequential(
+ normalization(ch),
+ nn.SiLU(),
+ zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)),
+ )
+ if self.predict_codebook_ids:
+ self.id_predictor = nn.Sequential(
+ normalization(ch),
+ conv_nd(dims, model_channels, n_embed, 1),
+ #nn.LogSoftmax(dim=1) # change to cross_entropy and produce non-normalized logits
+ )
+
+ def convert_to_fp16(self):
+ """
+ Convert the torso of the model to float16.
+ """
+ self.input_blocks.apply(convert_module_to_f16)
+ self.middle_block.apply(convert_module_to_f16)
+ self.output_blocks.apply(convert_module_to_f16)
+
+ def convert_to_fp32(self):
+ """
+ Convert the torso of the model to float32.
+ """
+ self.input_blocks.apply(convert_module_to_f32)
+ self.middle_block.apply(convert_module_to_f32)
+ self.output_blocks.apply(convert_module_to_f32)
+
+ def forward(self, x, timesteps=None, context=None, y=None,**kwargs):
+ """
+ Apply the model to an input batch.
+ :param x: an [N x C x ...] Tensor of inputs.
+ :param timesteps: a 1-D batch of timesteps.
+ :param context: conditioning plugged in via crossattn
+ :param y: an [N] Tensor of labels, if class-conditional.
+ :return: an [N x C x ...] Tensor of outputs.
+ """
+ assert (y is not None) == (
+ self.num_classes is not None
+ ), "must specify y if and only if the model is class-conditional"
+ hs = []
+ t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+ emb = self.time_embed(t_emb)
+
+ if self.num_classes is not None:
+ assert y.shape == (x.shape[0],)
+ emb = emb + self.label_emb(y)
+
+ h = x.type(self.dtype)
+ for module in self.input_blocks:
+ h = module(h, emb, context)
+ hs.append(h)
+ h = self.middle_block(h, emb, context)
+ for module in self.output_blocks:
+ h = th.cat([h, hs.pop()], dim=1)
+ h = module(h, emb, context)
+ h = h.type(x.dtype)
+ if self.predict_codebook_ids:
+ return self.id_predictor(h)
+ else:
+ return self.out(h)
+
+
+class EncoderUNetModel(nn.Module):
+ """
+ The half UNet model with attention and timestep embedding.
+ For usage, see UNet.
+ """
+
+ def __init__(
+ self,
+ image_size,
+ in_channels,
+ model_channels,
+ out_channels,
+ num_res_blocks,
+ attention_resolutions,
+ dropout=0,
+ channel_mult=(1, 2, 4, 8),
+ conv_resample=True,
+ dims=2,
+ use_checkpoint=False,
+ use_fp16=False,
+ num_heads=1,
+ num_head_channels=-1,
+ num_heads_upsample=-1,
+ use_scale_shift_norm=False,
+ resblock_updown=False,
+ use_new_attention_order=False,
+ pool="adaptive",
+ *args,
+ **kwargs
+ ):
+ super().__init__()
+
+ if num_heads_upsample == -1:
+ num_heads_upsample = num_heads
+
+ self.in_channels = in_channels
+ self.model_channels = model_channels
+ self.out_channels = out_channels
+ self.num_res_blocks = num_res_blocks
+ self.attention_resolutions = attention_resolutions
+ self.dropout = dropout
+ self.channel_mult = channel_mult
+ self.conv_resample = conv_resample
+ self.use_checkpoint = use_checkpoint
+ self.dtype = th.float16 if use_fp16 else th.float32
+ self.num_heads = num_heads
+ self.num_head_channels = num_head_channels
+ self.num_heads_upsample = num_heads_upsample
+
+ time_embed_dim = model_channels * 4
+ self.time_embed = nn.Sequential(
+ linear(model_channels, time_embed_dim),
+ nn.SiLU(),
+ linear(time_embed_dim, time_embed_dim),
+ )
+
+ self.input_blocks = nn.ModuleList(
+ [
+ TimestepEmbedSequential(
+ conv_nd(dims, in_channels, model_channels, 3, padding=1)
+ )
+ ]
+ )
+ self._feature_size = model_channels
+ input_block_chans = [model_channels]
+ ch = model_channels
+ ds = 1
+ for level, mult in enumerate(channel_mult):
+ for _ in range(num_res_blocks):
+ layers = [
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ out_channels=mult * model_channels,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ )
+ ]
+ ch = mult * model_channels
+ if ds in attention_resolutions:
+ layers.append(
+ AttentionBlock(
+ ch,
+ use_checkpoint=use_checkpoint,
+ num_heads=num_heads,
+ num_head_channels=num_head_channels,
+ use_new_attention_order=use_new_attention_order,
+ )
+ )
+ self.input_blocks.append(TimestepEmbedSequential(*layers))
+ self._feature_size += ch
+ input_block_chans.append(ch)
+ if level != len(channel_mult) - 1:
+ out_ch = ch
+ self.input_blocks.append(
+ TimestepEmbedSequential(
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ out_channels=out_ch,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ down=True,
+ )
+ if resblock_updown
+ else Downsample(
+ ch, conv_resample, dims=dims, out_channels=out_ch
+ )
+ )
+ )
+ ch = out_ch
+ input_block_chans.append(ch)
+ ds *= 2
+ self._feature_size += ch
+
+ self.middle_block = TimestepEmbedSequential(
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ ),
+ AttentionBlock(
+ ch,
+ use_checkpoint=use_checkpoint,
+ num_heads=num_heads,
+ num_head_channels=num_head_channels,
+ use_new_attention_order=use_new_attention_order,
+ ),
+ ResBlock(
+ ch,
+ time_embed_dim,
+ dropout,
+ dims=dims,
+ use_checkpoint=use_checkpoint,
+ use_scale_shift_norm=use_scale_shift_norm,
+ ),
+ )
+ self._feature_size += ch
+ self.pool = pool
+ if pool == "adaptive":
+ self.out = nn.Sequential(
+ normalization(ch),
+ nn.SiLU(),
+ nn.AdaptiveAvgPool2d((1, 1)),
+ zero_module(conv_nd(dims, ch, out_channels, 1)),
+ nn.Flatten(),
+ )
+ elif pool == "attention":
+ assert num_head_channels != -1
+ self.out = nn.Sequential(
+ normalization(ch),
+ nn.SiLU(),
+ AttentionPool2d(
+ (image_size // ds), ch, num_head_channels, out_channels
+ ),
+ )
+ elif pool == "spatial":
+ self.out = nn.Sequential(
+ nn.Linear(self._feature_size, 2048),
+ nn.ReLU(),
+ nn.Linear(2048, self.out_channels),
+ )
+ elif pool == "spatial_v2":
+ self.out = nn.Sequential(
+ nn.Linear(self._feature_size, 2048),
+ normalization(2048),
+ nn.SiLU(),
+ nn.Linear(2048, self.out_channels),
+ )
+ else:
+ raise NotImplementedError(f"Unexpected {pool} pooling")
+
+ def convert_to_fp16(self):
+ """
+ Convert the torso of the model to float16.
+ """
+ self.input_blocks.apply(convert_module_to_f16)
+ self.middle_block.apply(convert_module_to_f16)
+
+ def convert_to_fp32(self):
+ """
+ Convert the torso of the model to float32.
+ """
+ self.input_blocks.apply(convert_module_to_f32)
+ self.middle_block.apply(convert_module_to_f32)
+
+ def forward(self, x, timesteps):
+ """
+ Apply the model to an input batch.
+ :param x: an [N x C x ...] Tensor of inputs.
+ :param timesteps: a 1-D batch of timesteps.
+ :return: an [N x K] Tensor of outputs.
+ """
+ emb = self.time_embed(timestep_embedding(timesteps, self.model_channels))
+
+ results = []
+ h = x.type(self.dtype)
+ for module in self.input_blocks:
+ h = module(h, emb)
+ if self.pool.startswith("spatial"):
+ results.append(h.type(x.dtype).mean(dim=(2, 3)))
+ h = self.middle_block(h, emb)
+ if self.pool.startswith("spatial"):
+ results.append(h.type(x.dtype).mean(dim=(2, 3)))
+ h = th.cat(results, axis=-1)
+ return self.out(h)
+ else:
+ h = h.type(x.dtype)
+ return self.out(h)
+
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/util.py b/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/util.py
new file mode 100644
index 0000000..a952e6c
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/diffusionmodules/util.py
@@ -0,0 +1,267 @@
+# adopted from
+# https://github.com/openai/improved-diffusion/blob/main/improved_diffusion/gaussian_diffusion.py
+# and
+# https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+# and
+# https://github.com/openai/guided-diffusion/blob/0ba878e517b276c45d1195eb29f6f5f72659a05b/guided_diffusion/nn.py
+#
+# thanks!
+
+
+import os
+import math
+import torch
+import torch.nn as nn
+import numpy as np
+from einops import repeat
+
+from ldm.util import instantiate_from_config
+
+
+def make_beta_schedule(schedule, n_timestep, linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+ if schedule == "linear":
+ betas = (
+ torch.linspace(linear_start ** 0.5, linear_end ** 0.5, n_timestep, dtype=torch.float64) ** 2
+ )
+
+ elif schedule == "cosine":
+ timesteps = (
+ torch.arange(n_timestep + 1, dtype=torch.float64) / n_timestep + cosine_s
+ )
+ alphas = timesteps / (1 + cosine_s) * np.pi / 2
+ alphas = torch.cos(alphas).pow(2)
+ alphas = alphas / alphas[0]
+ betas = 1 - alphas[1:] / alphas[:-1]
+ betas = np.clip(betas, a_min=0, a_max=0.999)
+
+ elif schedule == "sqrt_linear":
+ betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64)
+ elif schedule == "sqrt":
+ betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64) ** 0.5
+ else:
+ raise ValueError(f"schedule '{schedule}' unknown.")
+ return betas.numpy()
+
+
+def make_ddim_timesteps(ddim_discr_method, num_ddim_timesteps, num_ddpm_timesteps, verbose=True):
+ if ddim_discr_method == 'uniform':
+ c = num_ddpm_timesteps // num_ddim_timesteps
+ ddim_timesteps = np.asarray(list(range(0, num_ddpm_timesteps, c)))
+ elif ddim_discr_method == 'quad':
+ ddim_timesteps = ((np.linspace(0, np.sqrt(num_ddpm_timesteps * .8), num_ddim_timesteps)) ** 2).astype(int)
+ else:
+ raise NotImplementedError(f'There is no ddim discretization method called "{ddim_discr_method}"')
+
+ # assert ddim_timesteps.shape[0] == num_ddim_timesteps
+ # add one to get the final alpha values right (the ones from first scale to data during sampling)
+ steps_out = ddim_timesteps + 1
+ if verbose:
+ print(f'Selected timesteps for ddim sampler: {steps_out}')
+ return steps_out
+
+
+def make_ddim_sampling_parameters(alphacums, ddim_timesteps, eta, verbose=True):
+ # select alphas for computing the variance schedule
+ alphas = alphacums[ddim_timesteps]
+ alphas_prev = np.asarray([alphacums[0]] + alphacums[ddim_timesteps[:-1]].tolist())
+
+ # according the the formula provided in https://arxiv.org/abs/2010.02502
+ sigmas = eta * np.sqrt((1 - alphas_prev) / (1 - alphas) * (1 - alphas / alphas_prev))
+ if verbose:
+ print(f'Selected alphas for ddim sampler: a_t: {alphas}; a_(t-1): {alphas_prev}')
+ print(f'For the chosen value of eta, which is {eta}, '
+ f'this results in the following sigma_t schedule for ddim sampler {sigmas}')
+ return sigmas, alphas, alphas_prev
+
+
+def betas_for_alpha_bar(num_diffusion_timesteps, alpha_bar, max_beta=0.999):
+ """
+ Create a beta schedule that discretizes the given alpha_t_bar function,
+ which defines the cumulative product of (1-beta) over time from t = [0,1].
+ :param num_diffusion_timesteps: the number of betas to produce.
+ :param alpha_bar: a lambda that takes an argument t from 0 to 1 and
+ produces the cumulative product of (1-beta) up to that
+ part of the diffusion process.
+ :param max_beta: the maximum beta to use; use values lower than 1 to
+ prevent singularities.
+ """
+ betas = []
+ for i in range(num_diffusion_timesteps):
+ t1 = i / num_diffusion_timesteps
+ t2 = (i + 1) / num_diffusion_timesteps
+ betas.append(min(1 - alpha_bar(t2) / alpha_bar(t1), max_beta))
+ return np.array(betas)
+
+
+def extract_into_tensor(a, t, x_shape):
+ b, *_ = t.shape
+ out = a.gather(-1, t)
+ return out.reshape(b, *((1,) * (len(x_shape) - 1)))
+
+
+def checkpoint(func, inputs, params, flag):
+ """
+ Evaluate a function without caching intermediate activations, allowing for
+ reduced memory at the expense of extra compute in the backward pass.
+ :param func: the function to evaluate.
+ :param inputs: the argument sequence to pass to `func`.
+ :param params: a sequence of parameters `func` depends on but does not
+ explicitly take as arguments.
+ :param flag: if False, disable gradient checkpointing.
+ """
+ if flag:
+ args = tuple(inputs) + tuple(params)
+ return CheckpointFunction.apply(func, len(inputs), *args)
+ else:
+ return func(*inputs)
+
+
+class CheckpointFunction(torch.autograd.Function):
+ @staticmethod
+ def forward(ctx, run_function, length, *args):
+ ctx.run_function = run_function
+ ctx.input_tensors = list(args[:length])
+ ctx.input_params = list(args[length:])
+
+ with torch.no_grad():
+ output_tensors = ctx.run_function(*ctx.input_tensors)
+ return output_tensors
+
+ @staticmethod
+ def backward(ctx, *output_grads):
+ ctx.input_tensors = [x.detach().requires_grad_(True) for x in ctx.input_tensors]
+ with torch.enable_grad():
+ # Fixes a bug where the first op in run_function modifies the
+ # Tensor storage in place, which is not allowed for detach()'d
+ # Tensors.
+ shallow_copies = [x.view_as(x) for x in ctx.input_tensors]
+ output_tensors = ctx.run_function(*shallow_copies)
+ input_grads = torch.autograd.grad(
+ output_tensors,
+ ctx.input_tensors + ctx.input_params,
+ output_grads,
+ allow_unused=True,
+ )
+ del ctx.input_tensors
+ del ctx.input_params
+ del output_tensors
+ return (None, None) + input_grads
+
+
+def timestep_embedding(timesteps, dim, max_period=10000, repeat_only=False):
+ """
+ Create sinusoidal timestep embeddings.
+ :param timesteps: a 1-D Tensor of N indices, one per batch element.
+ These may be fractional.
+ :param dim: the dimension of the output.
+ :param max_period: controls the minimum frequency of the embeddings.
+ :return: an [N x dim] Tensor of positional embeddings.
+ """
+ if not repeat_only:
+ half = dim // 2
+ freqs = torch.exp(
+ -math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half
+ ).to(device=timesteps.device)
+ args = timesteps[:, None].float() * freqs[None]
+ embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
+ if dim % 2:
+ embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
+ else:
+ embedding = repeat(timesteps, 'b -> b d', d=dim)
+ return embedding
+
+
+def zero_module(module):
+ """
+ Zero out the parameters of a module and return it.
+ """
+ for p in module.parameters():
+ p.detach().zero_()
+ return module
+
+
+def scale_module(module, scale):
+ """
+ Scale the parameters of a module and return it.
+ """
+ for p in module.parameters():
+ p.detach().mul_(scale)
+ return module
+
+
+def mean_flat(tensor):
+ """
+ Take the mean over all non-batch dimensions.
+ """
+ return tensor.mean(dim=list(range(1, len(tensor.shape))))
+
+
+def normalization(channels):
+ """
+ Make a standard normalization layer.
+ :param channels: number of input channels.
+ :return: an nn.Module for normalization.
+ """
+ return GroupNorm32(32, channels)
+
+
+# PyTorch 1.7 has SiLU, but we support PyTorch 1.5.
+class SiLU(nn.Module):
+ def forward(self, x):
+ return x * torch.sigmoid(x)
+
+
+class GroupNorm32(nn.GroupNorm):
+ def forward(self, x):
+ return super().forward(x.float()).type(x.dtype)
+
+def conv_nd(dims, *args, **kwargs):
+ """
+ Create a 1D, 2D, or 3D convolution module.
+ """
+ if dims == 1:
+ return nn.Conv1d(*args, **kwargs)
+ elif dims == 2:
+ return nn.Conv2d(*args, **kwargs)
+ elif dims == 3:
+ return nn.Conv3d(*args, **kwargs)
+ raise ValueError(f"unsupported dimensions: {dims}")
+
+
+def linear(*args, **kwargs):
+ """
+ Create a linear module.
+ """
+ return nn.Linear(*args, **kwargs)
+
+
+def avg_pool_nd(dims, *args, **kwargs):
+ """
+ Create a 1D, 2D, or 3D average pooling module.
+ """
+ if dims == 1:
+ return nn.AvgPool1d(*args, **kwargs)
+ elif dims == 2:
+ return nn.AvgPool2d(*args, **kwargs)
+ elif dims == 3:
+ return nn.AvgPool3d(*args, **kwargs)
+ raise ValueError(f"unsupported dimensions: {dims}")
+
+
+class HybridConditioner(nn.Module):
+
+ def __init__(self, c_concat_config, c_crossattn_config):
+ super().__init__()
+ self.concat_conditioner = instantiate_from_config(c_concat_config)
+ self.crossattn_conditioner = instantiate_from_config(c_crossattn_config)
+
+ def forward(self, c_concat, c_crossattn):
+ c_concat = self.concat_conditioner(c_concat)
+ c_crossattn = self.crossattn_conditioner(c_crossattn)
+ return {'c_concat': [c_concat], 'c_crossattn': [c_crossattn]}
+
+
+def noise_like(shape, device, repeat=False):
+ repeat_noise = lambda: torch.randn((1, *shape[1:]), device=device).repeat(shape[0], *((1,) * (len(shape) - 1)))
+ noise = lambda: torch.randn(shape, device=device)
+ return repeat_noise() if repeat else noise()
\ No newline at end of file
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/distributions/__init__.py b/OCR_earsing/latent_diffusion/ldm/modules/distributions/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/distributions/distributions.py b/OCR_earsing/latent_diffusion/ldm/modules/distributions/distributions.py
new file mode 100644
index 0000000..f2b8ef9
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/distributions/distributions.py
@@ -0,0 +1,92 @@
+import torch
+import numpy as np
+
+
+class AbstractDistribution:
+ def sample(self):
+ raise NotImplementedError()
+
+ def mode(self):
+ raise NotImplementedError()
+
+
+class DiracDistribution(AbstractDistribution):
+ def __init__(self, value):
+ self.value = value
+
+ def sample(self):
+ return self.value
+
+ def mode(self):
+ return self.value
+
+
+class DiagonalGaussianDistribution(object):
+ def __init__(self, parameters, deterministic=False):
+ self.parameters = parameters
+ self.mean, self.logvar = torch.chunk(parameters, 2, dim=1)
+ self.logvar = torch.clamp(self.logvar, -30.0, 20.0)
+ self.deterministic = deterministic
+ self.std = torch.exp(0.5 * self.logvar)
+ self.var = torch.exp(self.logvar)
+ if self.deterministic:
+ self.var = self.std = torch.zeros_like(self.mean).to(device=self.parameters.device)
+
+ def sample(self):
+ x = self.mean + self.std * torch.randn(self.mean.shape).to(device=self.parameters.device)
+ return x
+
+ def kl(self, other=None):
+ if self.deterministic:
+ return torch.Tensor([0.])
+ else:
+ if other is None:
+ return 0.5 * torch.sum(torch.pow(self.mean, 2)
+ + self.var - 1.0 - self.logvar,
+ dim=[1, 2, 3])
+ else:
+ return 0.5 * torch.sum(
+ torch.pow(self.mean - other.mean, 2) / other.var
+ + self.var / other.var - 1.0 - self.logvar + other.logvar,
+ dim=[1, 2, 3])
+
+ def nll(self, sample, dims=[1,2,3]):
+ if self.deterministic:
+ return torch.Tensor([0.])
+ logtwopi = np.log(2.0 * np.pi)
+ return 0.5 * torch.sum(
+ logtwopi + self.logvar + torch.pow(sample - self.mean, 2) / self.var,
+ dim=dims)
+
+ def mode(self):
+ return self.mean
+
+
+def normal_kl(mean1, logvar1, mean2, logvar2):
+ """
+ source: https://github.com/openai/guided-diffusion/blob/27c20a8fab9cb472df5d6bdd6c8d11c8f430b924/guided_diffusion/losses.py#L12
+ Compute the KL divergence between two gaussians.
+ Shapes are automatically broadcasted, so batches can be compared to
+ scalars, among other use cases.
+ """
+ tensor = None
+ for obj in (mean1, logvar1, mean2, logvar2):
+ if isinstance(obj, torch.Tensor):
+ tensor = obj
+ break
+ assert tensor is not None, "at least one argument must be a Tensor"
+
+ # Force variances to be Tensors. Broadcasting helps convert scalars to
+ # Tensors, but it does not work for torch.exp().
+ logvar1, logvar2 = [
+ x if isinstance(x, torch.Tensor) else torch.tensor(x).to(tensor)
+ for x in (logvar1, logvar2)
+ ]
+
+ return 0.5 * (
+ -1.0
+ + logvar2
+ - logvar1
+ + torch.exp(logvar1 - logvar2)
+ + ((mean1 - mean2) ** 2) * torch.exp(-logvar2)
+ )
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/ema.py b/OCR_earsing/latent_diffusion/ldm/modules/ema.py
new file mode 100644
index 0000000..c8c75af
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/ema.py
@@ -0,0 +1,76 @@
+import torch
+from torch import nn
+
+
+class LitEma(nn.Module):
+ def __init__(self, model, decay=0.9999, use_num_upates=True):
+ super().__init__()
+ if decay < 0.0 or decay > 1.0:
+ raise ValueError('Decay must be between 0 and 1')
+
+ self.m_name2s_name = {}
+ self.register_buffer('decay', torch.tensor(decay, dtype=torch.float32))
+ self.register_buffer('num_updates', torch.tensor(0,dtype=torch.int) if use_num_upates
+ else torch.tensor(-1,dtype=torch.int))
+
+ for name, p in model.named_parameters():
+ if p.requires_grad:
+ #remove as '.'-character is not allowed in buffers
+ s_name = name.replace('.','')
+ self.m_name2s_name.update({name:s_name})
+ self.register_buffer(s_name,p.clone().detach().data)
+
+ self.collected_params = []
+
+ def forward(self,model):
+ decay = self.decay
+
+ if self.num_updates >= 0:
+ self.num_updates += 1
+ decay = min(self.decay,(1 + self.num_updates) / (10 + self.num_updates))
+
+ one_minus_decay = 1.0 - decay
+
+ with torch.no_grad():
+ m_param = dict(model.named_parameters())
+ shadow_params = dict(self.named_buffers())
+
+ for key in m_param:
+ if m_param[key].requires_grad:
+ sname = self.m_name2s_name[key]
+ shadow_params[sname] = shadow_params[sname].type_as(m_param[key])
+ shadow_params[sname].sub_(one_minus_decay * (shadow_params[sname] - m_param[key]))
+ else:
+ assert not key in self.m_name2s_name
+
+ def copy_to(self, model):
+ m_param = dict(model.named_parameters())
+ shadow_params = dict(self.named_buffers())
+ for key in m_param:
+ if m_param[key].requires_grad:
+ m_param[key].data.copy_(shadow_params[self.m_name2s_name[key]].data)
+ else:
+ assert not key in self.m_name2s_name
+
+ def store(self, parameters):
+ """
+ Save the current parameters for restoring later.
+ Args:
+ parameters: Iterable of `torch.nn.Parameter`; the parameters to be
+ temporarily stored.
+ """
+ self.collected_params = [param.clone() for param in parameters]
+
+ def restore(self, parameters):
+ """
+ Restore the parameters stored with the `store` method.
+ Useful to validate the model with EMA parameters without affecting the
+ original optimization process. Store the parameters before the
+ `copy_to` method. After validation (or model saving), use this to
+ restore the former parameters.
+ Args:
+ parameters: Iterable of `torch.nn.Parameter`; the parameters to be
+ updated with the stored parameters.
+ """
+ for c_param, param in zip(self.collected_params, parameters):
+ param.data.copy_(c_param.data)
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/encoders/__init__.py b/OCR_earsing/latent_diffusion/ldm/modules/encoders/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/encoders/modules.py b/OCR_earsing/latent_diffusion/ldm/modules/encoders/modules.py
new file mode 100644
index 0000000..aa3031d
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/encoders/modules.py
@@ -0,0 +1,202 @@
+import torch
+import torch.nn as nn
+from functools import partial
+import clip
+from einops import rearrange, repeat
+import kornia
+
+
+from ldm.modules.x_transformer import Encoder, TransformerWrapper # TODO: can we directly rely on lucidrains code and simply add this as a reuirement? --> test
+
+
+class AbstractEncoder(nn.Module):
+ def __init__(self):
+ super().__init__()
+
+ def encode(self, *args, **kwargs):
+ raise NotImplementedError
+
+
+
+class ClassEmbedder(nn.Module):
+ def __init__(self, embed_dim, n_classes=1000, key='class'):
+ super().__init__()
+ self.key = key
+ self.embedding = nn.Embedding(n_classes, embed_dim)
+
+ def forward(self, batch, key=None):
+ if key is None:
+ key = self.key
+ # this is for use in crossattn
+ c = batch[key][:, None]
+ c = self.embedding(c)
+ return c
+
+
+class TransformerEmbedder(AbstractEncoder):
+ """Some transformer encoder layers"""
+ def __init__(self, n_embed, n_layer, vocab_size, max_seq_len=77, device="cuda"):
+ super().__init__()
+ self.device = device
+ self.transformer = TransformerWrapper(num_tokens=vocab_size, max_seq_len=max_seq_len,
+ attn_layers=Encoder(dim=n_embed, depth=n_layer))
+
+ def forward(self, tokens):
+ tokens = tokens.to(self.device) # meh
+ z = self.transformer(tokens, return_embeddings=True)
+ return z
+
+ def encode(self, x):
+ return self(x)
+
+
+class BERTTokenizer(AbstractEncoder):
+ """ Uses a pretrained BERT tokenizer by huggingface. Vocab size: 30522 (?)"""
+ def __init__(self, device="cuda", vq_interface=True, max_length=77):
+ super().__init__()
+ from transformers import BertTokenizerFast # TODO: add to reuquirements
+ self.tokenizer = BertTokenizerFast.from_pretrained("bert-base-uncased")
+ self.device = device
+ self.vq_interface = vq_interface
+ self.max_length = max_length
+
+ def forward(self, text):
+ batch_encoding = self.tokenizer(text, truncation=True, max_length=self.max_length, return_length=True,
+ return_overflowing_tokens=False, padding="max_length", return_tensors="pt")
+ tokens = batch_encoding["input_ids"].to(self.device)
+ return tokens
+
+ @torch.no_grad()
+ def encode(self, text):
+ tokens = self(text)
+ if not self.vq_interface:
+ return tokens
+ return None, None, [None, None, tokens]
+
+ def decode(self, text):
+ return text
+
+
+class BERTEmbedder(AbstractEncoder):
+ """Uses the BERT tokenizr model and add some transformer encoder layers"""
+ def __init__(self, n_embed, n_layer, vocab_size=30522, max_seq_len=77,
+ device="cuda",use_tokenizer=True, embedding_dropout=0.0):
+ super().__init__()
+ self.use_tknz_fn = use_tokenizer
+ if self.use_tknz_fn:
+ self.tknz_fn = BERTTokenizer(vq_interface=False, max_length=max_seq_len)
+ self.device = device
+ self.transformer = TransformerWrapper(num_tokens=vocab_size, max_seq_len=max_seq_len,
+ attn_layers=Encoder(dim=n_embed, depth=n_layer),
+ emb_dropout=embedding_dropout)
+
+ def forward(self, text):
+ if self.use_tknz_fn:
+ tokens = self.tknz_fn(text)#.to(self.device)
+ else:
+ tokens = text
+ z = self.transformer(tokens, return_embeddings=True)
+ return z
+
+ def encode(self, text):
+ # output of length 77
+ return self(text)
+
+
+class SpatialRescaler(nn.Module):
+ def __init__(self,
+ n_stages=1,
+ method='bilinear',
+ multiplier=0.5,
+ in_channels=3,
+ out_channels=None,
+ bias=False):
+ super().__init__()
+ self.n_stages = n_stages
+ assert self.n_stages >= 0
+ assert method in ['nearest','linear','bilinear','trilinear','bicubic','area']
+ self.multiplier = multiplier
+ self.interpolator = partial(torch.nn.functional.interpolate, mode=method)
+ self.remap_output = out_channels is not None
+ if self.remap_output:
+ print(f'Spatial Rescaler mapping from {in_channels} to {out_channels} channels after resizing.')
+ self.channel_mapper = nn.Conv2d(in_channels,out_channels,1,bias=bias)
+
+ def forward(self,x):
+ for stage in range(self.n_stages):
+ x = self.interpolator(x, scale_factor=self.multiplier)
+
+
+ if self.remap_output:
+ x = self.channel_mapper(x)
+ return x
+
+ def encode(self, x):
+ return self(x)
+
+
+class FrozenCLIPTextEmbedder(nn.Module):
+ """
+ Uses the CLIP transformer encoder for text.
+ """
+ def __init__(self, version='ViT-L/14', device="cuda", max_length=77, n_repeat=1, normalize=True):
+ super().__init__()
+ self.model, _ = clip.load(version, jit=False, device="cpu")
+ self.device = device
+ self.max_length = max_length
+ self.n_repeat = n_repeat
+ self.normalize = normalize
+
+ def freeze(self):
+ self.model = self.model.eval()
+ for param in self.parameters():
+ param.requires_grad = False
+
+ def forward(self, text):
+ tokens = clip.tokenize(text).to(self.device)
+ z = self.model.encode_text(tokens)
+ if self.normalize:
+ z = z / torch.linalg.norm(z, dim=1, keepdim=True)
+ return z
+
+ def encode(self, text):
+ z = self(text)
+ if z.ndim==2:
+ z = z[:, None, :]
+ z = repeat(z, 'b 1 d -> b k d', k=self.n_repeat)
+ return z
+
+
+class FrozenClipImageEmbedder(nn.Module):
+ """
+ Uses the CLIP image encoder.
+ """
+ def __init__(
+ self,
+ model,
+ jit=False,
+ device='cuda' if torch.cuda.is_available() else 'cpu',
+ antialias=False,
+ ):
+ super().__init__()
+ self.model, _ = clip.load(name=model, device=device, jit=jit)
+
+ self.antialias = antialias
+
+ self.register_buffer('mean', torch.Tensor([0.48145466, 0.4578275, 0.40821073]), persistent=False)
+ self.register_buffer('std', torch.Tensor([0.26862954, 0.26130258, 0.27577711]), persistent=False)
+
+ def preprocess(self, x):
+ # normalize to [0,1]
+ x = kornia.geometry.resize(x, (224, 224),
+ interpolation='bicubic',align_corners=True,
+ antialias=self.antialias)
+ x = (x + 1.) / 2.
+ # renormalize according to clip
+ x = kornia.enhance.normalize(x, self.mean, self.std)
+ return x
+
+ def forward(self, x):
+ # x is assumed to be in range [-1,1]
+ return self.model.encode_image(self.preprocess(x))
+
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/__init__.py b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/__init__.py
new file mode 100644
index 0000000..7836cad
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/__init__.py
@@ -0,0 +1,2 @@
+from ldm.modules.image_degradation.bsrgan import degradation_bsrgan_variant as degradation_fn_bsr
+from ldm.modules.image_degradation.bsrgan_light import degradation_bsrgan_variant as degradation_fn_bsr_light
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/bsrgan.py b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/bsrgan.py
new file mode 100644
index 0000000..32ef561
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/bsrgan.py
@@ -0,0 +1,730 @@
+# -*- coding: utf-8 -*-
+"""
+# --------------------------------------------
+# Super-Resolution
+# --------------------------------------------
+#
+# Kai Zhang (cskaizhang@gmail.com)
+# https://github.com/cszn
+# From 2019/03--2021/08
+# --------------------------------------------
+"""
+
+import numpy as np
+import cv2
+import torch
+
+from functools import partial
+import random
+from scipy import ndimage
+import scipy
+import scipy.stats as ss
+from scipy.interpolate import interp2d
+from scipy.linalg import orth
+import albumentations
+
+import ldm.modules.image_degradation.utils_image as util
+
+
+def modcrop_np(img, sf):
+ '''
+ Args:
+ img: numpy image, WxH or WxHxC
+ sf: scale factor
+ Return:
+ cropped image
+ '''
+ w, h = img.shape[:2]
+ im = np.copy(img)
+ return im[:w - w % sf, :h - h % sf, ...]
+
+
+"""
+# --------------------------------------------
+# anisotropic Gaussian kernels
+# --------------------------------------------
+"""
+
+
+def analytic_kernel(k):
+ """Calculate the X4 kernel from the X2 kernel (for proof see appendix in paper)"""
+ k_size = k.shape[0]
+ # Calculate the big kernels size
+ big_k = np.zeros((3 * k_size - 2, 3 * k_size - 2))
+ # Loop over the small kernel to fill the big one
+ for r in range(k_size):
+ for c in range(k_size):
+ big_k[2 * r:2 * r + k_size, 2 * c:2 * c + k_size] += k[r, c] * k
+ # Crop the edges of the big kernel to ignore very small values and increase run time of SR
+ crop = k_size // 2
+ cropped_big_k = big_k[crop:-crop, crop:-crop]
+ # Normalize to 1
+ return cropped_big_k / cropped_big_k.sum()
+
+
+def anisotropic_Gaussian(ksize=15, theta=np.pi, l1=6, l2=6):
+ """ generate an anisotropic Gaussian kernel
+ Args:
+ ksize : e.g., 15, kernel size
+ theta : [0, pi], rotation angle range
+ l1 : [0.1,50], scaling of eigenvalues
+ l2 : [0.1,l1], scaling of eigenvalues
+ If l1 = l2, will get an isotropic Gaussian kernel.
+ Returns:
+ k : kernel
+ """
+
+ v = np.dot(np.array([[np.cos(theta), -np.sin(theta)], [np.sin(theta), np.cos(theta)]]), np.array([1., 0.]))
+ V = np.array([[v[0], v[1]], [v[1], -v[0]]])
+ D = np.array([[l1, 0], [0, l2]])
+ Sigma = np.dot(np.dot(V, D), np.linalg.inv(V))
+ k = gm_blur_kernel(mean=[0, 0], cov=Sigma, size=ksize)
+
+ return k
+
+
+def gm_blur_kernel(mean, cov, size=15):
+ center = size / 2.0 + 0.5
+ k = np.zeros([size, size])
+ for y in range(size):
+ for x in range(size):
+ cy = y - center + 1
+ cx = x - center + 1
+ k[y, x] = ss.multivariate_normal.pdf([cx, cy], mean=mean, cov=cov)
+
+ k = k / np.sum(k)
+ return k
+
+
+def shift_pixel(x, sf, upper_left=True):
+ """shift pixel for super-resolution with different scale factors
+ Args:
+ x: WxHxC or WxH
+ sf: scale factor
+ upper_left: shift direction
+ """
+ h, w = x.shape[:2]
+ shift = (sf - 1) * 0.5
+ xv, yv = np.arange(0, w, 1.0), np.arange(0, h, 1.0)
+ if upper_left:
+ x1 = xv + shift
+ y1 = yv + shift
+ else:
+ x1 = xv - shift
+ y1 = yv - shift
+
+ x1 = np.clip(x1, 0, w - 1)
+ y1 = np.clip(y1, 0, h - 1)
+
+ if x.ndim == 2:
+ x = interp2d(xv, yv, x)(x1, y1)
+ if x.ndim == 3:
+ for i in range(x.shape[-1]):
+ x[:, :, i] = interp2d(xv, yv, x[:, :, i])(x1, y1)
+
+ return x
+
+
+def blur(x, k):
+ '''
+ x: image, NxcxHxW
+ k: kernel, Nx1xhxw
+ '''
+ n, c = x.shape[:2]
+ p1, p2 = (k.shape[-2] - 1) // 2, (k.shape[-1] - 1) // 2
+ x = torch.nn.functional.pad(x, pad=(p1, p2, p1, p2), mode='replicate')
+ k = k.repeat(1, c, 1, 1)
+ k = k.view(-1, 1, k.shape[2], k.shape[3])
+ x = x.view(1, -1, x.shape[2], x.shape[3])
+ x = torch.nn.functional.conv2d(x, k, bias=None, stride=1, padding=0, groups=n * c)
+ x = x.view(n, c, x.shape[2], x.shape[3])
+
+ return x
+
+
+def gen_kernel(k_size=np.array([15, 15]), scale_factor=np.array([4, 4]), min_var=0.6, max_var=10., noise_level=0):
+ """"
+ # modified version of https://github.com/assafshocher/BlindSR_dataset_generator
+ # Kai Zhang
+ # min_var = 0.175 * sf # variance of the gaussian kernel will be sampled between min_var and max_var
+ # max_var = 2.5 * sf
+ """
+ # Set random eigen-vals (lambdas) and angle (theta) for COV matrix
+ lambda_1 = min_var + np.random.rand() * (max_var - min_var)
+ lambda_2 = min_var + np.random.rand() * (max_var - min_var)
+ theta = np.random.rand() * np.pi # random theta
+ noise = -noise_level + np.random.rand(*k_size) * noise_level * 2
+
+ # Set COV matrix using Lambdas and Theta
+ LAMBDA = np.diag([lambda_1, lambda_2])
+ Q = np.array([[np.cos(theta), -np.sin(theta)],
+ [np.sin(theta), np.cos(theta)]])
+ SIGMA = Q @ LAMBDA @ Q.T
+ INV_SIGMA = np.linalg.inv(SIGMA)[None, None, :, :]
+
+ # Set expectation position (shifting kernel for aligned image)
+ MU = k_size // 2 - 0.5 * (scale_factor - 1) # - 0.5 * (scale_factor - k_size % 2)
+ MU = MU[None, None, :, None]
+
+ # Create meshgrid for Gaussian
+ [X, Y] = np.meshgrid(range(k_size[0]), range(k_size[1]))
+ Z = np.stack([X, Y], 2)[:, :, :, None]
+
+ # Calcualte Gaussian for every pixel of the kernel
+ ZZ = Z - MU
+ ZZ_t = ZZ.transpose(0, 1, 3, 2)
+ raw_kernel = np.exp(-0.5 * np.squeeze(ZZ_t @ INV_SIGMA @ ZZ)) * (1 + noise)
+
+ # shift the kernel so it will be centered
+ # raw_kernel_centered = kernel_shift(raw_kernel, scale_factor)
+
+ # Normalize the kernel and return
+ # kernel = raw_kernel_centered / np.sum(raw_kernel_centered)
+ kernel = raw_kernel / np.sum(raw_kernel)
+ return kernel
+
+
+def fspecial_gaussian(hsize, sigma):
+ hsize = [hsize, hsize]
+ siz = [(hsize[0] - 1.0) / 2.0, (hsize[1] - 1.0) / 2.0]
+ std = sigma
+ [x, y] = np.meshgrid(np.arange(-siz[1], siz[1] + 1), np.arange(-siz[0], siz[0] + 1))
+ arg = -(x * x + y * y) / (2 * std * std)
+ h = np.exp(arg)
+ h[h < scipy.finfo(float).eps * h.max()] = 0
+ sumh = h.sum()
+ if sumh != 0:
+ h = h / sumh
+ return h
+
+
+def fspecial_laplacian(alpha):
+ alpha = max([0, min([alpha, 1])])
+ h1 = alpha / (alpha + 1)
+ h2 = (1 - alpha) / (alpha + 1)
+ h = [[h1, h2, h1], [h2, -4 / (alpha + 1), h2], [h1, h2, h1]]
+ h = np.array(h)
+ return h
+
+
+def fspecial(filter_type, *args, **kwargs):
+ '''
+ python code from:
+ https://github.com/ronaldosena/imagens-medicas-2/blob/40171a6c259edec7827a6693a93955de2bd39e76/Aulas/aula_2_-_uniform_filter/matlab_fspecial.py
+ '''
+ if filter_type == 'gaussian':
+ return fspecial_gaussian(*args, **kwargs)
+ if filter_type == 'laplacian':
+ return fspecial_laplacian(*args, **kwargs)
+
+
+"""
+# --------------------------------------------
+# degradation models
+# --------------------------------------------
+"""
+
+
+def bicubic_degradation(x, sf=3):
+ '''
+ Args:
+ x: HxWxC image, [0, 1]
+ sf: down-scale factor
+ Return:
+ bicubicly downsampled LR image
+ '''
+ x = util.imresize_np(x, scale=1 / sf)
+ return x
+
+
+def srmd_degradation(x, k, sf=3):
+ ''' blur + bicubic downsampling
+ Args:
+ x: HxWxC image, [0, 1]
+ k: hxw, double
+ sf: down-scale factor
+ Return:
+ downsampled LR image
+ Reference:
+ @inproceedings{zhang2018learning,
+ title={Learning a single convolutional super-resolution network for multiple degradations},
+ author={Zhang, Kai and Zuo, Wangmeng and Zhang, Lei},
+ booktitle={IEEE Conference on Computer Vision and Pattern Recognition},
+ pages={3262--3271},
+ year={2018}
+ }
+ '''
+ x = ndimage.filters.convolve(x, np.expand_dims(k, axis=2), mode='wrap') # 'nearest' | 'mirror'
+ x = bicubic_degradation(x, sf=sf)
+ return x
+
+
+def dpsr_degradation(x, k, sf=3):
+ ''' bicubic downsampling + blur
+ Args:
+ x: HxWxC image, [0, 1]
+ k: hxw, double
+ sf: down-scale factor
+ Return:
+ downsampled LR image
+ Reference:
+ @inproceedings{zhang2019deep,
+ title={Deep Plug-and-Play Super-Resolution for Arbitrary Blur Kernels},
+ author={Zhang, Kai and Zuo, Wangmeng and Zhang, Lei},
+ booktitle={IEEE Conference on Computer Vision and Pattern Recognition},
+ pages={1671--1681},
+ year={2019}
+ }
+ '''
+ x = bicubic_degradation(x, sf=sf)
+ x = ndimage.filters.convolve(x, np.expand_dims(k, axis=2), mode='wrap')
+ return x
+
+
+def classical_degradation(x, k, sf=3):
+ ''' blur + downsampling
+ Args:
+ x: HxWxC image, [0, 1]/[0, 255]
+ k: hxw, double
+ sf: down-scale factor
+ Return:
+ downsampled LR image
+ '''
+ x = ndimage.filters.convolve(x, np.expand_dims(k, axis=2), mode='wrap')
+ # x = filters.correlate(x, np.expand_dims(np.flip(k), axis=2))
+ st = 0
+ return x[st::sf, st::sf, ...]
+
+
+def add_sharpening(img, weight=0.5, radius=50, threshold=10):
+ """USM sharpening. borrowed from real-ESRGAN
+ Input image: I; Blurry image: B.
+ 1. K = I + weight * (I - B)
+ 2. Mask = 1 if abs(I - B) > threshold, else: 0
+ 3. Blur mask:
+ 4. Out = Mask * K + (1 - Mask) * I
+ Args:
+ img (Numpy array): Input image, HWC, BGR; float32, [0, 1].
+ weight (float): Sharp weight. Default: 1.
+ radius (float): Kernel size of Gaussian blur. Default: 50.
+ threshold (int):
+ """
+ if radius % 2 == 0:
+ radius += 1
+ blur = cv2.GaussianBlur(img, (radius, radius), 0)
+ residual = img - blur
+ mask = np.abs(residual) * 255 > threshold
+ mask = mask.astype('float32')
+ soft_mask = cv2.GaussianBlur(mask, (radius, radius), 0)
+
+ K = img + weight * residual
+ K = np.clip(K, 0, 1)
+ return soft_mask * K + (1 - soft_mask) * img
+
+
+def add_blur(img, sf=4):
+ wd2 = 4.0 + sf
+ wd = 2.0 + 0.2 * sf
+ if random.random() < 0.5:
+ l1 = wd2 * random.random()
+ l2 = wd2 * random.random()
+ k = anisotropic_Gaussian(ksize=2 * random.randint(2, 11) + 3, theta=random.random() * np.pi, l1=l1, l2=l2)
+ else:
+ k = fspecial('gaussian', 2 * random.randint(2, 11) + 3, wd * random.random())
+ img = ndimage.filters.convolve(img, np.expand_dims(k, axis=2), mode='mirror')
+
+ return img
+
+
+def add_resize(img, sf=4):
+ rnum = np.random.rand()
+ if rnum > 0.8: # up
+ sf1 = random.uniform(1, 2)
+ elif rnum < 0.7: # down
+ sf1 = random.uniform(0.5 / sf, 1)
+ else:
+ sf1 = 1.0
+ img = cv2.resize(img, (int(sf1 * img.shape[1]), int(sf1 * img.shape[0])), interpolation=random.choice([1, 2, 3]))
+ img = np.clip(img, 0.0, 1.0)
+
+ return img
+
+
+# def add_Gaussian_noise(img, noise_level1=2, noise_level2=25):
+# noise_level = random.randint(noise_level1, noise_level2)
+# rnum = np.random.rand()
+# if rnum > 0.6: # add color Gaussian noise
+# img += np.random.normal(0, noise_level / 255.0, img.shape).astype(np.float32)
+# elif rnum < 0.4: # add grayscale Gaussian noise
+# img += np.random.normal(0, noise_level / 255.0, (*img.shape[:2], 1)).astype(np.float32)
+# else: # add noise
+# L = noise_level2 / 255.
+# D = np.diag(np.random.rand(3))
+# U = orth(np.random.rand(3, 3))
+# conv = np.dot(np.dot(np.transpose(U), D), U)
+# img += np.random.multivariate_normal([0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2]).astype(np.float32)
+# img = np.clip(img, 0.0, 1.0)
+# return img
+
+def add_Gaussian_noise(img, noise_level1=2, noise_level2=25):
+ noise_level = random.randint(noise_level1, noise_level2)
+ rnum = np.random.rand()
+ if rnum > 0.6: # add color Gaussian noise
+ img = img + np.random.normal(0, noise_level / 255.0, img.shape).astype(np.float32)
+ elif rnum < 0.4: # add grayscale Gaussian noise
+ img = img + np.random.normal(0, noise_level / 255.0, (*img.shape[:2], 1)).astype(np.float32)
+ else: # add noise
+ L = noise_level2 / 255.
+ D = np.diag(np.random.rand(3))
+ U = orth(np.random.rand(3, 3))
+ conv = np.dot(np.dot(np.transpose(U), D), U)
+ img = img + np.random.multivariate_normal([0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2]).astype(np.float32)
+ img = np.clip(img, 0.0, 1.0)
+ return img
+
+
+def add_speckle_noise(img, noise_level1=2, noise_level2=25):
+ noise_level = random.randint(noise_level1, noise_level2)
+ img = np.clip(img, 0.0, 1.0)
+ rnum = random.random()
+ if rnum > 0.6:
+ img += img * np.random.normal(0, noise_level / 255.0, img.shape).astype(np.float32)
+ elif rnum < 0.4:
+ img += img * np.random.normal(0, noise_level / 255.0, (*img.shape[:2], 1)).astype(np.float32)
+ else:
+ L = noise_level2 / 255.
+ D = np.diag(np.random.rand(3))
+ U = orth(np.random.rand(3, 3))
+ conv = np.dot(np.dot(np.transpose(U), D), U)
+ img += img * np.random.multivariate_normal([0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2]).astype(np.float32)
+ img = np.clip(img, 0.0, 1.0)
+ return img
+
+
+def add_Poisson_noise(img):
+ img = np.clip((img * 255.0).round(), 0, 255) / 255.
+ vals = 10 ** (2 * random.random() + 2.0) # [2, 4]
+ if random.random() < 0.5:
+ img = np.random.poisson(img * vals).astype(np.float32) / vals
+ else:
+ img_gray = np.dot(img[..., :3], [0.299, 0.587, 0.114])
+ img_gray = np.clip((img_gray * 255.0).round(), 0, 255) / 255.
+ noise_gray = np.random.poisson(img_gray * vals).astype(np.float32) / vals - img_gray
+ img += noise_gray[:, :, np.newaxis]
+ img = np.clip(img, 0.0, 1.0)
+ return img
+
+
+def add_JPEG_noise(img):
+ quality_factor = random.randint(30, 95)
+ img = cv2.cvtColor(util.single2uint(img), cv2.COLOR_RGB2BGR)
+ result, encimg = cv2.imencode('.jpg', img, [int(cv2.IMWRITE_JPEG_QUALITY), quality_factor])
+ img = cv2.imdecode(encimg, 1)
+ img = cv2.cvtColor(util.uint2single(img), cv2.COLOR_BGR2RGB)
+ return img
+
+
+def random_crop(lq, hq, sf=4, lq_patchsize=64):
+ h, w = lq.shape[:2]
+ rnd_h = random.randint(0, h - lq_patchsize)
+ rnd_w = random.randint(0, w - lq_patchsize)
+ lq = lq[rnd_h:rnd_h + lq_patchsize, rnd_w:rnd_w + lq_patchsize, :]
+
+ rnd_h_H, rnd_w_H = int(rnd_h * sf), int(rnd_w * sf)
+ hq = hq[rnd_h_H:rnd_h_H + lq_patchsize * sf, rnd_w_H:rnd_w_H + lq_patchsize * sf, :]
+ return lq, hq
+
+
+def degradation_bsrgan(img, sf=4, lq_patchsize=72, isp_model=None):
+ """
+ This is the degradation model of BSRGAN from the paper
+ "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution"
+ ----------
+ img: HXWXC, [0, 1], its size should be large than (lq_patchsizexsf)x(lq_patchsizexsf)
+ sf: scale factor
+ isp_model: camera ISP model
+ Returns
+ -------
+ img: low-quality patch, size: lq_patchsizeXlq_patchsizeXC, range: [0, 1]
+ hq: corresponding high-quality patch, size: (lq_patchsizexsf)X(lq_patchsizexsf)XC, range: [0, 1]
+ """
+ isp_prob, jpeg_prob, scale2_prob = 0.25, 0.9, 0.25
+ sf_ori = sf
+
+ h1, w1 = img.shape[:2]
+ img = img.copy()[:w1 - w1 % sf, :h1 - h1 % sf, ...] # mod crop
+ h, w = img.shape[:2]
+
+ if h < lq_patchsize * sf or w < lq_patchsize * sf:
+ raise ValueError(f'img size ({h1}X{w1}) is too small!')
+
+ hq = img.copy()
+
+ if sf == 4 and random.random() < scale2_prob: # downsample1
+ if np.random.rand() < 0.5:
+ img = cv2.resize(img, (int(1 / 2 * img.shape[1]), int(1 / 2 * img.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ img = util.imresize_np(img, 1 / 2, True)
+ img = np.clip(img, 0.0, 1.0)
+ sf = 2
+
+ shuffle_order = random.sample(range(7), 7)
+ idx1, idx2 = shuffle_order.index(2), shuffle_order.index(3)
+ if idx1 > idx2: # keep downsample3 last
+ shuffle_order[idx1], shuffle_order[idx2] = shuffle_order[idx2], shuffle_order[idx1]
+
+ for i in shuffle_order:
+
+ if i == 0:
+ img = add_blur(img, sf=sf)
+
+ elif i == 1:
+ img = add_blur(img, sf=sf)
+
+ elif i == 2:
+ a, b = img.shape[1], img.shape[0]
+ # downsample2
+ if random.random() < 0.75:
+ sf1 = random.uniform(1, 2 * sf)
+ img = cv2.resize(img, (int(1 / sf1 * img.shape[1]), int(1 / sf1 * img.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ k = fspecial('gaussian', 25, random.uniform(0.1, 0.6 * sf))
+ k_shifted = shift_pixel(k, sf)
+ k_shifted = k_shifted / k_shifted.sum() # blur with shifted kernel
+ img = ndimage.filters.convolve(img, np.expand_dims(k_shifted, axis=2), mode='mirror')
+ img = img[0::sf, 0::sf, ...] # nearest downsampling
+ img = np.clip(img, 0.0, 1.0)
+
+ elif i == 3:
+ # downsample3
+ img = cv2.resize(img, (int(1 / sf * a), int(1 / sf * b)), interpolation=random.choice([1, 2, 3]))
+ img = np.clip(img, 0.0, 1.0)
+
+ elif i == 4:
+ # add Gaussian noise
+ img = add_Gaussian_noise(img, noise_level1=2, noise_level2=25)
+
+ elif i == 5:
+ # add JPEG noise
+ if random.random() < jpeg_prob:
+ img = add_JPEG_noise(img)
+
+ elif i == 6:
+ # add processed camera sensor noise
+ if random.random() < isp_prob and isp_model is not None:
+ with torch.no_grad():
+ img, hq = isp_model.forward(img.copy(), hq)
+
+ # add final JPEG compression noise
+ img = add_JPEG_noise(img)
+
+ # random crop
+ img, hq = random_crop(img, hq, sf_ori, lq_patchsize)
+
+ return img, hq
+
+
+# todo no isp_model?
+def degradation_bsrgan_variant(image, sf=4, isp_model=None):
+ """
+ This is the degradation model of BSRGAN from the paper
+ "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution"
+ ----------
+ sf: scale factor
+ isp_model: camera ISP model
+ Returns
+ -------
+ img: low-quality patch, size: lq_patchsizeXlq_patchsizeXC, range: [0, 1]
+ hq: corresponding high-quality patch, size: (lq_patchsizexsf)X(lq_patchsizexsf)XC, range: [0, 1]
+ """
+ image = util.uint2single(image)
+ isp_prob, jpeg_prob, scale2_prob = 0.25, 0.9, 0.25
+ sf_ori = sf
+
+ h1, w1 = image.shape[:2]
+ image = image.copy()[:w1 - w1 % sf, :h1 - h1 % sf, ...] # mod crop
+ h, w = image.shape[:2]
+
+ hq = image.copy()
+
+ if sf == 4 and random.random() < scale2_prob: # downsample1
+ if np.random.rand() < 0.5:
+ image = cv2.resize(image, (int(1 / 2 * image.shape[1]), int(1 / 2 * image.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ image = util.imresize_np(image, 1 / 2, True)
+ image = np.clip(image, 0.0, 1.0)
+ sf = 2
+
+ shuffle_order = random.sample(range(7), 7)
+ idx1, idx2 = shuffle_order.index(2), shuffle_order.index(3)
+ if idx1 > idx2: # keep downsample3 last
+ shuffle_order[idx1], shuffle_order[idx2] = shuffle_order[idx2], shuffle_order[idx1]
+
+ for i in shuffle_order:
+
+ if i == 0:
+ image = add_blur(image, sf=sf)
+
+ elif i == 1:
+ image = add_blur(image, sf=sf)
+
+ elif i == 2:
+ a, b = image.shape[1], image.shape[0]
+ # downsample2
+ if random.random() < 0.75:
+ sf1 = random.uniform(1, 2 * sf)
+ image = cv2.resize(image, (int(1 / sf1 * image.shape[1]), int(1 / sf1 * image.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ k = fspecial('gaussian', 25, random.uniform(0.1, 0.6 * sf))
+ k_shifted = shift_pixel(k, sf)
+ k_shifted = k_shifted / k_shifted.sum() # blur with shifted kernel
+ image = ndimage.filters.convolve(image, np.expand_dims(k_shifted, axis=2), mode='mirror')
+ image = image[0::sf, 0::sf, ...] # nearest downsampling
+ image = np.clip(image, 0.0, 1.0)
+
+ elif i == 3:
+ # downsample3
+ image = cv2.resize(image, (int(1 / sf * a), int(1 / sf * b)), interpolation=random.choice([1, 2, 3]))
+ image = np.clip(image, 0.0, 1.0)
+
+ elif i == 4:
+ # add Gaussian noise
+ image = add_Gaussian_noise(image, noise_level1=2, noise_level2=25)
+
+ elif i == 5:
+ # add JPEG noise
+ if random.random() < jpeg_prob:
+ image = add_JPEG_noise(image)
+
+ # elif i == 6:
+ # # add processed camera sensor noise
+ # if random.random() < isp_prob and isp_model is not None:
+ # with torch.no_grad():
+ # img, hq = isp_model.forward(img.copy(), hq)
+
+ # add final JPEG compression noise
+ image = add_JPEG_noise(image)
+ image = util.single2uint(image)
+ example = {"image":image}
+ return example
+
+
+# TODO incase there is a pickle error one needs to replace a += x with a = a + x in add_speckle_noise etc...
+def degradation_bsrgan_plus(img, sf=4, shuffle_prob=0.5, use_sharp=True, lq_patchsize=64, isp_model=None):
+ """
+ This is an extended degradation model by combining
+ the degradation models of BSRGAN and Real-ESRGAN
+ ----------
+ img: HXWXC, [0, 1], its size should be large than (lq_patchsizexsf)x(lq_patchsizexsf)
+ sf: scale factor
+ use_shuffle: the degradation shuffle
+ use_sharp: sharpening the img
+ Returns
+ -------
+ img: low-quality patch, size: lq_patchsizeXlq_patchsizeXC, range: [0, 1]
+ hq: corresponding high-quality patch, size: (lq_patchsizexsf)X(lq_patchsizexsf)XC, range: [0, 1]
+ """
+
+ h1, w1 = img.shape[:2]
+ img = img.copy()[:w1 - w1 % sf, :h1 - h1 % sf, ...] # mod crop
+ h, w = img.shape[:2]
+
+ if h < lq_patchsize * sf or w < lq_patchsize * sf:
+ raise ValueError(f'img size ({h1}X{w1}) is too small!')
+
+ if use_sharp:
+ img = add_sharpening(img)
+ hq = img.copy()
+
+ if random.random() < shuffle_prob:
+ shuffle_order = random.sample(range(13), 13)
+ else:
+ shuffle_order = list(range(13))
+ # local shuffle for noise, JPEG is always the last one
+ shuffle_order[2:6] = random.sample(shuffle_order[2:6], len(range(2, 6)))
+ shuffle_order[9:13] = random.sample(shuffle_order[9:13], len(range(9, 13)))
+
+ poisson_prob, speckle_prob, isp_prob = 0.1, 0.1, 0.1
+
+ for i in shuffle_order:
+ if i == 0:
+ img = add_blur(img, sf=sf)
+ elif i == 1:
+ img = add_resize(img, sf=sf)
+ elif i == 2:
+ img = add_Gaussian_noise(img, noise_level1=2, noise_level2=25)
+ elif i == 3:
+ if random.random() < poisson_prob:
+ img = add_Poisson_noise(img)
+ elif i == 4:
+ if random.random() < speckle_prob:
+ img = add_speckle_noise(img)
+ elif i == 5:
+ if random.random() < isp_prob and isp_model is not None:
+ with torch.no_grad():
+ img, hq = isp_model.forward(img.copy(), hq)
+ elif i == 6:
+ img = add_JPEG_noise(img)
+ elif i == 7:
+ img = add_blur(img, sf=sf)
+ elif i == 8:
+ img = add_resize(img, sf=sf)
+ elif i == 9:
+ img = add_Gaussian_noise(img, noise_level1=2, noise_level2=25)
+ elif i == 10:
+ if random.random() < poisson_prob:
+ img = add_Poisson_noise(img)
+ elif i == 11:
+ if random.random() < speckle_prob:
+ img = add_speckle_noise(img)
+ elif i == 12:
+ if random.random() < isp_prob and isp_model is not None:
+ with torch.no_grad():
+ img, hq = isp_model.forward(img.copy(), hq)
+ else:
+ print('check the shuffle!')
+
+ # resize to desired size
+ img = cv2.resize(img, (int(1 / sf * hq.shape[1]), int(1 / sf * hq.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+
+ # add final JPEG compression noise
+ img = add_JPEG_noise(img)
+
+ # random crop
+ img, hq = random_crop(img, hq, sf, lq_patchsize)
+
+ return img, hq
+
+
+if __name__ == '__main__':
+ print("hey")
+ img = util.imread_uint('utils/test.png', 3)
+ print(img)
+ img = util.uint2single(img)
+ print(img)
+ img = img[:448, :448]
+ h = img.shape[0] // 4
+ print("resizing to", h)
+ sf = 4
+ deg_fn = partial(degradation_bsrgan_variant, sf=sf)
+ for i in range(20):
+ print(i)
+ img_lq = deg_fn(img)
+ print(img_lq)
+ img_lq_bicubic = albumentations.SmallestMaxSize(max_size=h, interpolation=cv2.INTER_CUBIC)(image=img)["image"]
+ print(img_lq.shape)
+ print("bicubic", img_lq_bicubic.shape)
+ print(img_hq.shape)
+ lq_nearest = cv2.resize(util.single2uint(img_lq), (int(sf * img_lq.shape[1]), int(sf * img_lq.shape[0])),
+ interpolation=0)
+ lq_bicubic_nearest = cv2.resize(util.single2uint(img_lq_bicubic), (int(sf * img_lq.shape[1]), int(sf * img_lq.shape[0])),
+ interpolation=0)
+ img_concat = np.concatenate([lq_bicubic_nearest, lq_nearest, util.single2uint(img_hq)], axis=1)
+ util.imsave(img_concat, str(i) + '.png')
+
+
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/bsrgan_light.py b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/bsrgan_light.py
new file mode 100644
index 0000000..9e1f823
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/bsrgan_light.py
@@ -0,0 +1,650 @@
+# -*- coding: utf-8 -*-
+import numpy as np
+import cv2
+import torch
+
+from functools import partial
+import random
+from scipy import ndimage
+import scipy
+import scipy.stats as ss
+from scipy.interpolate import interp2d
+from scipy.linalg import orth
+import albumentations
+
+import ldm.modules.image_degradation.utils_image as util
+
+"""
+# --------------------------------------------
+# Super-Resolution
+# --------------------------------------------
+#
+# Kai Zhang (cskaizhang@gmail.com)
+# https://github.com/cszn
+# From 2019/03--2021/08
+# --------------------------------------------
+"""
+
+
+def modcrop_np(img, sf):
+ '''
+ Args:
+ img: numpy image, WxH or WxHxC
+ sf: scale factor
+ Return:
+ cropped image
+ '''
+ w, h = img.shape[:2]
+ im = np.copy(img)
+ return im[:w - w % sf, :h - h % sf, ...]
+
+
+"""
+# --------------------------------------------
+# anisotropic Gaussian kernels
+# --------------------------------------------
+"""
+
+
+def analytic_kernel(k):
+ """Calculate the X4 kernel from the X2 kernel (for proof see appendix in paper)"""
+ k_size = k.shape[0]
+ # Calculate the big kernels size
+ big_k = np.zeros((3 * k_size - 2, 3 * k_size - 2))
+ # Loop over the small kernel to fill the big one
+ for r in range(k_size):
+ for c in range(k_size):
+ big_k[2 * r:2 * r + k_size, 2 * c:2 * c + k_size] += k[r, c] * k
+ # Crop the edges of the big kernel to ignore very small values and increase run time of SR
+ crop = k_size // 2
+ cropped_big_k = big_k[crop:-crop, crop:-crop]
+ # Normalize to 1
+ return cropped_big_k / cropped_big_k.sum()
+
+
+def anisotropic_Gaussian(ksize=15, theta=np.pi, l1=6, l2=6):
+ """ generate an anisotropic Gaussian kernel
+ Args:
+ ksize : e.g., 15, kernel size
+ theta : [0, pi], rotation angle range
+ l1 : [0.1,50], scaling of eigenvalues
+ l2 : [0.1,l1], scaling of eigenvalues
+ If l1 = l2, will get an isotropic Gaussian kernel.
+ Returns:
+ k : kernel
+ """
+
+ v = np.dot(np.array([[np.cos(theta), -np.sin(theta)], [np.sin(theta), np.cos(theta)]]), np.array([1., 0.]))
+ V = np.array([[v[0], v[1]], [v[1], -v[0]]])
+ D = np.array([[l1, 0], [0, l2]])
+ Sigma = np.dot(np.dot(V, D), np.linalg.inv(V))
+ k = gm_blur_kernel(mean=[0, 0], cov=Sigma, size=ksize)
+
+ return k
+
+
+def gm_blur_kernel(mean, cov, size=15):
+ center = size / 2.0 + 0.5
+ k = np.zeros([size, size])
+ for y in range(size):
+ for x in range(size):
+ cy = y - center + 1
+ cx = x - center + 1
+ k[y, x] = ss.multivariate_normal.pdf([cx, cy], mean=mean, cov=cov)
+
+ k = k / np.sum(k)
+ return k
+
+
+def shift_pixel(x, sf, upper_left=True):
+ """shift pixel for super-resolution with different scale factors
+ Args:
+ x: WxHxC or WxH
+ sf: scale factor
+ upper_left: shift direction
+ """
+ h, w = x.shape[:2]
+ shift = (sf - 1) * 0.5
+ xv, yv = np.arange(0, w, 1.0), np.arange(0, h, 1.0)
+ if upper_left:
+ x1 = xv + shift
+ y1 = yv + shift
+ else:
+ x1 = xv - shift
+ y1 = yv - shift
+
+ x1 = np.clip(x1, 0, w - 1)
+ y1 = np.clip(y1, 0, h - 1)
+
+ if x.ndim == 2:
+ x = interp2d(xv, yv, x)(x1, y1)
+ if x.ndim == 3:
+ for i in range(x.shape[-1]):
+ x[:, :, i] = interp2d(xv, yv, x[:, :, i])(x1, y1)
+
+ return x
+
+
+def blur(x, k):
+ '''
+ x: image, NxcxHxW
+ k: kernel, Nx1xhxw
+ '''
+ n, c = x.shape[:2]
+ p1, p2 = (k.shape[-2] - 1) // 2, (k.shape[-1] - 1) // 2
+ x = torch.nn.functional.pad(x, pad=(p1, p2, p1, p2), mode='replicate')
+ k = k.repeat(1, c, 1, 1)
+ k = k.view(-1, 1, k.shape[2], k.shape[3])
+ x = x.view(1, -1, x.shape[2], x.shape[3])
+ x = torch.nn.functional.conv2d(x, k, bias=None, stride=1, padding=0, groups=n * c)
+ x = x.view(n, c, x.shape[2], x.shape[3])
+
+ return x
+
+
+def gen_kernel(k_size=np.array([15, 15]), scale_factor=np.array([4, 4]), min_var=0.6, max_var=10., noise_level=0):
+ """"
+ # modified version of https://github.com/assafshocher/BlindSR_dataset_generator
+ # Kai Zhang
+ # min_var = 0.175 * sf # variance of the gaussian kernel will be sampled between min_var and max_var
+ # max_var = 2.5 * sf
+ """
+ # Set random eigen-vals (lambdas) and angle (theta) for COV matrix
+ lambda_1 = min_var + np.random.rand() * (max_var - min_var)
+ lambda_2 = min_var + np.random.rand() * (max_var - min_var)
+ theta = np.random.rand() * np.pi # random theta
+ noise = -noise_level + np.random.rand(*k_size) * noise_level * 2
+
+ # Set COV matrix using Lambdas and Theta
+ LAMBDA = np.diag([lambda_1, lambda_2])
+ Q = np.array([[np.cos(theta), -np.sin(theta)],
+ [np.sin(theta), np.cos(theta)]])
+ SIGMA = Q @ LAMBDA @ Q.T
+ INV_SIGMA = np.linalg.inv(SIGMA)[None, None, :, :]
+
+ # Set expectation position (shifting kernel for aligned image)
+ MU = k_size // 2 - 0.5 * (scale_factor - 1) # - 0.5 * (scale_factor - k_size % 2)
+ MU = MU[None, None, :, None]
+
+ # Create meshgrid for Gaussian
+ [X, Y] = np.meshgrid(range(k_size[0]), range(k_size[1]))
+ Z = np.stack([X, Y], 2)[:, :, :, None]
+
+ # Calcualte Gaussian for every pixel of the kernel
+ ZZ = Z - MU
+ ZZ_t = ZZ.transpose(0, 1, 3, 2)
+ raw_kernel = np.exp(-0.5 * np.squeeze(ZZ_t @ INV_SIGMA @ ZZ)) * (1 + noise)
+
+ # shift the kernel so it will be centered
+ # raw_kernel_centered = kernel_shift(raw_kernel, scale_factor)
+
+ # Normalize the kernel and return
+ # kernel = raw_kernel_centered / np.sum(raw_kernel_centered)
+ kernel = raw_kernel / np.sum(raw_kernel)
+ return kernel
+
+
+def fspecial_gaussian(hsize, sigma):
+ hsize = [hsize, hsize]
+ siz = [(hsize[0] - 1.0) / 2.0, (hsize[1] - 1.0) / 2.0]
+ std = sigma
+ [x, y] = np.meshgrid(np.arange(-siz[1], siz[1] + 1), np.arange(-siz[0], siz[0] + 1))
+ arg = -(x * x + y * y) / (2 * std * std)
+ h = np.exp(arg)
+ h[h < scipy.finfo(float).eps * h.max()] = 0
+ sumh = h.sum()
+ if sumh != 0:
+ h = h / sumh
+ return h
+
+
+def fspecial_laplacian(alpha):
+ alpha = max([0, min([alpha, 1])])
+ h1 = alpha / (alpha + 1)
+ h2 = (1 - alpha) / (alpha + 1)
+ h = [[h1, h2, h1], [h2, -4 / (alpha + 1), h2], [h1, h2, h1]]
+ h = np.array(h)
+ return h
+
+
+def fspecial(filter_type, *args, **kwargs):
+ '''
+ python code from:
+ https://github.com/ronaldosena/imagens-medicas-2/blob/40171a6c259edec7827a6693a93955de2bd39e76/Aulas/aula_2_-_uniform_filter/matlab_fspecial.py
+ '''
+ if filter_type == 'gaussian':
+ return fspecial_gaussian(*args, **kwargs)
+ if filter_type == 'laplacian':
+ return fspecial_laplacian(*args, **kwargs)
+
+
+"""
+# --------------------------------------------
+# degradation models
+# --------------------------------------------
+"""
+
+
+def bicubic_degradation(x, sf=3):
+ '''
+ Args:
+ x: HxWxC image, [0, 1]
+ sf: down-scale factor
+ Return:
+ bicubicly downsampled LR image
+ '''
+ x = util.imresize_np(x, scale=1 / sf)
+ return x
+
+
+def srmd_degradation(x, k, sf=3):
+ ''' blur + bicubic downsampling
+ Args:
+ x: HxWxC image, [0, 1]
+ k: hxw, double
+ sf: down-scale factor
+ Return:
+ downsampled LR image
+ Reference:
+ @inproceedings{zhang2018learning,
+ title={Learning a single convolutional super-resolution network for multiple degradations},
+ author={Zhang, Kai and Zuo, Wangmeng and Zhang, Lei},
+ booktitle={IEEE Conference on Computer Vision and Pattern Recognition},
+ pages={3262--3271},
+ year={2018}
+ }
+ '''
+ x = ndimage.filters.convolve(x, np.expand_dims(k, axis=2), mode='wrap') # 'nearest' | 'mirror'
+ x = bicubic_degradation(x, sf=sf)
+ return x
+
+
+def dpsr_degradation(x, k, sf=3):
+ ''' bicubic downsampling + blur
+ Args:
+ x: HxWxC image, [0, 1]
+ k: hxw, double
+ sf: down-scale factor
+ Return:
+ downsampled LR image
+ Reference:
+ @inproceedings{zhang2019deep,
+ title={Deep Plug-and-Play Super-Resolution for Arbitrary Blur Kernels},
+ author={Zhang, Kai and Zuo, Wangmeng and Zhang, Lei},
+ booktitle={IEEE Conference on Computer Vision and Pattern Recognition},
+ pages={1671--1681},
+ year={2019}
+ }
+ '''
+ x = bicubic_degradation(x, sf=sf)
+ x = ndimage.filters.convolve(x, np.expand_dims(k, axis=2), mode='wrap')
+ return x
+
+
+def classical_degradation(x, k, sf=3):
+ ''' blur + downsampling
+ Args:
+ x: HxWxC image, [0, 1]/[0, 255]
+ k: hxw, double
+ sf: down-scale factor
+ Return:
+ downsampled LR image
+ '''
+ x = ndimage.filters.convolve(x, np.expand_dims(k, axis=2), mode='wrap')
+ # x = filters.correlate(x, np.expand_dims(np.flip(k), axis=2))
+ st = 0
+ return x[st::sf, st::sf, ...]
+
+
+def add_sharpening(img, weight=0.5, radius=50, threshold=10):
+ """USM sharpening. borrowed from real-ESRGAN
+ Input image: I; Blurry image: B.
+ 1. K = I + weight * (I - B)
+ 2. Mask = 1 if abs(I - B) > threshold, else: 0
+ 3. Blur mask:
+ 4. Out = Mask * K + (1 - Mask) * I
+ Args:
+ img (Numpy array): Input image, HWC, BGR; float32, [0, 1].
+ weight (float): Sharp weight. Default: 1.
+ radius (float): Kernel size of Gaussian blur. Default: 50.
+ threshold (int):
+ """
+ if radius % 2 == 0:
+ radius += 1
+ blur = cv2.GaussianBlur(img, (radius, radius), 0)
+ residual = img - blur
+ mask = np.abs(residual) * 255 > threshold
+ mask = mask.astype('float32')
+ soft_mask = cv2.GaussianBlur(mask, (radius, radius), 0)
+
+ K = img + weight * residual
+ K = np.clip(K, 0, 1)
+ return soft_mask * K + (1 - soft_mask) * img
+
+
+def add_blur(img, sf=4):
+ wd2 = 4.0 + sf
+ wd = 2.0 + 0.2 * sf
+
+ wd2 = wd2/4
+ wd = wd/4
+
+ if random.random() < 0.5:
+ l1 = wd2 * random.random()
+ l2 = wd2 * random.random()
+ k = anisotropic_Gaussian(ksize=random.randint(2, 11) + 3, theta=random.random() * np.pi, l1=l1, l2=l2)
+ else:
+ k = fspecial('gaussian', random.randint(2, 4) + 3, wd * random.random())
+ img = ndimage.filters.convolve(img, np.expand_dims(k, axis=2), mode='mirror')
+
+ return img
+
+
+def add_resize(img, sf=4):
+ rnum = np.random.rand()
+ if rnum > 0.8: # up
+ sf1 = random.uniform(1, 2)
+ elif rnum < 0.7: # down
+ sf1 = random.uniform(0.5 / sf, 1)
+ else:
+ sf1 = 1.0
+ img = cv2.resize(img, (int(sf1 * img.shape[1]), int(sf1 * img.shape[0])), interpolation=random.choice([1, 2, 3]))
+ img = np.clip(img, 0.0, 1.0)
+
+ return img
+
+
+# def add_Gaussian_noise(img, noise_level1=2, noise_level2=25):
+# noise_level = random.randint(noise_level1, noise_level2)
+# rnum = np.random.rand()
+# if rnum > 0.6: # add color Gaussian noise
+# img += np.random.normal(0, noise_level / 255.0, img.shape).astype(np.float32)
+# elif rnum < 0.4: # add grayscale Gaussian noise
+# img += np.random.normal(0, noise_level / 255.0, (*img.shape[:2], 1)).astype(np.float32)
+# else: # add noise
+# L = noise_level2 / 255.
+# D = np.diag(np.random.rand(3))
+# U = orth(np.random.rand(3, 3))
+# conv = np.dot(np.dot(np.transpose(U), D), U)
+# img += np.random.multivariate_normal([0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2]).astype(np.float32)
+# img = np.clip(img, 0.0, 1.0)
+# return img
+
+def add_Gaussian_noise(img, noise_level1=2, noise_level2=25):
+ noise_level = random.randint(noise_level1, noise_level2)
+ rnum = np.random.rand()
+ if rnum > 0.6: # add color Gaussian noise
+ img = img + np.random.normal(0, noise_level / 255.0, img.shape).astype(np.float32)
+ elif rnum < 0.4: # add grayscale Gaussian noise
+ img = img + np.random.normal(0, noise_level / 255.0, (*img.shape[:2], 1)).astype(np.float32)
+ else: # add noise
+ L = noise_level2 / 255.
+ D = np.diag(np.random.rand(3))
+ U = orth(np.random.rand(3, 3))
+ conv = np.dot(np.dot(np.transpose(U), D), U)
+ img = img + np.random.multivariate_normal([0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2]).astype(np.float32)
+ img = np.clip(img, 0.0, 1.0)
+ return img
+
+
+def add_speckle_noise(img, noise_level1=2, noise_level2=25):
+ noise_level = random.randint(noise_level1, noise_level2)
+ img = np.clip(img, 0.0, 1.0)
+ rnum = random.random()
+ if rnum > 0.6:
+ img += img * np.random.normal(0, noise_level / 255.0, img.shape).astype(np.float32)
+ elif rnum < 0.4:
+ img += img * np.random.normal(0, noise_level / 255.0, (*img.shape[:2], 1)).astype(np.float32)
+ else:
+ L = noise_level2 / 255.
+ D = np.diag(np.random.rand(3))
+ U = orth(np.random.rand(3, 3))
+ conv = np.dot(np.dot(np.transpose(U), D), U)
+ img += img * np.random.multivariate_normal([0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2]).astype(np.float32)
+ img = np.clip(img, 0.0, 1.0)
+ return img
+
+
+def add_Poisson_noise(img):
+ img = np.clip((img * 255.0).round(), 0, 255) / 255.
+ vals = 10 ** (2 * random.random() + 2.0) # [2, 4]
+ if random.random() < 0.5:
+ img = np.random.poisson(img * vals).astype(np.float32) / vals
+ else:
+ img_gray = np.dot(img[..., :3], [0.299, 0.587, 0.114])
+ img_gray = np.clip((img_gray * 255.0).round(), 0, 255) / 255.
+ noise_gray = np.random.poisson(img_gray * vals).astype(np.float32) / vals - img_gray
+ img += noise_gray[:, :, np.newaxis]
+ img = np.clip(img, 0.0, 1.0)
+ return img
+
+
+def add_JPEG_noise(img):
+ quality_factor = random.randint(80, 95)
+ img = cv2.cvtColor(util.single2uint(img), cv2.COLOR_RGB2BGR)
+ result, encimg = cv2.imencode('.jpg', img, [int(cv2.IMWRITE_JPEG_QUALITY), quality_factor])
+ img = cv2.imdecode(encimg, 1)
+ img = cv2.cvtColor(util.uint2single(img), cv2.COLOR_BGR2RGB)
+ return img
+
+
+def random_crop(lq, hq, sf=4, lq_patchsize=64):
+ h, w = lq.shape[:2]
+ rnd_h = random.randint(0, h - lq_patchsize)
+ rnd_w = random.randint(0, w - lq_patchsize)
+ lq = lq[rnd_h:rnd_h + lq_patchsize, rnd_w:rnd_w + lq_patchsize, :]
+
+ rnd_h_H, rnd_w_H = int(rnd_h * sf), int(rnd_w * sf)
+ hq = hq[rnd_h_H:rnd_h_H + lq_patchsize * sf, rnd_w_H:rnd_w_H + lq_patchsize * sf, :]
+ return lq, hq
+
+
+def degradation_bsrgan(img, sf=4, lq_patchsize=72, isp_model=None):
+ """
+ This is the degradation model of BSRGAN from the paper
+ "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution"
+ ----------
+ img: HXWXC, [0, 1], its size should be large than (lq_patchsizexsf)x(lq_patchsizexsf)
+ sf: scale factor
+ isp_model: camera ISP model
+ Returns
+ -------
+ img: low-quality patch, size: lq_patchsizeXlq_patchsizeXC, range: [0, 1]
+ hq: corresponding high-quality patch, size: (lq_patchsizexsf)X(lq_patchsizexsf)XC, range: [0, 1]
+ """
+ isp_prob, jpeg_prob, scale2_prob = 0.25, 0.9, 0.25
+ sf_ori = sf
+
+ h1, w1 = img.shape[:2]
+ img = img.copy()[:w1 - w1 % sf, :h1 - h1 % sf, ...] # mod crop
+ h, w = img.shape[:2]
+
+ if h < lq_patchsize * sf or w < lq_patchsize * sf:
+ raise ValueError(f'img size ({h1}X{w1}) is too small!')
+
+ hq = img.copy()
+
+ if sf == 4 and random.random() < scale2_prob: # downsample1
+ if np.random.rand() < 0.5:
+ img = cv2.resize(img, (int(1 / 2 * img.shape[1]), int(1 / 2 * img.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ img = util.imresize_np(img, 1 / 2, True)
+ img = np.clip(img, 0.0, 1.0)
+ sf = 2
+
+ shuffle_order = random.sample(range(7), 7)
+ idx1, idx2 = shuffle_order.index(2), shuffle_order.index(3)
+ if idx1 > idx2: # keep downsample3 last
+ shuffle_order[idx1], shuffle_order[idx2] = shuffle_order[idx2], shuffle_order[idx1]
+
+ for i in shuffle_order:
+
+ if i == 0:
+ img = add_blur(img, sf=sf)
+
+ elif i == 1:
+ img = add_blur(img, sf=sf)
+
+ elif i == 2:
+ a, b = img.shape[1], img.shape[0]
+ # downsample2
+ if random.random() < 0.75:
+ sf1 = random.uniform(1, 2 * sf)
+ img = cv2.resize(img, (int(1 / sf1 * img.shape[1]), int(1 / sf1 * img.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ k = fspecial('gaussian', 25, random.uniform(0.1, 0.6 * sf))
+ k_shifted = shift_pixel(k, sf)
+ k_shifted = k_shifted / k_shifted.sum() # blur with shifted kernel
+ img = ndimage.filters.convolve(img, np.expand_dims(k_shifted, axis=2), mode='mirror')
+ img = img[0::sf, 0::sf, ...] # nearest downsampling
+ img = np.clip(img, 0.0, 1.0)
+
+ elif i == 3:
+ # downsample3
+ img = cv2.resize(img, (int(1 / sf * a), int(1 / sf * b)), interpolation=random.choice([1, 2, 3]))
+ img = np.clip(img, 0.0, 1.0)
+
+ elif i == 4:
+ # add Gaussian noise
+ img = add_Gaussian_noise(img, noise_level1=2, noise_level2=8)
+
+ elif i == 5:
+ # add JPEG noise
+ if random.random() < jpeg_prob:
+ img = add_JPEG_noise(img)
+
+ elif i == 6:
+ # add processed camera sensor noise
+ if random.random() < isp_prob and isp_model is not None:
+ with torch.no_grad():
+ img, hq = isp_model.forward(img.copy(), hq)
+
+ # add final JPEG compression noise
+ img = add_JPEG_noise(img)
+
+ # random crop
+ img, hq = random_crop(img, hq, sf_ori, lq_patchsize)
+
+ return img, hq
+
+
+# todo no isp_model?
+def degradation_bsrgan_variant(image, sf=4, isp_model=None):
+ """
+ This is the degradation model of BSRGAN from the paper
+ "Designing a Practical Degradation Model for Deep Blind Image Super-Resolution"
+ ----------
+ sf: scale factor
+ isp_model: camera ISP model
+ Returns
+ -------
+ img: low-quality patch, size: lq_patchsizeXlq_patchsizeXC, range: [0, 1]
+ hq: corresponding high-quality patch, size: (lq_patchsizexsf)X(lq_patchsizexsf)XC, range: [0, 1]
+ """
+ image = util.uint2single(image)
+ isp_prob, jpeg_prob, scale2_prob = 0.25, 0.9, 0.25
+ sf_ori = sf
+
+ h1, w1 = image.shape[:2]
+ image = image.copy()[:w1 - w1 % sf, :h1 - h1 % sf, ...] # mod crop
+ h, w = image.shape[:2]
+
+ hq = image.copy()
+
+ if sf == 4 and random.random() < scale2_prob: # downsample1
+ if np.random.rand() < 0.5:
+ image = cv2.resize(image, (int(1 / 2 * image.shape[1]), int(1 / 2 * image.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ image = util.imresize_np(image, 1 / 2, True)
+ image = np.clip(image, 0.0, 1.0)
+ sf = 2
+
+ shuffle_order = random.sample(range(7), 7)
+ idx1, idx2 = shuffle_order.index(2), shuffle_order.index(3)
+ if idx1 > idx2: # keep downsample3 last
+ shuffle_order[idx1], shuffle_order[idx2] = shuffle_order[idx2], shuffle_order[idx1]
+
+ for i in shuffle_order:
+
+ if i == 0:
+ image = add_blur(image, sf=sf)
+
+ # elif i == 1:
+ # image = add_blur(image, sf=sf)
+
+ if i == 0:
+ pass
+
+ elif i == 2:
+ a, b = image.shape[1], image.shape[0]
+ # downsample2
+ if random.random() < 0.8:
+ sf1 = random.uniform(1, 2 * sf)
+ image = cv2.resize(image, (int(1 / sf1 * image.shape[1]), int(1 / sf1 * image.shape[0])),
+ interpolation=random.choice([1, 2, 3]))
+ else:
+ k = fspecial('gaussian', 25, random.uniform(0.1, 0.6 * sf))
+ k_shifted = shift_pixel(k, sf)
+ k_shifted = k_shifted / k_shifted.sum() # blur with shifted kernel
+ image = ndimage.filters.convolve(image, np.expand_dims(k_shifted, axis=2), mode='mirror')
+ image = image[0::sf, 0::sf, ...] # nearest downsampling
+
+ image = np.clip(image, 0.0, 1.0)
+
+ elif i == 3:
+ # downsample3
+ image = cv2.resize(image, (int(1 / sf * a), int(1 / sf * b)), interpolation=random.choice([1, 2, 3]))
+ image = np.clip(image, 0.0, 1.0)
+
+ elif i == 4:
+ # add Gaussian noise
+ image = add_Gaussian_noise(image, noise_level1=1, noise_level2=2)
+
+ elif i == 5:
+ # add JPEG noise
+ if random.random() < jpeg_prob:
+ image = add_JPEG_noise(image)
+ #
+ # elif i == 6:
+ # # add processed camera sensor noise
+ # if random.random() < isp_prob and isp_model is not None:
+ # with torch.no_grad():
+ # img, hq = isp_model.forward(img.copy(), hq)
+
+ # add final JPEG compression noise
+ image = add_JPEG_noise(image)
+ image = util.single2uint(image)
+ example = {"image": image}
+ return example
+
+
+
+
+if __name__ == '__main__':
+ print("hey")
+ img = util.imread_uint('utils/test.png', 3)
+ img = img[:448, :448]
+ h = img.shape[0] // 4
+ print("resizing to", h)
+ sf = 4
+ deg_fn = partial(degradation_bsrgan_variant, sf=sf)
+ for i in range(20):
+ print(i)
+ img_hq = img
+ img_lq = deg_fn(img)["image"]
+ img_hq, img_lq = util.uint2single(img_hq), util.uint2single(img_lq)
+ print(img_lq)
+ img_lq_bicubic = albumentations.SmallestMaxSize(max_size=h, interpolation=cv2.INTER_CUBIC)(image=img_hq)["image"]
+ print(img_lq.shape)
+ print("bicubic", img_lq_bicubic.shape)
+ print(img_hq.shape)
+ lq_nearest = cv2.resize(util.single2uint(img_lq), (int(sf * img_lq.shape[1]), int(sf * img_lq.shape[0])),
+ interpolation=0)
+ lq_bicubic_nearest = cv2.resize(util.single2uint(img_lq_bicubic),
+ (int(sf * img_lq.shape[1]), int(sf * img_lq.shape[0])),
+ interpolation=0)
+ img_concat = np.concatenate([lq_bicubic_nearest, lq_nearest, util.single2uint(img_hq)], axis=1)
+ util.imsave(img_concat, str(i) + '.png')
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/utils/test.png b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/utils/test.png
new file mode 100644
index 0000000..4249b43
Binary files /dev/null and b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/utils/test.png differ
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/utils_image.py b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/utils_image.py
new file mode 100644
index 0000000..0175f15
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/image_degradation/utils_image.py
@@ -0,0 +1,916 @@
+import os
+import math
+import random
+import numpy as np
+import torch
+import cv2
+from torchvision.utils import make_grid
+from datetime import datetime
+#import matplotlib.pyplot as plt # TODO: check with Dominik, also bsrgan.py vs bsrgan_light.py
+
+
+os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
+
+
+'''
+# --------------------------------------------
+# Kai Zhang (github: https://github.com/cszn)
+# 03/Mar/2019
+# --------------------------------------------
+# https://github.com/twhui/SRGAN-pyTorch
+# https://github.com/xinntao/BasicSR
+# --------------------------------------------
+'''
+
+
+IMG_EXTENSIONS = ['.jpg', '.JPG', '.jpeg', '.JPEG', '.png', '.PNG', '.ppm', '.PPM', '.bmp', '.BMP', '.tif']
+
+
+def is_image_file(filename):
+ return any(filename.endswith(extension) for extension in IMG_EXTENSIONS)
+
+
+def get_timestamp():
+ return datetime.now().strftime('%y%m%d-%H%M%S')
+
+
+def imshow(x, title=None, cbar=False, figsize=None):
+ plt.figure(figsize=figsize)
+ plt.imshow(np.squeeze(x), interpolation='nearest', cmap='gray')
+ if title:
+ plt.title(title)
+ if cbar:
+ plt.colorbar()
+ plt.show()
+
+
+def surf(Z, cmap='rainbow', figsize=None):
+ plt.figure(figsize=figsize)
+ ax3 = plt.axes(projection='3d')
+
+ w, h = Z.shape[:2]
+ xx = np.arange(0,w,1)
+ yy = np.arange(0,h,1)
+ X, Y = np.meshgrid(xx, yy)
+ ax3.plot_surface(X,Y,Z,cmap=cmap)
+ #ax3.contour(X,Y,Z, zdim='z',offset=-2，cmap=cmap)
+ plt.show()
+
+
+'''
+# --------------------------------------------
+# get image pathes
+# --------------------------------------------
+'''
+
+
+def get_image_paths(dataroot):
+ paths = None # return None if dataroot is None
+ if dataroot is not None:
+ paths = sorted(_get_paths_from_images(dataroot))
+ return paths
+
+
+def _get_paths_from_images(path):
+ assert os.path.isdir(path), '{:s} is not a valid directory'.format(path)
+ images = []
+ for dirpath, _, fnames in sorted(os.walk(path)):
+ for fname in sorted(fnames):
+ if is_image_file(fname):
+ img_path = os.path.join(dirpath, fname)
+ images.append(img_path)
+ assert images, '{:s} has no valid image file'.format(path)
+ return images
+
+
+'''
+# --------------------------------------------
+# split large images into small images
+# --------------------------------------------
+'''
+
+
+def patches_from_image(img, p_size=512, p_overlap=64, p_max=800):
+ w, h = img.shape[:2]
+ patches = []
+ if w > p_max and h > p_max:
+ w1 = list(np.arange(0, w-p_size, p_size-p_overlap, dtype=np.int))
+ h1 = list(np.arange(0, h-p_size, p_size-p_overlap, dtype=np.int))
+ w1.append(w-p_size)
+ h1.append(h-p_size)
+# print(w1)
+# print(h1)
+ for i in w1:
+ for j in h1:
+ patches.append(img[i:i+p_size, j:j+p_size,:])
+ else:
+ patches.append(img)
+
+ return patches
+
+
+def imssave(imgs, img_path):
+ """
+ imgs: list, N images of size WxHxC
+ """
+ img_name, ext = os.path.splitext(os.path.basename(img_path))
+
+ for i, img in enumerate(imgs):
+ if img.ndim == 3:
+ img = img[:, :, [2, 1, 0]]
+ new_path = os.path.join(os.path.dirname(img_path), img_name+str('_s{:04d}'.format(i))+'.png')
+ cv2.imwrite(new_path, img)
+
+
+def split_imageset(original_dataroot, taget_dataroot, n_channels=3, p_size=800, p_overlap=96, p_max=1000):
+ """
+ split the large images from original_dataroot into small overlapped images with size (p_size)x(p_size),
+ and save them into taget_dataroot; only the images with larger size than (p_max)x(p_max)
+ will be splitted.
+ Args:
+ original_dataroot:
+ taget_dataroot:
+ p_size: size of small images
+ p_overlap: patch size in training is a good choice
+ p_max: images with smaller size than (p_max)x(p_max) keep unchanged.
+ """
+ paths = get_image_paths(original_dataroot)
+ for img_path in paths:
+ # img_name, ext = os.path.splitext(os.path.basename(img_path))
+ img = imread_uint(img_path, n_channels=n_channels)
+ patches = patches_from_image(img, p_size, p_overlap, p_max)
+ imssave(patches, os.path.join(taget_dataroot,os.path.basename(img_path)))
+ #if original_dataroot == taget_dataroot:
+ #del img_path
+
+'''
+# --------------------------------------------
+# makedir
+# --------------------------------------------
+'''
+
+
+def mkdir(path):
+ if not os.path.exists(path):
+ os.makedirs(path)
+
+
+def mkdirs(paths):
+ if isinstance(paths, str):
+ mkdir(paths)
+ else:
+ for path in paths:
+ mkdir(path)
+
+
+def mkdir_and_rename(path):
+ if os.path.exists(path):
+ new_name = path + '_archived_' + get_timestamp()
+ print('Path already exists. Rename it to [{:s}]'.format(new_name))
+ os.rename(path, new_name)
+ os.makedirs(path)
+
+
+'''
+# --------------------------------------------
+# read image from path
+# opencv is fast, but read BGR numpy image
+# --------------------------------------------
+'''
+
+
+# --------------------------------------------
+# get uint8 image of size HxWxn_channles (RGB)
+# --------------------------------------------
+def imread_uint(path, n_channels=3):
+ # input: path
+ # output: HxWx3(RGB or GGG), or HxWx1 (G)
+ if n_channels == 1:
+ img = cv2.imread(path, 0) # cv2.IMREAD_GRAYSCALE
+ img = np.expand_dims(img, axis=2) # HxWx1
+ elif n_channels == 3:
+ img = cv2.imread(path, cv2.IMREAD_UNCHANGED) # BGR or G
+ if img.ndim == 2:
+ img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB) # GGG
+ else:
+ img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) # RGB
+ return img
+
+
+# --------------------------------------------
+# matlab's imwrite
+# --------------------------------------------
+def imsave(img, img_path):
+ img = np.squeeze(img)
+ if img.ndim == 3:
+ img = img[:, :, [2, 1, 0]]
+ cv2.imwrite(img_path, img)
+
+def imwrite(img, img_path):
+ img = np.squeeze(img)
+ if img.ndim == 3:
+ img = img[:, :, [2, 1, 0]]
+ cv2.imwrite(img_path, img)
+
+
+
+# --------------------------------------------
+# get single image of size HxWxn_channles (BGR)
+# --------------------------------------------
+def read_img(path):
+ # read image by cv2
+ # return: Numpy float32, HWC, BGR, [0,1]
+ img = cv2.imread(path, cv2.IMREAD_UNCHANGED) # cv2.IMREAD_GRAYSCALE
+ img = img.astype(np.float32) / 255.
+ if img.ndim == 2:
+ img = np.expand_dims(img, axis=2)
+ # some images have 4 channels
+ if img.shape[2] > 3:
+ img = img[:, :, :3]
+ return img
+
+
+'''
+# --------------------------------------------
+# image format conversion
+# --------------------------------------------
+# numpy(single) <---> numpy(unit)
+# numpy(single) <---> tensor
+# numpy(unit) <---> tensor
+# --------------------------------------------
+'''
+
+
+# --------------------------------------------
+# numpy(single) [0, 1] <---> numpy(unit)
+# --------------------------------------------
+
+
+def uint2single(img):
+
+ return np.float32(img/255.)
+
+
+def single2uint(img):
+
+ return np.uint8((img.clip(0, 1)*255.).round())
+
+
+def uint162single(img):
+
+ return np.float32(img/65535.)
+
+
+def single2uint16(img):
+
+ return np.uint16((img.clip(0, 1)*65535.).round())
+
+
+# --------------------------------------------
+# numpy(unit) (HxWxC or HxW) <---> tensor
+# --------------------------------------------
+
+
+# convert uint to 4-dimensional torch tensor
+def uint2tensor4(img):
+ if img.ndim == 2:
+ img = np.expand_dims(img, axis=2)
+ return torch.from_numpy(np.ascontiguousarray(img)).permute(2, 0, 1).float().div(255.).unsqueeze(0)
+
+
+# convert uint to 3-dimensional torch tensor
+def uint2tensor3(img):
+ if img.ndim == 2:
+ img = np.expand_dims(img, axis=2)
+ return torch.from_numpy(np.ascontiguousarray(img)).permute(2, 0, 1).float().div(255.)
+
+
+# convert 2/3/4-dimensional torch tensor to uint
+def tensor2uint(img):
+ img = img.data.squeeze().float().clamp_(0, 1).cpu().numpy()
+ if img.ndim == 3:
+ img = np.transpose(img, (1, 2, 0))
+ return np.uint8((img*255.0).round())
+
+
+# --------------------------------------------
+# numpy(single) (HxWxC) <---> tensor
+# --------------------------------------------
+
+
+# convert single (HxWxC) to 3-dimensional torch tensor
+def single2tensor3(img):
+ return torch.from_numpy(np.ascontiguousarray(img)).permute(2, 0, 1).float()
+
+
+# convert single (HxWxC) to 4-dimensional torch tensor
+def single2tensor4(img):
+ return torch.from_numpy(np.ascontiguousarray(img)).permute(2, 0, 1).float().unsqueeze(0)
+
+
+# convert torch tensor to single
+def tensor2single(img):
+ img = img.data.squeeze().float().cpu().numpy()
+ if img.ndim == 3:
+ img = np.transpose(img, (1, 2, 0))
+
+ return img
+
+# convert torch tensor to single
+def tensor2single3(img):
+ img = img.data.squeeze().float().cpu().numpy()
+ if img.ndim == 3:
+ img = np.transpose(img, (1, 2, 0))
+ elif img.ndim == 2:
+ img = np.expand_dims(img, axis=2)
+ return img
+
+
+def single2tensor5(img):
+ return torch.from_numpy(np.ascontiguousarray(img)).permute(2, 0, 1, 3).float().unsqueeze(0)
+
+
+def single32tensor5(img):
+ return torch.from_numpy(np.ascontiguousarray(img)).float().unsqueeze(0).unsqueeze(0)
+
+
+def single42tensor4(img):
+ return torch.from_numpy(np.ascontiguousarray(img)).permute(2, 0, 1, 3).float()
+
+
+# from skimage.io import imread, imsave
+def tensor2img(tensor, out_type=np.uint8, min_max=(0, 1)):
+ '''
+ Converts a torch Tensor into an image Numpy array of BGR channel order
+ Input: 4D(B,(3/1),H,W), 3D(C,H,W), or 2D(H,W), any range, RGB channel order
+ Output: 3D(H,W,C) or 2D(H,W), [0,255], np.uint8 (default)
+ '''
+ tensor = tensor.squeeze().float().cpu().clamp_(*min_max) # squeeze first, then clamp
+ tensor = (tensor - min_max[0]) / (min_max[1] - min_max[0]) # to range [0,1]
+ n_dim = tensor.dim()
+ if n_dim == 4:
+ n_img = len(tensor)
+ img_np = make_grid(tensor, nrow=int(math.sqrt(n_img)), normalize=False).numpy()
+ img_np = np.transpose(img_np[[2, 1, 0], :, :], (1, 2, 0)) # HWC, BGR
+ elif n_dim == 3:
+ img_np = tensor.numpy()
+ img_np = np.transpose(img_np[[2, 1, 0], :, :], (1, 2, 0)) # HWC, BGR
+ elif n_dim == 2:
+ img_np = tensor.numpy()
+ else:
+ raise TypeError(
+ 'Only support 4D, 3D and 2D tensor. But received with dimension: {:d}'.format(n_dim))
+ if out_type == np.uint8:
+ img_np = (img_np * 255.0).round()
+ # Important. Unlike matlab, numpy.unit8() WILL NOT round by default.
+ return img_np.astype(out_type)
+
+
+'''
+# --------------------------------------------
+# Augmentation, flipe and/or rotate
+# --------------------------------------------
+# The following two are enough.
+# (1) augmet_img: numpy image of WxHxC or WxH
+# (2) augment_img_tensor4: tensor image 1xCxWxH
+# --------------------------------------------
+'''
+
+
+def augment_img(img, mode=0):
+ '''Kai Zhang (github: https://github.com/cszn)
+ '''
+ if mode == 0:
+ return img
+ elif mode == 1:
+ return np.flipud(np.rot90(img))
+ elif mode == 2:
+ return np.flipud(img)
+ elif mode == 3:
+ return np.rot90(img, k=3)
+ elif mode == 4:
+ return np.flipud(np.rot90(img, k=2))
+ elif mode == 5:
+ return np.rot90(img)
+ elif mode == 6:
+ return np.rot90(img, k=2)
+ elif mode == 7:
+ return np.flipud(np.rot90(img, k=3))
+
+
+def augment_img_tensor4(img, mode=0):
+ '''Kai Zhang (github: https://github.com/cszn)
+ '''
+ if mode == 0:
+ return img
+ elif mode == 1:
+ return img.rot90(1, [2, 3]).flip([2])
+ elif mode == 2:
+ return img.flip([2])
+ elif mode == 3:
+ return img.rot90(3, [2, 3])
+ elif mode == 4:
+ return img.rot90(2, [2, 3]).flip([2])
+ elif mode == 5:
+ return img.rot90(1, [2, 3])
+ elif mode == 6:
+ return img.rot90(2, [2, 3])
+ elif mode == 7:
+ return img.rot90(3, [2, 3]).flip([2])
+
+
+def augment_img_tensor(img, mode=0):
+ '''Kai Zhang (github: https://github.com/cszn)
+ '''
+ img_size = img.size()
+ img_np = img.data.cpu().numpy()
+ if len(img_size) == 3:
+ img_np = np.transpose(img_np, (1, 2, 0))
+ elif len(img_size) == 4:
+ img_np = np.transpose(img_np, (2, 3, 1, 0))
+ img_np = augment_img(img_np, mode=mode)
+ img_tensor = torch.from_numpy(np.ascontiguousarray(img_np))
+ if len(img_size) == 3:
+ img_tensor = img_tensor.permute(2, 0, 1)
+ elif len(img_size) == 4:
+ img_tensor = img_tensor.permute(3, 2, 0, 1)
+
+ return img_tensor.type_as(img)
+
+
+def augment_img_np3(img, mode=0):
+ if mode == 0:
+ return img
+ elif mode == 1:
+ return img.transpose(1, 0, 2)
+ elif mode == 2:
+ return img[::-1, :, :]
+ elif mode == 3:
+ img = img[::-1, :, :]
+ img = img.transpose(1, 0, 2)
+ return img
+ elif mode == 4:
+ return img[:, ::-1, :]
+ elif mode == 5:
+ img = img[:, ::-1, :]
+ img = img.transpose(1, 0, 2)
+ return img
+ elif mode == 6:
+ img = img[:, ::-1, :]
+ img = img[::-1, :, :]
+ return img
+ elif mode == 7:
+ img = img[:, ::-1, :]
+ img = img[::-1, :, :]
+ img = img.transpose(1, 0, 2)
+ return img
+
+
+def augment_imgs(img_list, hflip=True, rot=True):
+ # horizontal flip OR rotate
+ hflip = hflip and random.random() < 0.5
+ vflip = rot and random.random() < 0.5
+ rot90 = rot and random.random() < 0.5
+
+ def _augment(img):
+ if hflip:
+ img = img[:, ::-1, :]
+ if vflip:
+ img = img[::-1, :, :]
+ if rot90:
+ img = img.transpose(1, 0, 2)
+ return img
+
+ return [_augment(img) for img in img_list]
+
+
+'''
+# --------------------------------------------
+# modcrop and shave
+# --------------------------------------------
+'''
+
+
+def modcrop(img_in, scale):
+ # img_in: Numpy, HWC or HW
+ img = np.copy(img_in)
+ if img.ndim == 2:
+ H, W = img.shape
+ H_r, W_r = H % scale, W % scale
+ img = img[:H - H_r, :W - W_r]
+ elif img.ndim == 3:
+ H, W, C = img.shape
+ H_r, W_r = H % scale, W % scale
+ img = img[:H - H_r, :W - W_r, :]
+ else:
+ raise ValueError('Wrong img ndim: [{:d}].'.format(img.ndim))
+ return img
+
+
+def shave(img_in, border=0):
+ # img_in: Numpy, HWC or HW
+ img = np.copy(img_in)
+ h, w = img.shape[:2]
+ img = img[border:h-border, border:w-border]
+ return img
+
+
+'''
+# --------------------------------------------
+# image processing process on numpy image
+# channel_convert(in_c, tar_type, img_list):
+# rgb2ycbcr(img, only_y=True):
+# bgr2ycbcr(img, only_y=True):
+# ycbcr2rgb(img):
+# --------------------------------------------
+'''
+
+
+def rgb2ycbcr(img, only_y=True):
+ '''same as matlab rgb2ycbcr
+ only_y: only return Y channel
+ Input:
+ uint8, [0, 255]
+ float, [0, 1]
+ '''
+ in_img_type = img.dtype
+ img.astype(np.float32)
+ if in_img_type != np.uint8:
+ img *= 255.
+ # convert
+ if only_y:
+ rlt = np.dot(img, [65.481, 128.553, 24.966]) / 255.0 + 16.0
+ else:
+ rlt = np.matmul(img, [[65.481, -37.797, 112.0], [128.553, -74.203, -93.786],
+ [24.966, 112.0, -18.214]]) / 255.0 + [16, 128, 128]
+ if in_img_type == np.uint8:
+ rlt = rlt.round()
+ else:
+ rlt /= 255.
+ return rlt.astype(in_img_type)
+
+
+def ycbcr2rgb(img):
+ '''same as matlab ycbcr2rgb
+ Input:
+ uint8, [0, 255]
+ float, [0, 1]
+ '''
+ in_img_type = img.dtype
+ img.astype(np.float32)
+ if in_img_type != np.uint8:
+ img *= 255.
+ # convert
+ rlt = np.matmul(img, [[0.00456621, 0.00456621, 0.00456621], [0, -0.00153632, 0.00791071],
+ [0.00625893, -0.00318811, 0]]) * 255.0 + [-222.921, 135.576, -276.836]
+ if in_img_type == np.uint8:
+ rlt = rlt.round()
+ else:
+ rlt /= 255.
+ return rlt.astype(in_img_type)
+
+
+def bgr2ycbcr(img, only_y=True):
+ '''bgr version of rgb2ycbcr
+ only_y: only return Y channel
+ Input:
+ uint8, [0, 255]
+ float, [0, 1]
+ '''
+ in_img_type = img.dtype
+ img.astype(np.float32)
+ if in_img_type != np.uint8:
+ img *= 255.
+ # convert
+ if only_y:
+ rlt = np.dot(img, [24.966, 128.553, 65.481]) / 255.0 + 16.0
+ else:
+ rlt = np.matmul(img, [[24.966, 112.0, -18.214], [128.553, -74.203, -93.786],
+ [65.481, -37.797, 112.0]]) / 255.0 + [16, 128, 128]
+ if in_img_type == np.uint8:
+ rlt = rlt.round()
+ else:
+ rlt /= 255.
+ return rlt.astype(in_img_type)
+
+
+def channel_convert(in_c, tar_type, img_list):
+ # conversion among BGR, gray and y
+ if in_c == 3 and tar_type == 'gray': # BGR to gray
+ gray_list = [cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) for img in img_list]
+ return [np.expand_dims(img, axis=2) for img in gray_list]
+ elif in_c == 3 and tar_type == 'y': # BGR to y
+ y_list = [bgr2ycbcr(img, only_y=True) for img in img_list]
+ return [np.expand_dims(img, axis=2) for img in y_list]
+ elif in_c == 1 and tar_type == 'RGB': # gray/y to BGR
+ return [cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) for img in img_list]
+ else:
+ return img_list
+
+
+'''
+# --------------------------------------------
+# metric, PSNR and SSIM
+# --------------------------------------------
+'''
+
+
+# --------------------------------------------
+# PSNR
+# --------------------------------------------
+def calculate_psnr(img1, img2, border=0):
+ # img1 and img2 have range [0, 255]
+ #img1 = img1.squeeze()
+ #img2 = img2.squeeze()
+ if not img1.shape == img2.shape:
+ raise ValueError('Input images must have the same dimensions.')
+ h, w = img1.shape[:2]
+ img1 = img1[border:h-border, border:w-border]
+ img2 = img2[border:h-border, border:w-border]
+
+ img1 = img1.astype(np.float64)
+ img2 = img2.astype(np.float64)
+ mse = np.mean((img1 - img2)**2)
+ if mse == 0:
+ return float('inf')
+ return 20 * math.log10(255.0 / math.sqrt(mse))
+
+
+# --------------------------------------------
+# SSIM
+# --------------------------------------------
+def calculate_ssim(img1, img2, border=0):
+ '''calculate SSIM
+ the same outputs as MATLAB's
+ img1, img2: [0, 255]
+ '''
+ #img1 = img1.squeeze()
+ #img2 = img2.squeeze()
+ if not img1.shape == img2.shape:
+ raise ValueError('Input images must have the same dimensions.')
+ h, w = img1.shape[:2]
+ img1 = img1[border:h-border, border:w-border]
+ img2 = img2[border:h-border, border:w-border]
+
+ if img1.ndim == 2:
+ return ssim(img1, img2)
+ elif img1.ndim == 3:
+ if img1.shape[2] == 3:
+ ssims = []
+ for i in range(3):
+ ssims.append(ssim(img1[:,:,i], img2[:,:,i]))
+ return np.array(ssims).mean()
+ elif img1.shape[2] == 1:
+ return ssim(np.squeeze(img1), np.squeeze(img2))
+ else:
+ raise ValueError('Wrong input image dimensions.')
+
+
+def ssim(img1, img2):
+ C1 = (0.01 * 255)**2
+ C2 = (0.03 * 255)**2
+
+ img1 = img1.astype(np.float64)
+ img2 = img2.astype(np.float64)
+ kernel = cv2.getGaussianKernel(11, 1.5)
+ window = np.outer(kernel, kernel.transpose())
+
+ mu1 = cv2.filter2D(img1, -1, window)[5:-5, 5:-5] # valid
+ mu2 = cv2.filter2D(img2, -1, window)[5:-5, 5:-5]
+ mu1_sq = mu1**2
+ mu2_sq = mu2**2
+ mu1_mu2 = mu1 * mu2
+ sigma1_sq = cv2.filter2D(img1**2, -1, window)[5:-5, 5:-5] - mu1_sq
+ sigma2_sq = cv2.filter2D(img2**2, -1, window)[5:-5, 5:-5] - mu2_sq
+ sigma12 = cv2.filter2D(img1 * img2, -1, window)[5:-5, 5:-5] - mu1_mu2
+
+ ssim_map = ((2 * mu1_mu2 + C1) * (2 * sigma12 + C2)) / ((mu1_sq + mu2_sq + C1) *
+ (sigma1_sq + sigma2_sq + C2))
+ return ssim_map.mean()
+
+
+'''
+# --------------------------------------------
+# matlab's bicubic imresize (numpy and torch) [0, 1]
+# --------------------------------------------
+'''
+
+
+# matlab 'imresize' function, now only support 'bicubic'
+def cubic(x):
+ absx = torch.abs(x)
+ absx2 = absx**2
+ absx3 = absx**3
+ return (1.5*absx3 - 2.5*absx2 + 1) * ((absx <= 1).type_as(absx)) + \
+ (-0.5*absx3 + 2.5*absx2 - 4*absx + 2) * (((absx > 1)*(absx <= 2)).type_as(absx))
+
+
+def calculate_weights_indices(in_length, out_length, scale, kernel, kernel_width, antialiasing):
+ if (scale < 1) and (antialiasing):
+ # Use a modified kernel to simultaneously interpolate and antialias- larger kernel width
+ kernel_width = kernel_width / scale
+
+ # Output-space coordinates
+ x = torch.linspace(1, out_length, out_length)
+
+ # Input-space coordinates. Calculate the inverse mapping such that 0.5
+ # in output space maps to 0.5 in input space, and 0.5+scale in output
+ # space maps to 1.5 in input space.
+ u = x / scale + 0.5 * (1 - 1 / scale)
+
+ # What is the left-most pixel that can be involved in the computation?
+ left = torch.floor(u - kernel_width / 2)
+
+ # What is the maximum number of pixels that can be involved in the
+ # computation? Note: it's OK to use an extra pixel here; if the
+ # corresponding weights are all zero, it will be eliminated at the end
+ # of this function.
+ P = math.ceil(kernel_width) + 2
+
+ # The indices of the input pixels involved in computing the k-th output
+ # pixel are in row k of the indices matrix.
+ indices = left.view(out_length, 1).expand(out_length, P) + torch.linspace(0, P - 1, P).view(
+ 1, P).expand(out_length, P)
+
+ # The weights used to compute the k-th output pixel are in row k of the
+ # weights matrix.
+ distance_to_center = u.view(out_length, 1).expand(out_length, P) - indices
+ # apply cubic kernel
+ if (scale < 1) and (antialiasing):
+ weights = scale * cubic(distance_to_center * scale)
+ else:
+ weights = cubic(distance_to_center)
+ # Normalize the weights matrix so that each row sums to 1.
+ weights_sum = torch.sum(weights, 1).view(out_length, 1)
+ weights = weights / weights_sum.expand(out_length, P)
+
+ # If a column in weights is all zero, get rid of it. only consider the first and last column.
+ weights_zero_tmp = torch.sum((weights == 0), 0)
+ if not math.isclose(weights_zero_tmp[0], 0, rel_tol=1e-6):
+ indices = indices.narrow(1, 1, P - 2)
+ weights = weights.narrow(1, 1, P - 2)
+ if not math.isclose(weights_zero_tmp[-1], 0, rel_tol=1e-6):
+ indices = indices.narrow(1, 0, P - 2)
+ weights = weights.narrow(1, 0, P - 2)
+ weights = weights.contiguous()
+ indices = indices.contiguous()
+ sym_len_s = -indices.min() + 1
+ sym_len_e = indices.max() - in_length
+ indices = indices + sym_len_s - 1
+ return weights, indices, int(sym_len_s), int(sym_len_e)
+
+
+# --------------------------------------------
+# imresize for tensor image [0, 1]
+# --------------------------------------------
+def imresize(img, scale, antialiasing=True):
+ # Now the scale should be the same for H and W
+ # input: img: pytorch tensor, CHW or HW [0,1]
+ # output: CHW or HW [0,1] w/o round
+ need_squeeze = True if img.dim() == 2 else False
+ if need_squeeze:
+ img.unsqueeze_(0)
+ in_C, in_H, in_W = img.size()
+ out_C, out_H, out_W = in_C, math.ceil(in_H * scale), math.ceil(in_W * scale)
+ kernel_width = 4
+ kernel = 'cubic'
+
+ # Return the desired dimension order for performing the resize. The
+ # strategy is to perform the resize first along the dimension with the
+ # smallest scale factor.
+ # Now we do not support this.
+
+ # get weights and indices
+ weights_H, indices_H, sym_len_Hs, sym_len_He = calculate_weights_indices(
+ in_H, out_H, scale, kernel, kernel_width, antialiasing)
+ weights_W, indices_W, sym_len_Ws, sym_len_We = calculate_weights_indices(
+ in_W, out_W, scale, kernel, kernel_width, antialiasing)
+ # process H dimension
+ # symmetric copying
+ img_aug = torch.FloatTensor(in_C, in_H + sym_len_Hs + sym_len_He, in_W)
+ img_aug.narrow(1, sym_len_Hs, in_H).copy_(img)
+
+ sym_patch = img[:, :sym_len_Hs, :]
+ inv_idx = torch.arange(sym_patch.size(1) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(1, inv_idx)
+ img_aug.narrow(1, 0, sym_len_Hs).copy_(sym_patch_inv)
+
+ sym_patch = img[:, -sym_len_He:, :]
+ inv_idx = torch.arange(sym_patch.size(1) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(1, inv_idx)
+ img_aug.narrow(1, sym_len_Hs + in_H, sym_len_He).copy_(sym_patch_inv)
+
+ out_1 = torch.FloatTensor(in_C, out_H, in_W)
+ kernel_width = weights_H.size(1)
+ for i in range(out_H):
+ idx = int(indices_H[i][0])
+ for j in range(out_C):
+ out_1[j, i, :] = img_aug[j, idx:idx + kernel_width, :].transpose(0, 1).mv(weights_H[i])
+
+ # process W dimension
+ # symmetric copying
+ out_1_aug = torch.FloatTensor(in_C, out_H, in_W + sym_len_Ws + sym_len_We)
+ out_1_aug.narrow(2, sym_len_Ws, in_W).copy_(out_1)
+
+ sym_patch = out_1[:, :, :sym_len_Ws]
+ inv_idx = torch.arange(sym_patch.size(2) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(2, inv_idx)
+ out_1_aug.narrow(2, 0, sym_len_Ws).copy_(sym_patch_inv)
+
+ sym_patch = out_1[:, :, -sym_len_We:]
+ inv_idx = torch.arange(sym_patch.size(2) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(2, inv_idx)
+ out_1_aug.narrow(2, sym_len_Ws + in_W, sym_len_We).copy_(sym_patch_inv)
+
+ out_2 = torch.FloatTensor(in_C, out_H, out_W)
+ kernel_width = weights_W.size(1)
+ for i in range(out_W):
+ idx = int(indices_W[i][0])
+ for j in range(out_C):
+ out_2[j, :, i] = out_1_aug[j, :, idx:idx + kernel_width].mv(weights_W[i])
+ if need_squeeze:
+ out_2.squeeze_()
+ return out_2
+
+
+# --------------------------------------------
+# imresize for numpy image [0, 1]
+# --------------------------------------------
+def imresize_np(img, scale, antialiasing=True):
+ # Now the scale should be the same for H and W
+ # input: img: Numpy, HWC or HW [0,1]
+ # output: HWC or HW [0,1] w/o round
+ img = torch.from_numpy(img)
+ need_squeeze = True if img.dim() == 2 else False
+ if need_squeeze:
+ img.unsqueeze_(2)
+
+ in_H, in_W, in_C = img.size()
+ out_C, out_H, out_W = in_C, math.ceil(in_H * scale), math.ceil(in_W * scale)
+ kernel_width = 4
+ kernel = 'cubic'
+
+ # Return the desired dimension order for performing the resize. The
+ # strategy is to perform the resize first along the dimension with the
+ # smallest scale factor.
+ # Now we do not support this.
+
+ # get weights and indices
+ weights_H, indices_H, sym_len_Hs, sym_len_He = calculate_weights_indices(
+ in_H, out_H, scale, kernel, kernel_width, antialiasing)
+ weights_W, indices_W, sym_len_Ws, sym_len_We = calculate_weights_indices(
+ in_W, out_W, scale, kernel, kernel_width, antialiasing)
+ # process H dimension
+ # symmetric copying
+ img_aug = torch.FloatTensor(in_H + sym_len_Hs + sym_len_He, in_W, in_C)
+ img_aug.narrow(0, sym_len_Hs, in_H).copy_(img)
+
+ sym_patch = img[:sym_len_Hs, :, :]
+ inv_idx = torch.arange(sym_patch.size(0) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(0, inv_idx)
+ img_aug.narrow(0, 0, sym_len_Hs).copy_(sym_patch_inv)
+
+ sym_patch = img[-sym_len_He:, :, :]
+ inv_idx = torch.arange(sym_patch.size(0) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(0, inv_idx)
+ img_aug.narrow(0, sym_len_Hs + in_H, sym_len_He).copy_(sym_patch_inv)
+
+ out_1 = torch.FloatTensor(out_H, in_W, in_C)
+ kernel_width = weights_H.size(1)
+ for i in range(out_H):
+ idx = int(indices_H[i][0])
+ for j in range(out_C):
+ out_1[i, :, j] = img_aug[idx:idx + kernel_width, :, j].transpose(0, 1).mv(weights_H[i])
+
+ # process W dimension
+ # symmetric copying
+ out_1_aug = torch.FloatTensor(out_H, in_W + sym_len_Ws + sym_len_We, in_C)
+ out_1_aug.narrow(1, sym_len_Ws, in_W).copy_(out_1)
+
+ sym_patch = out_1[:, :sym_len_Ws, :]
+ inv_idx = torch.arange(sym_patch.size(1) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(1, inv_idx)
+ out_1_aug.narrow(1, 0, sym_len_Ws).copy_(sym_patch_inv)
+
+ sym_patch = out_1[:, -sym_len_We:, :]
+ inv_idx = torch.arange(sym_patch.size(1) - 1, -1, -1).long()
+ sym_patch_inv = sym_patch.index_select(1, inv_idx)
+ out_1_aug.narrow(1, sym_len_Ws + in_W, sym_len_We).copy_(sym_patch_inv)
+
+ out_2 = torch.FloatTensor(out_H, out_W, in_C)
+ kernel_width = weights_W.size(1)
+ for i in range(out_W):
+ idx = int(indices_W[i][0])
+ for j in range(out_C):
+ out_2[:, i, j] = out_1_aug[:, idx:idx + kernel_width, j].mv(weights_W[i])
+ if need_squeeze:
+ out_2.squeeze_()
+
+ return out_2.numpy()
+
+
+if __name__ == '__main__':
+ print('---')
+# img = imread_uint('test.bmp', 3)
+# img = uint2single(img)
+# img_bicubic = imresize_np(img, 1/4)
\ No newline at end of file
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/losses/__init__.py b/OCR_earsing/latent_diffusion/ldm/modules/losses/__init__.py
new file mode 100644
index 0000000..876d7c5
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/losses/__init__.py
@@ -0,0 +1 @@
+from ldm.modules.losses.contperceptual import LPIPSWithDiscriminator
\ No newline at end of file
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/losses/contperceptual.py b/OCR_earsing/latent_diffusion/ldm/modules/losses/contperceptual.py
new file mode 100644
index 0000000..672c1e3
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/losses/contperceptual.py
@@ -0,0 +1,111 @@
+import torch
+import torch.nn as nn
+
+from taming.modules.losses.vqperceptual import * # TODO: taming dependency yes/no?
+
+
+class LPIPSWithDiscriminator(nn.Module):
+ def __init__(self, disc_start, logvar_init=0.0, kl_weight=1.0, pixelloss_weight=1.0,
+ disc_num_layers=3, disc_in_channels=3, disc_factor=1.0, disc_weight=1.0,
+ perceptual_weight=1.0, use_actnorm=False, disc_conditional=False,
+ disc_loss="hinge"):
+
+ super().__init__()
+ assert disc_loss in ["hinge", "vanilla"]
+ self.kl_weight = kl_weight
+ self.pixel_weight = pixelloss_weight
+ self.perceptual_loss = LPIPS().eval()
+ self.perceptual_weight = perceptual_weight
+ # output log variance
+ self.logvar = nn.Parameter(torch.ones(size=()) * logvar_init)
+
+ self.discriminator = NLayerDiscriminator(input_nc=disc_in_channels,
+ n_layers=disc_num_layers,
+ use_actnorm=use_actnorm
+ ).apply(weights_init)
+ self.discriminator_iter_start = disc_start
+ self.disc_loss = hinge_d_loss if disc_loss == "hinge" else vanilla_d_loss
+ self.disc_factor = disc_factor
+ self.discriminator_weight = disc_weight
+ self.disc_conditional = disc_conditional
+
+ def calculate_adaptive_weight(self, nll_loss, g_loss, last_layer=None):
+ if last_layer is not None:
+ nll_grads = torch.autograd.grad(nll_loss, last_layer, retain_graph=True)[0]
+ g_grads = torch.autograd.grad(g_loss, last_layer, retain_graph=True)[0]
+ else:
+ nll_grads = torch.autograd.grad(nll_loss, self.last_layer[0], retain_graph=True)[0]
+ g_grads = torch.autograd.grad(g_loss, self.last_layer[0], retain_graph=True)[0]
+
+ d_weight = torch.norm(nll_grads) / (torch.norm(g_grads) + 1e-4)
+ d_weight = torch.clamp(d_weight, 0.0, 1e4).detach()
+ d_weight = d_weight * self.discriminator_weight
+ return d_weight
+
+ def forward(self, inputs, reconstructions, posteriors, optimizer_idx,
+ global_step, last_layer=None, cond=None, split="train",
+ weights=None):
+ rec_loss = torch.abs(inputs.contiguous() - reconstructions.contiguous())
+ if self.perceptual_weight > 0:
+ p_loss = self.perceptual_loss(inputs.contiguous(), reconstructions.contiguous())
+ rec_loss = rec_loss + self.perceptual_weight * p_loss
+
+ nll_loss = rec_loss / torch.exp(self.logvar) + self.logvar
+ weighted_nll_loss = nll_loss
+ if weights is not None:
+ weighted_nll_loss = weights*nll_loss
+ weighted_nll_loss = torch.sum(weighted_nll_loss) / weighted_nll_loss.shape[0]
+ nll_loss = torch.sum(nll_loss) / nll_loss.shape[0]
+ kl_loss = posteriors.kl()
+ kl_loss = torch.sum(kl_loss) / kl_loss.shape[0]
+
+ # now the GAN part
+ if optimizer_idx == 0:
+ # generator update
+ if cond is None:
+ assert not self.disc_conditional
+ logits_fake = self.discriminator(reconstructions.contiguous())
+ else:
+ assert self.disc_conditional
+ logits_fake = self.discriminator(torch.cat((reconstructions.contiguous(), cond), dim=1))
+ g_loss = -torch.mean(logits_fake)
+
+ if self.disc_factor > 0.0:
+ try:
+ d_weight = self.calculate_adaptive_weight(nll_loss, g_loss, last_layer=last_layer)
+ except RuntimeError:
+ assert not self.training
+ d_weight = torch.tensor(0.0)
+ else:
+ d_weight = torch.tensor(0.0)
+
+ disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+ loss = weighted_nll_loss + self.kl_weight * kl_loss + d_weight * disc_factor * g_loss
+
+ log = {"{}/total_loss".format(split): loss.clone().detach().mean(), "{}/logvar".format(split): self.logvar.detach(),
+ "{}/kl_loss".format(split): kl_loss.detach().mean(), "{}/nll_loss".format(split): nll_loss.detach().mean(),
+ "{}/rec_loss".format(split): rec_loss.detach().mean(),
+ "{}/d_weight".format(split): d_weight.detach(),
+ "{}/disc_factor".format(split): torch.tensor(disc_factor),
+ "{}/g_loss".format(split): g_loss.detach().mean(),
+ }
+ return loss, log
+
+ if optimizer_idx == 1:
+ # second pass for discriminator update
+ if cond is None:
+ logits_real = self.discriminator(inputs.contiguous().detach())
+ logits_fake = self.discriminator(reconstructions.contiguous().detach())
+ else:
+ logits_real = self.discriminator(torch.cat((inputs.contiguous().detach(), cond), dim=1))
+ logits_fake = self.discriminator(torch.cat((reconstructions.contiguous().detach(), cond), dim=1))
+
+ disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+ d_loss = disc_factor * self.disc_loss(logits_real, logits_fake)
+
+ log = {"{}/disc_loss".format(split): d_loss.clone().detach().mean(),
+ "{}/logits_real".format(split): logits_real.detach().mean(),
+ "{}/logits_fake".format(split): logits_fake.detach().mean()
+ }
+ return d_loss, log
+
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/losses/vqperceptual.py b/OCR_earsing/latent_diffusion/ldm/modules/losses/vqperceptual.py
new file mode 100644
index 0000000..f699817
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/losses/vqperceptual.py
@@ -0,0 +1,167 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+from einops import repeat
+
+from taming.modules.discriminator.model import NLayerDiscriminator, weights_init
+from taming.modules.losses.lpips import LPIPS
+from taming.modules.losses.vqperceptual import hinge_d_loss, vanilla_d_loss
+
+
+def hinge_d_loss_with_exemplar_weights(logits_real, logits_fake, weights):
+ assert weights.shape[0] == logits_real.shape[0] == logits_fake.shape[0]
+ loss_real = torch.mean(F.relu(1. - logits_real), dim=[1,2,3])
+ loss_fake = torch.mean(F.relu(1. + logits_fake), dim=[1,2,3])
+ loss_real = (weights * loss_real).sum() / weights.sum()
+ loss_fake = (weights * loss_fake).sum() / weights.sum()
+ d_loss = 0.5 * (loss_real + loss_fake)
+ return d_loss
+
+def adopt_weight(weight, global_step, threshold=0, value=0.):
+ if global_step < threshold:
+ weight = value
+ return weight
+
+
+def measure_perplexity(predicted_indices, n_embed):
+ # src: https://github.com/karpathy/deep-vector-quantization/blob/main/model.py
+ # eval cluster perplexity. when perplexity == num_embeddings then all clusters are used exactly equally
+ encodings = F.one_hot(predicted_indices, n_embed).float().reshape(-1, n_embed)
+ avg_probs = encodings.mean(0)
+ perplexity = (-(avg_probs * torch.log(avg_probs + 1e-10)).sum()).exp()
+ cluster_use = torch.sum(avg_probs > 0)
+ return perplexity, cluster_use
+
+def l1(x, y):
+ return torch.abs(x-y)
+
+
+def l2(x, y):
+ return torch.pow((x-y), 2)
+
+
+class VQLPIPSWithDiscriminator(nn.Module):
+ def __init__(self, disc_start, codebook_weight=1.0, pixelloss_weight=1.0,
+ disc_num_layers=3, disc_in_channels=3, disc_factor=1.0, disc_weight=1.0,
+ perceptual_weight=1.0, use_actnorm=False, disc_conditional=False,
+ disc_ndf=64, disc_loss="hinge", n_classes=None, perceptual_loss="lpips",
+ pixel_loss="l1"):
+ super().__init__()
+ assert disc_loss in ["hinge", "vanilla"]
+ assert perceptual_loss in ["lpips", "clips", "dists"]
+ assert pixel_loss in ["l1", "l2"]
+ self.codebook_weight = codebook_weight
+ self.pixel_weight = pixelloss_weight
+ if perceptual_loss == "lpips":
+ print(f"{self.__class__.__name__}: Running with LPIPS.")
+ self.perceptual_loss = LPIPS().eval()
+ else:
+ raise ValueError(f"Unknown perceptual loss: >> {perceptual_loss} <<")
+ self.perceptual_weight = perceptual_weight
+
+ if pixel_loss == "l1":
+ self.pixel_loss = l1
+ else:
+ self.pixel_loss = l2
+
+ self.discriminator = NLayerDiscriminator(input_nc=disc_in_channels,
+ n_layers=disc_num_layers,
+ use_actnorm=use_actnorm,
+ ndf=disc_ndf
+ ).apply(weights_init)
+ self.discriminator_iter_start = disc_start
+ if disc_loss == "hinge":
+ self.disc_loss = hinge_d_loss
+ elif disc_loss == "vanilla":
+ self.disc_loss = vanilla_d_loss
+ else:
+ raise ValueError(f"Unknown GAN loss '{disc_loss}'.")
+ print(f"VQLPIPSWithDiscriminator running with {disc_loss} loss.")
+ self.disc_factor = disc_factor
+ self.discriminator_weight = disc_weight
+ self.disc_conditional = disc_conditional
+ self.n_classes = n_classes
+
+ def calculate_adaptive_weight(self, nll_loss, g_loss, last_layer=None):
+ if last_layer is not None:
+ nll_grads = torch.autograd.grad(nll_loss, last_layer, retain_graph=True)[0]
+ g_grads = torch.autograd.grad(g_loss, last_layer, retain_graph=True)[0]
+ else:
+ nll_grads = torch.autograd.grad(nll_loss, self.last_layer[0], retain_graph=True)[0]
+ g_grads = torch.autograd.grad(g_loss, self.last_layer[0], retain_graph=True)[0]
+
+ d_weight = torch.norm(nll_grads) / (torch.norm(g_grads) + 1e-4)
+ d_weight = torch.clamp(d_weight, 0.0, 1e4).detach()
+ d_weight = d_weight * self.discriminator_weight
+ return d_weight
+
+ def forward(self, codebook_loss, inputs, reconstructions, optimizer_idx,
+ global_step, last_layer=None, cond=None, split="train", predicted_indices=None):
+ if not exists(codebook_loss):
+ codebook_loss = torch.tensor([0.]).to(inputs.device)
+ #rec_loss = torch.abs(inputs.contiguous() - reconstructions.contiguous())
+ rec_loss = self.pixel_loss(inputs.contiguous(), reconstructions.contiguous())
+ if self.perceptual_weight > 0:
+ p_loss = self.perceptual_loss(inputs.contiguous(), reconstructions.contiguous())
+ rec_loss = rec_loss + self.perceptual_weight * p_loss
+ else:
+ p_loss = torch.tensor([0.0])
+
+ nll_loss = rec_loss
+ #nll_loss = torch.sum(nll_loss) / nll_loss.shape[0]
+ nll_loss = torch.mean(nll_loss)
+
+ # now the GAN part
+ if optimizer_idx == 0:
+ # generator update
+ if cond is None:
+ assert not self.disc_conditional
+ logits_fake = self.discriminator(reconstructions.contiguous())
+ else:
+ assert self.disc_conditional
+ logits_fake = self.discriminator(torch.cat((reconstructions.contiguous(), cond), dim=1))
+ g_loss = -torch.mean(logits_fake)
+
+ try:
+ d_weight = self.calculate_adaptive_weight(nll_loss, g_loss, last_layer=last_layer)
+ except RuntimeError:
+ assert not self.training
+ d_weight = torch.tensor(0.0)
+
+ disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+ loss = nll_loss + d_weight * disc_factor * g_loss + self.codebook_weight * codebook_loss.mean()
+
+ log = {"{}/total_loss".format(split): loss.clone().detach().mean(),
+ "{}/quant_loss".format(split): codebook_loss.detach().mean(),
+ "{}/nll_loss".format(split): nll_loss.detach().mean(),
+ "{}/rec_loss".format(split): rec_loss.detach().mean(),
+ "{}/p_loss".format(split): p_loss.detach().mean(),
+ "{}/d_weight".format(split): d_weight.detach(),
+ "{}/disc_factor".format(split): torch.tensor(disc_factor),
+ "{}/g_loss".format(split): g_loss.detach().mean(),
+ }
+ if predicted_indices is not None:
+ assert self.n_classes is not None
+ with torch.no_grad():
+ perplexity, cluster_usage = measure_perplexity(predicted_indices, self.n_classes)
+ log[f"{split}/perplexity"] = perplexity
+ log[f"{split}/cluster_usage"] = cluster_usage
+ return loss, log
+
+ if optimizer_idx == 1:
+ # second pass for discriminator update
+ if cond is None:
+ logits_real = self.discriminator(inputs.contiguous().detach())
+ logits_fake = self.discriminator(reconstructions.contiguous().detach())
+ else:
+ logits_real = self.discriminator(torch.cat((inputs.contiguous().detach(), cond), dim=1))
+ logits_fake = self.discriminator(torch.cat((reconstructions.contiguous().detach(), cond), dim=1))
+
+ disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+ d_loss = disc_factor * self.disc_loss(logits_real, logits_fake)
+
+ log = {"{}/disc_loss".format(split): d_loss.clone().detach().mean(),
+ "{}/logits_real".format(split): logits_real.detach().mean(),
+ "{}/logits_fake".format(split): logits_fake.detach().mean()
+ }
+ return d_loss, log
diff --git a/OCR_earsing/latent_diffusion/ldm/modules/x_transformer.py b/OCR_earsing/latent_diffusion/ldm/modules/x_transformer.py
new file mode 100644
index 0000000..5fc15bf
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/modules/x_transformer.py
@@ -0,0 +1,641 @@
+"""shout-out to https://github.com/lucidrains/x-transformers/tree/main/x_transformers"""
+import torch
+from torch import nn, einsum
+import torch.nn.functional as F
+from functools import partial
+from inspect import isfunction
+from collections import namedtuple
+from einops import rearrange, repeat, reduce
+
+# constants
+
+DEFAULT_DIM_HEAD = 64
+
+Intermediates = namedtuple('Intermediates', [
+ 'pre_softmax_attn',
+ 'post_softmax_attn'
+])
+
+LayerIntermediates = namedtuple('Intermediates', [
+ 'hiddens',
+ 'attn_intermediates'
+])
+
+
+class AbsolutePositionalEmbedding(nn.Module):
+ def __init__(self, dim, max_seq_len):
+ super().__init__()
+ self.emb = nn.Embedding(max_seq_len, dim)
+ self.init_()
+
+ def init_(self):
+ nn.init.normal_(self.emb.weight, std=0.02)
+
+ def forward(self, x):
+ n = torch.arange(x.shape[1], device=x.device)
+ return self.emb(n)[None, :, :]
+
+
+class FixedPositionalEmbedding(nn.Module):
+ def __init__(self, dim):
+ super().__init__()
+ inv_freq = 1. / (10000 ** (torch.arange(0, dim, 2).float() / dim))
+ self.register_buffer('inv_freq', inv_freq)
+
+ def forward(self, x, seq_dim=1, offset=0):
+ t = torch.arange(x.shape[seq_dim], device=x.device).type_as(self.inv_freq) + offset
+ sinusoid_inp = torch.einsum('i , j -> i j', t, self.inv_freq)
+ emb = torch.cat((sinusoid_inp.sin(), sinusoid_inp.cos()), dim=-1)
+ return emb[None, :, :]
+
+
+# helpers
+
+def exists(val):
+ return val is not None
+
+
+def default(val, d):
+ if exists(val):
+ return val
+ return d() if isfunction(d) else d
+
+
+def always(val):
+ def inner(*args, **kwargs):
+ return val
+ return inner
+
+
+def not_equals(val):
+ def inner(x):
+ return x != val
+ return inner
+
+
+def equals(val):
+ def inner(x):
+ return x == val
+ return inner
+
+
+def max_neg_value(tensor):
+ return -torch.finfo(tensor.dtype).max
+
+
+# keyword argument helpers
+
+def pick_and_pop(keys, d):
+ values = list(map(lambda key: d.pop(key), keys))
+ return dict(zip(keys, values))
+
+
+def group_dict_by_key(cond, d):
+ return_val = [dict(), dict()]
+ for key in d.keys():
+ match = bool(cond(key))
+ ind = int(not match)
+ return_val[ind][key] = d[key]
+ return (*return_val,)
+
+
+def string_begins_with(prefix, str):
+ return str.startswith(prefix)
+
+
+def group_by_key_prefix(prefix, d):
+ return group_dict_by_key(partial(string_begins_with, prefix), d)
+
+
+def groupby_prefix_and_trim(prefix, d):
+ kwargs_with_prefix, kwargs = group_dict_by_key(partial(string_begins_with, prefix), d)
+ kwargs_without_prefix = dict(map(lambda x: (x[0][len(prefix):], x[1]), tuple(kwargs_with_prefix.items())))
+ return kwargs_without_prefix, kwargs
+
+
+# classes
+class Scale(nn.Module):
+ def __init__(self, value, fn):
+ super().__init__()
+ self.value = value
+ self.fn = fn
+
+ def forward(self, x, **kwargs):
+ x, *rest = self.fn(x, **kwargs)
+ return (x * self.value, *rest)
+
+
+class Rezero(nn.Module):
+ def __init__(self, fn):
+ super().__init__()
+ self.fn = fn
+ self.g = nn.Parameter(torch.zeros(1))
+
+ def forward(self, x, **kwargs):
+ x, *rest = self.fn(x, **kwargs)
+ return (x * self.g, *rest)
+
+
+class ScaleNorm(nn.Module):
+ def __init__(self, dim, eps=1e-5):
+ super().__init__()
+ self.scale = dim ** -0.5
+ self.eps = eps
+ self.g = nn.Parameter(torch.ones(1))
+
+ def forward(self, x):
+ norm = torch.norm(x, dim=-1, keepdim=True) * self.scale
+ return x / norm.clamp(min=self.eps) * self.g
+
+
+class RMSNorm(nn.Module):
+ def __init__(self, dim, eps=1e-8):
+ super().__init__()
+ self.scale = dim ** -0.5
+ self.eps = eps
+ self.g = nn.Parameter(torch.ones(dim))
+
+ def forward(self, x):
+ norm = torch.norm(x, dim=-1, keepdim=True) * self.scale
+ return x / norm.clamp(min=self.eps) * self.g
+
+
+class Residual(nn.Module):
+ def forward(self, x, residual):
+ return x + residual
+
+
+class GRUGating(nn.Module):
+ def __init__(self, dim):
+ super().__init__()
+ self.gru = nn.GRUCell(dim, dim)
+
+ def forward(self, x, residual):
+ gated_output = self.gru(
+ rearrange(x, 'b n d -> (b n) d'),
+ rearrange(residual, 'b n d -> (b n) d')
+ )
+
+ return gated_output.reshape_as(x)
+
+
+# feedforward
+
+class GEGLU(nn.Module):
+ def __init__(self, dim_in, dim_out):
+ super().__init__()
+ self.proj = nn.Linear(dim_in, dim_out * 2)
+
+ def forward(self, x):
+ x, gate = self.proj(x).chunk(2, dim=-1)
+ return x * F.gelu(gate)
+
+
+class FeedForward(nn.Module):
+ def __init__(self, dim, dim_out=None, mult=4, glu=False, dropout=0.):
+ super().__init__()
+ inner_dim = int(dim * mult)
+ dim_out = default(dim_out, dim)
+ project_in = nn.Sequential(
+ nn.Linear(dim, inner_dim),
+ nn.GELU()
+ ) if not glu else GEGLU(dim, inner_dim)
+
+ self.net = nn.Sequential(
+ project_in,
+ nn.Dropout(dropout),
+ nn.Linear(inner_dim, dim_out)
+ )
+
+ def forward(self, x):
+ return self.net(x)
+
+
+# attention.
+class Attention(nn.Module):
+ def __init__(
+ self,
+ dim,
+ dim_head=DEFAULT_DIM_HEAD,
+ heads=8,
+ causal=False,
+ mask=None,
+ talking_heads=False,
+ sparse_topk=None,
+ use_entmax15=False,
+ num_mem_kv=0,
+ dropout=0.,
+ on_attn=False
+ ):
+ super().__init__()
+ if use_entmax15:
+ raise NotImplementedError("Check out entmax activation instead of softmax activation!")
+ self.scale = dim_head ** -0.5
+ self.heads = heads
+ self.causal = causal
+ self.mask = mask
+
+ inner_dim = dim_head * heads
+
+ self.to_q = nn.Linear(dim, inner_dim, bias=False)
+ self.to_k = nn.Linear(dim, inner_dim, bias=False)
+ self.to_v = nn.Linear(dim, inner_dim, bias=False)
+ self.dropout = nn.Dropout(dropout)
+
+ # talking heads
+ self.talking_heads = talking_heads
+ if talking_heads:
+ self.pre_softmax_proj = nn.Parameter(torch.randn(heads, heads))
+ self.post_softmax_proj = nn.Parameter(torch.randn(heads, heads))
+
+ # explicit topk sparse attention
+ self.sparse_topk = sparse_topk
+
+ # entmax
+ #self.attn_fn = entmax15 if use_entmax15 else F.softmax
+ self.attn_fn = F.softmax
+
+ # add memory key / values
+ self.num_mem_kv = num_mem_kv
+ if num_mem_kv > 0:
+ self.mem_k = nn.Parameter(torch.randn(heads, num_mem_kv, dim_head))
+ self.mem_v = nn.Parameter(torch.randn(heads, num_mem_kv, dim_head))
+
+ # attention on attention
+ self.attn_on_attn = on_attn
+ self.to_out = nn.Sequential(nn.Linear(inner_dim, dim * 2), nn.GLU()) if on_attn else nn.Linear(inner_dim, dim)
+
+ def forward(
+ self,
+ x,
+ context=None,
+ mask=None,
+ context_mask=None,
+ rel_pos=None,
+ sinusoidal_emb=None,
+ prev_attn=None,
+ mem=None
+ ):
+ b, n, _, h, talking_heads, device = *x.shape, self.heads, self.talking_heads, x.device
+ kv_input = default(context, x)
+
+ q_input = x
+ k_input = kv_input
+ v_input = kv_input
+
+ if exists(mem):
+ k_input = torch.cat((mem, k_input), dim=-2)
+ v_input = torch.cat((mem, v_input), dim=-2)
+
+ if exists(sinusoidal_emb):
+ # in shortformer, the query would start at a position offset depending on the past cached memory
+ offset = k_input.shape[-2] - q_input.shape[-2]
+ q_input = q_input + sinusoidal_emb(q_input, offset=offset)
+ k_input = k_input + sinusoidal_emb(k_input)
+
+ q = self.to_q(q_input)
+ k = self.to_k(k_input)
+ v = self.to_v(v_input)
+
+ q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h=h), (q, k, v))
+
+ input_mask = None
+ if any(map(exists, (mask, context_mask))):
+ q_mask = default(mask, lambda: torch.ones((b, n), device=device).bool())
+ k_mask = q_mask if not exists(context) else context_mask
+ k_mask = default(k_mask, lambda: torch.ones((b, k.shape[-2]), device=device).bool())
+ q_mask = rearrange(q_mask, 'b i -> b () i ()')
+ k_mask = rearrange(k_mask, 'b j -> b () () j')
+ input_mask = q_mask * k_mask
+
+ if self.num_mem_kv > 0:
+ mem_k, mem_v = map(lambda t: repeat(t, 'h n d -> b h n d', b=b), (self.mem_k, self.mem_v))
+ k = torch.cat((mem_k, k), dim=-2)
+ v = torch.cat((mem_v, v), dim=-2)
+ if exists(input_mask):
+ input_mask = F.pad(input_mask, (self.num_mem_kv, 0), value=True)
+
+ dots = einsum('b h i d, b h j d -> b h i j', q, k) * self.scale
+ mask_value = max_neg_value(dots)
+
+ if exists(prev_attn):
+ dots = dots + prev_attn
+
+ pre_softmax_attn = dots
+
+ if talking_heads:
+ dots = einsum('b h i j, h k -> b k i j', dots, self.pre_softmax_proj).contiguous()
+
+ if exists(rel_pos):
+ dots = rel_pos(dots)
+
+ if exists(input_mask):
+ dots.masked_fill_(~input_mask, mask_value)
+ del input_mask
+
+ if self.causal:
+ i, j = dots.shape[-2:]
+ r = torch.arange(i, device=device)
+ mask = rearrange(r, 'i -> () () i ()') < rearrange(r, 'j -> () () () j')
+ mask = F.pad(mask, (j - i, 0), value=False)
+ dots.masked_fill_(mask, mask_value)
+ del mask
+
+ if exists(self.sparse_topk) and self.sparse_topk < dots.shape[-1]:
+ top, _ = dots.topk(self.sparse_topk, dim=-1)
+ vk = top[..., -1].unsqueeze(-1).expand_as(dots)
+ mask = dots < vk
+ dots.masked_fill_(mask, mask_value)
+ del mask
+
+ attn = self.attn_fn(dots, dim=-1)
+ post_softmax_attn = attn
+
+ attn = self.dropout(attn)
+
+ if talking_heads:
+ attn = einsum('b h i j, h k -> b k i j', attn, self.post_softmax_proj).contiguous()
+
+ out = einsum('b h i j, b h j d -> b h i d', attn, v)
+ out = rearrange(out, 'b h n d -> b n (h d)')
+
+ intermediates = Intermediates(
+ pre_softmax_attn=pre_softmax_attn,
+ post_softmax_attn=post_softmax_attn
+ )
+
+ return self.to_out(out), intermediates
+
+
+class AttentionLayers(nn.Module):
+ def __init__(
+ self,
+ dim,
+ depth,
+ heads=8,
+ causal=False,
+ cross_attend=False,
+ only_cross=False,
+ use_scalenorm=False,
+ use_rmsnorm=False,
+ use_rezero=False,
+ rel_pos_num_buckets=32,
+ rel_pos_max_distance=128,
+ position_infused_attn=False,
+ custom_layers=None,
+ sandwich_coef=None,
+ par_ratio=None,
+ residual_attn=False,
+ cross_residual_attn=False,
+ macaron=False,
+ pre_norm=True,
+ gate_residual=False,
+ **kwargs
+ ):
+ super().__init__()
+ ff_kwargs, kwargs = groupby_prefix_and_trim('ff_', kwargs)
+ attn_kwargs, _ = groupby_prefix_and_trim('attn_', kwargs)
+
+ dim_head = attn_kwargs.get('dim_head', DEFAULT_DIM_HEAD)
+
+ self.dim = dim
+ self.depth = depth
+ self.layers = nn.ModuleList([])
+
+ self.has_pos_emb = position_infused_attn
+ self.pia_pos_emb = FixedPositionalEmbedding(dim) if position_infused_attn else None
+ self.rotary_pos_emb = always(None)
+
+ assert rel_pos_num_buckets <= rel_pos_max_distance, 'number of relative position buckets must be less than the relative position max distance'
+ self.rel_pos = None
+
+ self.pre_norm = pre_norm
+
+ self.residual_attn = residual_attn
+ self.cross_residual_attn = cross_residual_attn
+
+ norm_class = ScaleNorm if use_scalenorm else nn.LayerNorm
+ norm_class = RMSNorm if use_rmsnorm else norm_class
+ norm_fn = partial(norm_class, dim)
+
+ norm_fn = nn.Identity if use_rezero else norm_fn
+ branch_fn = Rezero if use_rezero else None
+
+ if cross_attend and not only_cross:
+ default_block = ('a', 'c', 'f')
+ elif cross_attend and only_cross:
+ default_block = ('c', 'f')
+ else:
+ default_block = ('a', 'f')
+
+ if macaron:
+ default_block = ('f',) + default_block
+
+ if exists(custom_layers):
+ layer_types = custom_layers
+ elif exists(par_ratio):
+ par_depth = depth * len(default_block)
+ assert 1 < par_ratio <= par_depth, 'par ratio out of range'
+ default_block = tuple(filter(not_equals('f'), default_block))
+ par_attn = par_depth // par_ratio
+ depth_cut = par_depth * 2 // 3 # 2 / 3 attention layer cutoff suggested by PAR paper
+ par_width = (depth_cut + depth_cut // par_attn) // par_attn
+ assert len(default_block) <= par_width, 'default block is too large for par_ratio'
+ par_block = default_block + ('f',) * (par_width - len(default_block))
+ par_head = par_block * par_attn
+ layer_types = par_head + ('f',) * (par_depth - len(par_head))
+ elif exists(sandwich_coef):
+ assert sandwich_coef > 0 and sandwich_coef <= depth, 'sandwich coefficient should be less than the depth'
+ layer_types = ('a',) * sandwich_coef + default_block * (depth - sandwich_coef) + ('f',) * sandwich_coef
+ else:
+ layer_types = default_block * depth
+
+ self.layer_types = layer_types
+ self.num_attn_layers = len(list(filter(equals('a'), layer_types)))
+
+ for layer_type in self.layer_types:
+ if layer_type == 'a':
+ layer = Attention(dim, heads=heads, causal=causal, **attn_kwargs)
+ elif layer_type == 'c':
+ layer = Attention(dim, heads=heads, **attn_kwargs)
+ elif layer_type == 'f':
+ layer = FeedForward(dim, **ff_kwargs)
+ layer = layer if not macaron else Scale(0.5, layer)
+ else:
+ raise Exception(f'invalid layer type {layer_type}')
+
+ if isinstance(layer, Attention) and exists(branch_fn):
+ layer = branch_fn(layer)
+
+ if gate_residual:
+ residual_fn = GRUGating(dim)
+ else:
+ residual_fn = Residual()
+
+ self.layers.append(nn.ModuleList([
+ norm_fn(),
+ layer,
+ residual_fn
+ ]))
+
+ def forward(
+ self,
+ x,
+ context=None,
+ mask=None,
+ context_mask=None,
+ mems=None,
+ return_hiddens=False
+ ):
+ hiddens = []
+ intermediates = []
+ prev_attn = None
+ prev_cross_attn = None
+
+ mems = mems.copy() if exists(mems) else [None] * self.num_attn_layers
+
+ for ind, (layer_type, (norm, block, residual_fn)) in enumerate(zip(self.layer_types, self.layers)):
+ is_last = ind == (len(self.layers) - 1)
+
+ if layer_type == 'a':
+ hiddens.append(x)
+ layer_mem = mems.pop(0)
+
+ residual = x
+
+ if self.pre_norm:
+ x = norm(x)
+
+ if layer_type == 'a':
+ out, inter = block(x, mask=mask, sinusoidal_emb=self.pia_pos_emb, rel_pos=self.rel_pos,
+ prev_attn=prev_attn, mem=layer_mem)
+ elif layer_type == 'c':
+ out, inter = block(x, context=context, mask=mask, context_mask=context_mask, prev_attn=prev_cross_attn)
+ elif layer_type == 'f':
+ out = block(x)
+
+ x = residual_fn(out, residual)
+
+ if layer_type in ('a', 'c'):
+ intermediates.append(inter)
+
+ if layer_type == 'a' and self.residual_attn:
+ prev_attn = inter.pre_softmax_attn
+ elif layer_type == 'c' and self.cross_residual_attn:
+ prev_cross_attn = inter.pre_softmax_attn
+
+ if not self.pre_norm and not is_last:
+ x = norm(x)
+
+ if return_hiddens:
+ intermediates = LayerIntermediates(
+ hiddens=hiddens,
+ attn_intermediates=intermediates
+ )
+
+ return x, intermediates
+
+ return x
+
+
+class Encoder(AttentionLayers):
+ def __init__(self, **kwargs):
+ assert 'causal' not in kwargs, 'cannot set causality on encoder'
+ super().__init__(causal=False, **kwargs)
+
+
+
+class TransformerWrapper(nn.Module):
+ def __init__(
+ self,
+ *,
+ num_tokens,
+ max_seq_len,
+ attn_layers,
+ emb_dim=None,
+ max_mem_len=0.,
+ emb_dropout=0.,
+ num_memory_tokens=None,
+ tie_embedding=False,
+ use_pos_emb=True
+ ):
+ super().__init__()
+ assert isinstance(attn_layers, AttentionLayers), 'attention layers must be one of Encoder or Decoder'
+
+ dim = attn_layers.dim
+ emb_dim = default(emb_dim, dim)
+
+ self.max_seq_len = max_seq_len
+ self.max_mem_len = max_mem_len
+ self.num_tokens = num_tokens
+
+ self.token_emb = nn.Embedding(num_tokens, emb_dim)
+ self.pos_emb = AbsolutePositionalEmbedding(emb_dim, max_seq_len) if (
+ use_pos_emb and not attn_layers.has_pos_emb) else always(0)
+ self.emb_dropout = nn.Dropout(emb_dropout)
+
+ self.project_emb = nn.Linear(emb_dim, dim) if emb_dim != dim else nn.Identity()
+ self.attn_layers = attn_layers
+ self.norm = nn.LayerNorm(dim)
+
+ self.init_()
+
+ self.to_logits = nn.Linear(dim, num_tokens) if not tie_embedding else lambda t: t @ self.token_emb.weight.t()
+
+ # memory tokens (like [cls]) from Memory Transformers paper
+ num_memory_tokens = default(num_memory_tokens, 0)
+ self.num_memory_tokens = num_memory_tokens
+ if num_memory_tokens > 0:
+ self.memory_tokens = nn.Parameter(torch.randn(num_memory_tokens, dim))
+
+ # let funnel encoder know number of memory tokens, if specified
+ if hasattr(attn_layers, 'num_memory_tokens'):
+ attn_layers.num_memory_tokens = num_memory_tokens
+
+ def init_(self):
+ nn.init.normal_(self.token_emb.weight, std=0.02)
+
+ def forward(
+ self,
+ x,
+ return_embeddings=False,
+ mask=None,
+ return_mems=False,
+ return_attn=False,
+ mems=None,
+ **kwargs
+ ):
+ b, n, device, num_mem = *x.shape, x.device, self.num_memory_tokens
+ x = self.token_emb(x)
+ x += self.pos_emb(x)
+ x = self.emb_dropout(x)
+
+ x = self.project_emb(x)
+
+ if num_mem > 0:
+ mem = repeat(self.memory_tokens, 'n d -> b n d', b=b)
+ x = torch.cat((mem, x), dim=1)
+
+ # auto-handle masking after appending memory tokens
+ if exists(mask):
+ mask = F.pad(mask, (num_mem, 0), value=True)
+
+ x, intermediates = self.attn_layers(x, mask=mask, mems=mems, return_hiddens=True, **kwargs)
+ x = self.norm(x)
+
+ mem, x = x[:, :num_mem], x[:, num_mem:]
+
+ out = self.to_logits(x) if not return_embeddings else x
+
+ if return_mems:
+ hiddens = intermediates.hiddens
+ new_mems = list(map(lambda pair: torch.cat(pair, dim=-2), zip(mems, hiddens))) if exists(mems) else hiddens
+ new_mems = list(map(lambda t: t[..., -self.max_mem_len:, :].detach(), new_mems))
+ return out, new_mems
+
+ if return_attn:
+ attn_maps = list(map(lambda t: t.post_softmax_attn, intermediates.attn_intermediates))
+ return out, attn_maps
+
+ return out
+
diff --git a/OCR_earsing/latent_diffusion/ldm/util.py b/OCR_earsing/latent_diffusion/ldm/util.py
new file mode 100644
index 0000000..8ba3885
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm/util.py
@@ -0,0 +1,203 @@
+import importlib
+
+import torch
+import numpy as np
+from collections import abc
+from einops import rearrange
+from functools import partial
+
+import multiprocessing as mp
+from threading import Thread
+from queue import Queue
+
+from inspect import isfunction
+from PIL import Image, ImageDraw, ImageFont
+
+
+def log_txt_as_img(wh, xc, size=10):
+ # wh a tuple of (width, height)
+ # xc a list of captions to plot
+ b = len(xc)
+ txts = list()
+ for bi in range(b):
+ txt = Image.new("RGB", wh, color="white")
+ draw = ImageDraw.Draw(txt)
+ font = ImageFont.truetype('data/DejaVuSans.ttf', size=size)
+ nc = int(40 * (wh[0] / 256))
+ lines = "\n".join(xc[bi][start:start + nc] for start in range(0, len(xc[bi]), nc))
+
+ try:
+ draw.text((0, 0), lines, fill="black", font=font)
+ except UnicodeEncodeError:
+ print("Cant encode string for logging. Skipping.")
+
+ txt = np.array(txt).transpose(2, 0, 1) / 127.5 - 1.0
+ txts.append(txt)
+ txts = np.stack(txts)
+ txts = torch.tensor(txts)
+ return txts
+
+
+def ismap(x):
+ if not isinstance(x, torch.Tensor):
+ return False
+ return (len(x.shape) == 4) and (x.shape[1] > 3)
+
+
+def isimage(x):
+ if not isinstance(x, torch.Tensor):
+ return False
+ return (len(x.shape) == 4) and (x.shape[1] == 3 or x.shape[1] == 1)
+
+
+def exists(x):
+ return x is not None
+
+
+def default(val, d):
+ if exists(val):
+ return val
+ return d() if isfunction(d) else d
+
+
+def mean_flat(tensor):
+ """
+ https://github.com/openai/guided-diffusion/blob/27c20a8fab9cb472df5d6bdd6c8d11c8f430b924/guided_diffusion/nn.py#L86
+ Take the mean over all non-batch dimensions.
+ """
+ return tensor.mean(dim=list(range(1, len(tensor.shape))))
+
+
+def count_params(model, verbose=False):
+ total_params = sum(p.numel() for p in model.parameters())
+ if verbose:
+ print(f"{model.__class__.__name__} has {total_params * 1.e-6:.2f} M params.")
+ return total_params
+
+
+def instantiate_from_config(config):
+ if not "target" in config:
+ if config == '__is_first_stage__':
+ return None
+ elif config == "__is_unconditional__":
+ return None
+ raise KeyError("Expected key `target` to instantiate.")
+ return get_obj_from_str(config["target"])(**config.get("params", dict()))
+
+
+def get_obj_from_str(string, reload=False):
+ module, cls = string.rsplit(".", 1)
+ if reload:
+ module_imp = importlib.import_module(module)
+ importlib.reload(module_imp)
+ return getattr(importlib.import_module(module, package=None), cls)
+
+
+def _do_parallel_data_prefetch(func, Q, data, idx, idx_to_fn=False):
+ # create dummy dataset instance
+
+ # run prefetching
+ if idx_to_fn:
+ res = func(data, worker_id=idx)
+ else:
+ res = func(data)
+ Q.put([idx, res])
+ Q.put("Done")
+
+
+def parallel_data_prefetch(
+ func: callable, data, n_proc, target_data_type="ndarray", cpu_intensive=True, use_worker_id=False
+):
+ # if target_data_type not in ["ndarray", "list"]:
+ # raise ValueError(
+ # "Data, which is passed to parallel_data_prefetch has to be either of type list or ndarray."
+ # )
+ if isinstance(data, np.ndarray) and target_data_type == "list":
+ raise ValueError("list expected but function got ndarray.")
+ elif isinstance(data, abc.Iterable):
+ if isinstance(data, dict):
+ print(
+ f'WARNING:"data" argument passed to parallel_data_prefetch is a dict: Using only its values and disregarding keys.'
+ )
+ data = list(data.values())
+ if target_data_type == "ndarray":
+ data = np.asarray(data)
+ else:
+ data = list(data)
+ else:
+ raise TypeError(
+ f"The data, that shall be processed parallel has to be either an np.ndarray or an Iterable, but is actually {type(data)}."
+ )
+
+ if cpu_intensive:
+ Q = mp.Queue(1000)
+ proc = mp.Process
+ else:
+ Q = Queue(1000)
+ proc = Thread
+ # spawn processes
+ if target_data_type == "ndarray":
+ arguments = [
+ [func, Q, part, i, use_worker_id]
+ for i, part in enumerate(np.array_split(data, n_proc))
+ ]
+ else:
+ step = (
+ int(len(data) / n_proc + 1)
+ if len(data) % n_proc != 0
+ else int(len(data) / n_proc)
+ )
+ arguments = [
+ [func, Q, part, i, use_worker_id]
+ for i, part in enumerate(
+ [data[i: i + step] for i in range(0, len(data), step)]
+ )
+ ]
+ processes = []
+ for i in range(n_proc):
+ p = proc(target=_do_parallel_data_prefetch, args=arguments[i])
+ processes += [p]
+
+ # start processes
+ print(f"Start prefetching...")
+ import time
+
+ start = time.time()
+ gather_res = [[] for _ in range(n_proc)]
+ try:
+ for p in processes:
+ p.start()
+
+ k = 0
+ while k < n_proc:
+ # get result
+ res = Q.get()
+ if res == "Done":
+ k += 1
+ else:
+ gather_res[res[0]] = res[1]
+
+ except Exception as e:
+ print("Exception: ", e)
+ for p in processes:
+ p.terminate()
+
+ raise e
+ finally:
+ for p in processes:
+ p.join()
+ print(f"Prefetching complete. [{time.time() - start} sec.]")
+
+ if target_data_type == 'ndarray':
+ if not isinstance(gather_res[0], np.ndarray):
+ return np.concatenate([np.asarray(r) for r in gather_res], axis=0)
+
+ # order outputs
+ return np.concatenate(gather_res, axis=0)
+ elif target_data_type == 'list':
+ out = []
+ for r in gather_res:
+ out.extend(r)
+ return out
+ else:
+ return gather_res
diff --git a/OCR_earsing/latent_diffusion/ldm_erase_text.py b/OCR_earsing/latent_diffusion/ldm_erase_text.py
new file mode 100644
index 0000000..503ded6
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/ldm_erase_text.py
@@ -0,0 +1,96 @@
+import argparse, os, sys, glob
+from omegaconf import OmegaConf
+from PIL import Image
+from tqdm import tqdm
+import numpy as np
+import torch
+from ldm.util import instantiate_from_config
+from ldm.models.diffusion.ddim import DDIMSampler
+
+
+def make_batch(image, mask_pil_image, img_size, device):
+ if isinstance(image, str):
+ if img_size is not None:
+ image = np.array(Image.open(image).convert("RGB").resize(img_size))
+ else:
+ image = np.array(Image.open(image).convert("RGB")) # need to resize to a image_size
+ else:
+ if img_size is not None:
+ image = np.array(image.convert("RGB").resize(img_size))
+ else:
+ image = np.array(image.convert("RGB")) # need to resize to a image_size
+ image = image.astype(np.float32) / 255.0
+ image = image[None].transpose(0, 3, 1, 2)
+ image = torch.from_numpy(image)
+
+ if img_size is not None:
+ mask = np.array(mask_pil_image.convert("L").resize(img_size))
+ else:
+ mask = np.array(mask_pil_image.convert("L"))
+ mask = mask.astype(np.float32) / 255.0
+ mask = mask[None, None]
+ mask[mask < 0.5] = 0
+ mask[mask >= 0.5] = 1
+ mask = torch.from_numpy(mask)
+
+ masked_image = (1 - mask) * image
+
+ batch = {"image": image, "mask": mask, "masked_image": masked_image}
+ for k in batch:
+ batch[k] = batch[k].to(device=device)
+ batch[k] = batch[k] * 2.0 - 1.0
+ return batch
+
+
+def erase_text_from_image(img_path,
+ mask_pil_img,
+ model,
+ device,
+ opt,
+ img_size=None,
+ steps=None):
+ sampler = DDIMSampler(model)
+ with torch.no_grad():
+ with model.ema_scope():
+
+ if img_size is None:
+ batch = make_batch(
+ img_path,
+ mask_pil_img,
+ img_size=opt.img_size,
+ device=device)
+ else:
+ batch = make_batch(
+ img_path, mask_pil_img, img_size=img_size, device=device)
+ # encode masked image and concat downsampled mask
+ c = model.cond_stage_model.encode(batch["masked_image"])
+ cc = torch.nn.functional.interpolate(
+ batch["mask"], size=c.shape[-2:])
+ c = torch.cat((c, cc), dim=1)
+
+ shape = (c.shape[1] - 1, ) + c.shape[2:]
+ if steps is None:
+ samples_ddim, _ = sampler.sample(
+ S=opt.steps,
+ conditioning=c,
+ batch_size=c.shape[0],
+ shape=shape,
+ verbose=False)
+ else:
+ samples_ddim, _ = sampler.sample(
+ S=steps,
+ conditioning=c,
+ batch_size=c.shape[0],
+ shape=shape,
+ verbose=False)
+ x_samples_ddim = model.decode_first_stage(samples_ddim)
+
+ image = torch.clamp((batch["image"] + 1.0) / 2.0, min=0.0, max=1.0)
+ mask = torch.clamp((batch["mask"] + 1.0) / 2.0, min=0.0, max=1.0)
+ predicted_image = torch.clamp(
+ (x_samples_ddim + 1.0) / 2.0, min=0.0, max=1.0)
+
+ inpainted = (1 - mask) * image + mask * predicted_image
+ inpainted = inpainted.cpu().numpy().transpose(0, 2, 3, 1)[0] * 255
+
+ return Image.fromarray(inpainted.astype(np.uint8))
diff --git a/OCR_earsing/latent_diffusion/taming/__init__.py b/OCR_earsing/latent_diffusion/taming/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/models/__init__.py b/OCR_earsing/latent_diffusion/taming/models/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/models/cond_transformer.py b/OCR_earsing/latent_diffusion/taming/models/cond_transformer.py
new file mode 100644
index 0000000..e4c6373
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/models/cond_transformer.py
@@ -0,0 +1,352 @@
+import os, math
+import torch
+import torch.nn.functional as F
+import pytorch_lightning as pl
+
+from main import instantiate_from_config
+from taming.modules.util import SOSProvider
+
+
+def disabled_train(self, mode=True):
+ """Overwrite model.train with this function to make sure train/eval mode
+ does not change anymore."""
+ return self
+
+
+class Net2NetTransformer(pl.LightningModule):
+ def __init__(self,
+ transformer_config,
+ first_stage_config,
+ cond_stage_config,
+ permuter_config=None,
+ ckpt_path=None,
+ ignore_keys=[],
+ first_stage_key="image",
+ cond_stage_key="depth",
+ downsample_cond_size=-1,
+ pkeep=1.0,
+ sos_token=0,
+ unconditional=False,
+ ):
+ super().__init__()
+ self.be_unconditional = unconditional
+ self.sos_token = sos_token
+ self.first_stage_key = first_stage_key
+ self.cond_stage_key = cond_stage_key
+ self.init_first_stage_from_ckpt(first_stage_config)
+ self.init_cond_stage_from_ckpt(cond_stage_config)
+ if permuter_config is None:
+ permuter_config = {"target": "taming.modules.transformer.permuter.Identity"}
+ self.permuter = instantiate_from_config(config=permuter_config)
+ self.transformer = instantiate_from_config(config=transformer_config)
+
+ if ckpt_path is not None:
+ self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys)
+ self.downsample_cond_size = downsample_cond_size
+ self.pkeep = pkeep
+
+ def init_from_ckpt(self, path, ignore_keys=list()):
+ sd = torch.load(path, map_location="cpu")["state_dict"]
+ for k in sd.keys():
+ for ik in ignore_keys:
+ if k.startswith(ik):
+ self.print("Deleting key {} from state_dict.".format(k))
+ del sd[k]
+ self.load_state_dict(sd, strict=False)
+ print(f"Restored from {path}")
+
+ def init_first_stage_from_ckpt(self, config):
+ model = instantiate_from_config(config)
+ model = model.eval()
+ model.train = disabled_train
+ self.first_stage_model = model
+
+ def init_cond_stage_from_ckpt(self, config):
+ if config == "__is_first_stage__":
+ print("Using first stage also as cond stage.")
+ self.cond_stage_model = self.first_stage_model
+ elif config == "__is_unconditional__" or self.be_unconditional:
+ print(f"Using no cond stage. Assuming the training is intended to be unconditional. "
+ f"Prepending {self.sos_token} as a sos token.")
+ self.be_unconditional = True
+ self.cond_stage_key = self.first_stage_key
+ self.cond_stage_model = SOSProvider(self.sos_token)
+ else:
+ model = instantiate_from_config(config)
+ model = model.eval()
+ model.train = disabled_train
+ self.cond_stage_model = model
+
+ def forward(self, x, c):
+ # one step to produce the logits
+ _, z_indices = self.encode_to_z(x)
+ _, c_indices = self.encode_to_c(c)
+
+ if self.training and self.pkeep < 1.0:
+ mask = torch.bernoulli(self.pkeep*torch.ones(z_indices.shape,
+ device=z_indices.device))
+ mask = mask.round().to(dtype=torch.int64)
+ r_indices = torch.randint_like(z_indices, self.transformer.config.vocab_size)
+ a_indices = mask*z_indices+(1-mask)*r_indices
+ else:
+ a_indices = z_indices
+
+ cz_indices = torch.cat((c_indices, a_indices), dim=1)
+
+ # target includes all sequence elements (no need to handle first one
+ # differently because we are conditioning)
+ target = z_indices
+ # make the prediction
+ logits, _ = self.transformer(cz_indices[:, :-1])
+ # cut off conditioning outputs - output i corresponds to p(z_i | z_{ -1:
+ c = F.interpolate(c, size=(self.downsample_cond_size, self.downsample_cond_size))
+ quant_c, _, [_,_,indices] = self.cond_stage_model.encode(c)
+ if len(indices.shape) > 2:
+ indices = indices.view(c.shape[0], -1)
+ return quant_c, indices
+
+ @torch.no_grad()
+ def decode_to_img(self, index, zshape):
+ index = self.permuter(index, reverse=True)
+ bhwc = (zshape[0],zshape[2],zshape[3],zshape[1])
+ quant_z = self.first_stage_model.quantize.get_codebook_entry(
+ index.reshape(-1), shape=bhwc)
+ x = self.first_stage_model.decode(quant_z)
+ return x
+
+ @torch.no_grad()
+ def log_images(self, batch, temperature=None, top_k=None, callback=None, lr_interface=False, **kwargs):
+ log = dict()
+
+ N = 4
+ if lr_interface:
+ x, c = self.get_xc(batch, N, diffuse=False, upsample_factor=8)
+ else:
+ x, c = self.get_xc(batch, N)
+ x = x.to(device=self.device)
+ c = c.to(device=self.device)
+
+ quant_z, z_indices = self.encode_to_z(x)
+ quant_c, c_indices = self.encode_to_c(c)
+
+ # create a "half"" sample
+ z_start_indices = z_indices[:,:z_indices.shape[1]//2]
+ index_sample = self.sample(z_start_indices, c_indices,
+ steps=z_indices.shape[1]-z_start_indices.shape[1],
+ temperature=temperature if temperature is not None else 1.0,
+ sample=True,
+ top_k=top_k if top_k is not None else 100,
+ callback=callback if callback is not None else lambda k: None)
+ x_sample = self.decode_to_img(index_sample, quant_z.shape)
+
+ # sample
+ z_start_indices = z_indices[:, :0]
+ index_sample = self.sample(z_start_indices, c_indices,
+ steps=z_indices.shape[1],
+ temperature=temperature if temperature is not None else 1.0,
+ sample=True,
+ top_k=top_k if top_k is not None else 100,
+ callback=callback if callback is not None else lambda k: None)
+ x_sample_nopix = self.decode_to_img(index_sample, quant_z.shape)
+
+ # det sample
+ z_start_indices = z_indices[:, :0]
+ index_sample = self.sample(z_start_indices, c_indices,
+ steps=z_indices.shape[1],
+ sample=False,
+ callback=callback if callback is not None else lambda k: None)
+ x_sample_det = self.decode_to_img(index_sample, quant_z.shape)
+
+ # reconstruction
+ x_rec = self.decode_to_img(z_indices, quant_z.shape)
+
+ log["inputs"] = x
+ log["reconstructions"] = x_rec
+
+ if self.cond_stage_key in ["objects_bbox", "objects_center_points"]:
+ figure_size = (x_rec.shape[2], x_rec.shape[3])
+ dataset = kwargs["pl_module"].trainer.datamodule.datasets["validation"]
+ label_for_category_no = dataset.get_textual_label_for_category_no
+ plotter = dataset.conditional_builders[self.cond_stage_key].plot
+ log["conditioning"] = torch.zeros_like(log["reconstructions"])
+ for i in range(quant_c.shape[0]):
+ log["conditioning"][i] = plotter(quant_c[i], label_for_category_no, figure_size)
+ log["conditioning_rec"] = log["conditioning"]
+ elif self.cond_stage_key != "image":
+ cond_rec = self.cond_stage_model.decode(quant_c)
+ if self.cond_stage_key == "segmentation":
+ # get image from segmentation mask
+ num_classes = cond_rec.shape[1]
+
+ c = torch.argmax(c, dim=1, keepdim=True)
+ c = F.one_hot(c, num_classes=num_classes)
+ c = c.squeeze(1).permute(0, 3, 1, 2).float()
+ c = self.cond_stage_model.to_rgb(c)
+
+ cond_rec = torch.argmax(cond_rec, dim=1, keepdim=True)
+ cond_rec = F.one_hot(cond_rec, num_classes=num_classes)
+ cond_rec = cond_rec.squeeze(1).permute(0, 3, 1, 2).float()
+ cond_rec = self.cond_stage_model.to_rgb(cond_rec)
+ log["conditioning_rec"] = cond_rec
+ log["conditioning"] = c
+
+ log["samples_half"] = x_sample
+ log["samples_nopix"] = x_sample_nopix
+ log["samples_det"] = x_sample_det
+ return log
+
+ def get_input(self, key, batch):
+ x = batch[key]
+ if len(x.shape) == 3:
+ x = x[..., None]
+ if len(x.shape) == 4:
+ x = x.permute(0, 3, 1, 2).to(memory_format=torch.contiguous_format)
+ if x.dtype == torch.double:
+ x = x.float()
+ return x
+
+ def get_xc(self, batch, N=None):
+ x = self.get_input(self.first_stage_key, batch)
+ c = self.get_input(self.cond_stage_key, batch)
+ if N is not None:
+ x = x[:N]
+ c = c[:N]
+ return x, c
+
+ def shared_step(self, batch, batch_idx):
+ x, c = self.get_xc(batch)
+ logits, target = self(x, c)
+ loss = F.cross_entropy(logits.reshape(-1, logits.size(-1)), target.reshape(-1))
+ return loss
+
+ def training_step(self, batch, batch_idx):
+ loss = self.shared_step(batch, batch_idx)
+ self.log("train/loss", loss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ return loss
+
+ def validation_step(self, batch, batch_idx):
+ loss = self.shared_step(batch, batch_idx)
+ self.log("val/loss", loss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ return loss
+
+ def configure_optimizers(self):
+ """
+ Following minGPT:
+ This long function is unfortunately doing something very simple and is being very defensive:
+ We are separating out all parameters of the model into two buckets: those that will experience
+ weight decay for regularization and those that won't (biases, and layernorm/embedding weights).
+ We are then returning the PyTorch optimizer object.
+ """
+ # separate out all parameters to those that will and won't experience regularizing weight decay
+ decay = set()
+ no_decay = set()
+ whitelist_weight_modules = (torch.nn.Linear, )
+ blacklist_weight_modules = (torch.nn.LayerNorm, torch.nn.Embedding)
+ for mn, m in self.transformer.named_modules():
+ for pn, p in m.named_parameters():
+ fpn = '%s.%s' % (mn, pn) if mn else pn # full param name
+
+ if pn.endswith('bias'):
+ # all biases will not be decayed
+ no_decay.add(fpn)
+ elif pn.endswith('weight') and isinstance(m, whitelist_weight_modules):
+ # weights of whitelist modules will be weight decayed
+ decay.add(fpn)
+ elif pn.endswith('weight') and isinstance(m, blacklist_weight_modules):
+ # weights of blacklist modules will NOT be weight decayed
+ no_decay.add(fpn)
+
+ # special case the position embedding parameter in the root GPT module as not decayed
+ no_decay.add('pos_emb')
+
+ # validate that we considered every parameter
+ param_dict = {pn: p for pn, p in self.transformer.named_parameters()}
+ inter_params = decay & no_decay
+ union_params = decay | no_decay
+ assert len(inter_params) == 0, "parameters %s made it into both decay/no_decay sets!" % (str(inter_params), )
+ assert len(param_dict.keys() - union_params) == 0, "parameters %s were not separated into either decay/no_decay set!" \
+ % (str(param_dict.keys() - union_params), )
+
+ # create the pytorch optimizer object
+ optim_groups = [
+ {"params": [param_dict[pn] for pn in sorted(list(decay))], "weight_decay": 0.01},
+ {"params": [param_dict[pn] for pn in sorted(list(no_decay))], "weight_decay": 0.0},
+ ]
+ optimizer = torch.optim.AdamW(optim_groups, lr=self.learning_rate, betas=(0.9, 0.95))
+ return optimizer
diff --git a/OCR_earsing/latent_diffusion/taming/models/dummy_cond_stage.py b/OCR_earsing/latent_diffusion/taming/models/dummy_cond_stage.py
new file mode 100644
index 0000000..6e19938
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/models/dummy_cond_stage.py
@@ -0,0 +1,22 @@
+from torch import Tensor
+
+
+class DummyCondStage:
+ def __init__(self, conditional_key):
+ self.conditional_key = conditional_key
+ self.train = None
+
+ def eval(self):
+ return self
+
+ @staticmethod
+ def encode(c: Tensor):
+ return c, None, (None, None, c)
+
+ @staticmethod
+ def decode(c: Tensor):
+ return c
+
+ @staticmethod
+ def to_rgb(c: Tensor):
+ return c
diff --git a/OCR_earsing/latent_diffusion/taming/models/vqgan.py b/OCR_earsing/latent_diffusion/taming/models/vqgan.py
new file mode 100644
index 0000000..a6950ba
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/models/vqgan.py
@@ -0,0 +1,404 @@
+import torch
+import torch.nn.functional as F
+import pytorch_lightning as pl
+
+from main import instantiate_from_config
+
+from taming.modules.diffusionmodules.model import Encoder, Decoder
+from taming.modules.vqvae.quantize import VectorQuantizer2 as VectorQuantizer
+from taming.modules.vqvae.quantize import GumbelQuantize
+from taming.modules.vqvae.quantize import EMAVectorQuantizer
+
+class VQModel(pl.LightningModule):
+ def __init__(self,
+ ddconfig,
+ lossconfig,
+ n_embed,
+ embed_dim,
+ ckpt_path=None,
+ ignore_keys=[],
+ image_key="image",
+ colorize_nlabels=None,
+ monitor=None,
+ remap=None,
+ sane_index_shape=False, # tell vector quantizer to return indices as bhw
+ ):
+ super().__init__()
+ self.image_key = image_key
+ self.encoder = Encoder(**ddconfig)
+ self.decoder = Decoder(**ddconfig)
+ self.loss = instantiate_from_config(lossconfig)
+ self.quantize = VectorQuantizer(n_embed, embed_dim, beta=0.25,
+ remap=remap, sane_index_shape=sane_index_shape)
+ self.quant_conv = torch.nn.Conv2d(ddconfig["z_channels"], embed_dim, 1)
+ self.post_quant_conv = torch.nn.Conv2d(embed_dim, ddconfig["z_channels"], 1)
+ if ckpt_path is not None:
+ self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys)
+ self.image_key = image_key
+ if colorize_nlabels is not None:
+ assert type(colorize_nlabels)==int
+ self.register_buffer("colorize", torch.randn(3, colorize_nlabels, 1, 1))
+ if monitor is not None:
+ self.monitor = monitor
+
+ def init_from_ckpt(self, path, ignore_keys=list()):
+ sd = torch.load(path, map_location="cpu")["state_dict"]
+ keys = list(sd.keys())
+ for k in keys:
+ for ik in ignore_keys:
+ if k.startswith(ik):
+ print("Deleting key {} from state_dict.".format(k))
+ del sd[k]
+ self.load_state_dict(sd, strict=False)
+ print(f"Restored from {path}")
+
+ def encode(self, x):
+ h = self.encoder(x)
+ h = self.quant_conv(h)
+ quant, emb_loss, info = self.quantize(h)
+ return quant, emb_loss, info
+
+ def decode(self, quant):
+ quant = self.post_quant_conv(quant)
+ dec = self.decoder(quant)
+ return dec
+
+ def decode_code(self, code_b):
+ quant_b = self.quantize.embed_code(code_b)
+ dec = self.decode(quant_b)
+ return dec
+
+ def forward(self, input):
+ quant, diff, _ = self.encode(input)
+ dec = self.decode(quant)
+ return dec, diff
+
+ def get_input(self, batch, k):
+ x = batch[k]
+ if len(x.shape) == 3:
+ x = x[..., None]
+ x = x.permute(0, 3, 1, 2).to(memory_format=torch.contiguous_format)
+ return x.float()
+
+ def training_step(self, batch, batch_idx, optimizer_idx):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x)
+
+ if optimizer_idx == 0:
+ # autoencode
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train")
+
+ self.log("train/aeloss", aeloss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return aeloss
+
+ if optimizer_idx == 1:
+ # discriminator
+ discloss, log_dict_disc = self.loss(qloss, x, xrec, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train")
+ self.log("train/discloss", discloss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ self.log_dict(log_dict_disc, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return discloss
+
+ def validation_step(self, batch, batch_idx):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x)
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, 0, self.global_step,
+ last_layer=self.get_last_layer(), split="val")
+
+ discloss, log_dict_disc = self.loss(qloss, x, xrec, 1, self.global_step,
+ last_layer=self.get_last_layer(), split="val")
+ rec_loss = log_dict_ae["val/rec_loss"]
+ self.log("val/rec_loss", rec_loss,
+ prog_bar=True, logger=True, on_step=True, on_epoch=True, sync_dist=True)
+ self.log("val/aeloss", aeloss,
+ prog_bar=True, logger=True, on_step=True, on_epoch=True, sync_dist=True)
+ self.log_dict(log_dict_ae)
+ self.log_dict(log_dict_disc)
+ return self.log_dict
+
+ def configure_optimizers(self):
+ lr = self.learning_rate
+ opt_ae = torch.optim.Adam(list(self.encoder.parameters())+
+ list(self.decoder.parameters())+
+ list(self.quantize.parameters())+
+ list(self.quant_conv.parameters())+
+ list(self.post_quant_conv.parameters()),
+ lr=lr, betas=(0.5, 0.9))
+ opt_disc = torch.optim.Adam(self.loss.discriminator.parameters(),
+ lr=lr, betas=(0.5, 0.9))
+ return [opt_ae, opt_disc], []
+
+ def get_last_layer(self):
+ return self.decoder.conv_out.weight
+
+ def log_images(self, batch, **kwargs):
+ log = dict()
+ x = self.get_input(batch, self.image_key)
+ x = x.to(self.device)
+ xrec, _ = self(x)
+ if x.shape[1] > 3:
+ # colorize with random projection
+ assert xrec.shape[1] > 3
+ x = self.to_rgb(x)
+ xrec = self.to_rgb(xrec)
+ log["inputs"] = x
+ log["reconstructions"] = xrec
+ return log
+
+ def to_rgb(self, x):
+ assert self.image_key == "segmentation"
+ if not hasattr(self, "colorize"):
+ self.register_buffer("colorize", torch.randn(3, x.shape[1], 1, 1).to(x))
+ x = F.conv2d(x, weight=self.colorize)
+ x = 2.*(x-x.min())/(x.max()-x.min()) - 1.
+ return x
+
+
+class VQSegmentationModel(VQModel):
+ def __init__(self, n_labels, *args, **kwargs):
+ super().__init__(*args, **kwargs)
+ self.register_buffer("colorize", torch.randn(3, n_labels, 1, 1))
+
+ def configure_optimizers(self):
+ lr = self.learning_rate
+ opt_ae = torch.optim.Adam(list(self.encoder.parameters())+
+ list(self.decoder.parameters())+
+ list(self.quantize.parameters())+
+ list(self.quant_conv.parameters())+
+ list(self.post_quant_conv.parameters()),
+ lr=lr, betas=(0.5, 0.9))
+ return opt_ae
+
+ def training_step(self, batch, batch_idx):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x)
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, split="train")
+ self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return aeloss
+
+ def validation_step(self, batch, batch_idx):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x)
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, split="val")
+ self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ total_loss = log_dict_ae["val/total_loss"]
+ self.log("val/total_loss", total_loss,
+ prog_bar=True, logger=True, on_step=True, on_epoch=True, sync_dist=True)
+ return aeloss
+
+ @torch.no_grad()
+ def log_images(self, batch, **kwargs):
+ log = dict()
+ x = self.get_input(batch, self.image_key)
+ x = x.to(self.device)
+ xrec, _ = self(x)
+ if x.shape[1] > 3:
+ # colorize with random projection
+ assert xrec.shape[1] > 3
+ # convert logits to indices
+ xrec = torch.argmax(xrec, dim=1, keepdim=True)
+ xrec = F.one_hot(xrec, num_classes=x.shape[1])
+ xrec = xrec.squeeze(1).permute(0, 3, 1, 2).float()
+ x = self.to_rgb(x)
+ xrec = self.to_rgb(xrec)
+ log["inputs"] = x
+ log["reconstructions"] = xrec
+ return log
+
+
+class VQNoDiscModel(VQModel):
+ def __init__(self,
+ ddconfig,
+ lossconfig,
+ n_embed,
+ embed_dim,
+ ckpt_path=None,
+ ignore_keys=[],
+ image_key="image",
+ colorize_nlabels=None
+ ):
+ super().__init__(ddconfig=ddconfig, lossconfig=lossconfig, n_embed=n_embed, embed_dim=embed_dim,
+ ckpt_path=ckpt_path, ignore_keys=ignore_keys, image_key=image_key,
+ colorize_nlabels=colorize_nlabels)
+
+ def training_step(self, batch, batch_idx):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x)
+ # autoencode
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, self.global_step, split="train")
+ output = pl.TrainResult(minimize=aeloss)
+ output.log("train/aeloss", aeloss,
+ prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ output.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return output
+
+ def validation_step(self, batch, batch_idx):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x)
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, self.global_step, split="val")
+ rec_loss = log_dict_ae["val/rec_loss"]
+ output = pl.EvalResult(checkpoint_on=rec_loss)
+ output.log("val/rec_loss", rec_loss,
+ prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ output.log("val/aeloss", aeloss,
+ prog_bar=True, logger=True, on_step=True, on_epoch=True)
+ output.log_dict(log_dict_ae)
+
+ return output
+
+ def configure_optimizers(self):
+ optimizer = torch.optim.Adam(list(self.encoder.parameters())+
+ list(self.decoder.parameters())+
+ list(self.quantize.parameters())+
+ list(self.quant_conv.parameters())+
+ list(self.post_quant_conv.parameters()),
+ lr=self.learning_rate, betas=(0.5, 0.9))
+ return optimizer
+
+
+class GumbelVQ(VQModel):
+ def __init__(self,
+ ddconfig,
+ lossconfig,
+ n_embed,
+ embed_dim,
+ temperature_scheduler_config,
+ ckpt_path=None,
+ ignore_keys=[],
+ image_key="image",
+ colorize_nlabels=None,
+ monitor=None,
+ kl_weight=1e-8,
+ remap=None,
+ ):
+
+ z_channels = ddconfig["z_channels"]
+ super().__init__(ddconfig,
+ lossconfig,
+ n_embed,
+ embed_dim,
+ ckpt_path=None,
+ ignore_keys=ignore_keys,
+ image_key=image_key,
+ colorize_nlabels=colorize_nlabels,
+ monitor=monitor,
+ )
+
+ self.loss.n_classes = n_embed
+ self.vocab_size = n_embed
+
+ self.quantize = GumbelQuantize(z_channels, embed_dim,
+ n_embed=n_embed,
+ kl_weight=kl_weight, temp_init=1.0,
+ remap=remap)
+
+ self.temperature_scheduler = instantiate_from_config(temperature_scheduler_config) # annealing of temp
+
+ if ckpt_path is not None:
+ self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys)
+
+ def temperature_scheduling(self):
+ self.quantize.temperature = self.temperature_scheduler(self.global_step)
+
+ def encode_to_prequant(self, x):
+ h = self.encoder(x)
+ h = self.quant_conv(h)
+ return h
+
+ def decode_code(self, code_b):
+ raise NotImplementedError
+
+ def training_step(self, batch, batch_idx, optimizer_idx):
+ self.temperature_scheduling()
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x)
+
+ if optimizer_idx == 0:
+ # autoencode
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train")
+
+ self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ self.log("temperature", self.quantize.temperature, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return aeloss
+
+ if optimizer_idx == 1:
+ # discriminator
+ discloss, log_dict_disc = self.loss(qloss, x, xrec, optimizer_idx, self.global_step,
+ last_layer=self.get_last_layer(), split="train")
+ self.log_dict(log_dict_disc, prog_bar=False, logger=True, on_step=True, on_epoch=True)
+ return discloss
+
+ def validation_step(self, batch, batch_idx):
+ x = self.get_input(batch, self.image_key)
+ xrec, qloss = self(x, return_pred_indices=True)
+ aeloss, log_dict_ae = self.loss(qloss, x, xrec, 0, self.global_step,
+ last_layer=self.get_last_layer(), split="val")
+
+ discloss, log_dict_disc = self.loss(qloss, x, xrec, 1, self.global_step,
+ last_layer=self.get_last_layer(), split="val")
+ rec_loss = log_dict_ae["val/rec_loss"]
+ self.log("val/rec_loss", rec_loss,
+ prog_bar=True, logger=True, on_step=False, on_epoch=True, sync_dist=True)
+ self.log("val/aeloss", aeloss,
+ prog_bar=True, logger=True, on_step=False, on_epoch=True, sync_dist=True)
+ self.log_dict(log_dict_ae)
+ self.log_dict(log_dict_disc)
+ return self.log_dict
+
+ def log_images(self, batch, **kwargs):
+ log = dict()
+ x = self.get_input(batch, self.image_key)
+ x = x.to(self.device)
+ # encode
+ h = self.encoder(x)
+ h = self.quant_conv(h)
+ quant, _, _ = self.quantize(h)
+ # decode
+ x_rec = self.decode(quant)
+ log["inputs"] = x
+ log["reconstructions"] = x_rec
+ return log
+
+
+class EMAVQ(VQModel):
+ def __init__(self,
+ ddconfig,
+ lossconfig,
+ n_embed,
+ embed_dim,
+ ckpt_path=None,
+ ignore_keys=[],
+ image_key="image",
+ colorize_nlabels=None,
+ monitor=None,
+ remap=None,
+ sane_index_shape=False, # tell vector quantizer to return indices as bhw
+ ):
+ super().__init__(ddconfig,
+ lossconfig,
+ n_embed,
+ embed_dim,
+ ckpt_path=None,
+ ignore_keys=ignore_keys,
+ image_key=image_key,
+ colorize_nlabels=colorize_nlabels,
+ monitor=monitor,
+ )
+ self.quantize = EMAVectorQuantizer(n_embed=n_embed,
+ embedding_dim=embed_dim,
+ beta=0.25,
+ remap=remap)
+ def configure_optimizers(self):
+ lr = self.learning_rate
+ #Remove self.quantize from parameter list since it is updated via EMA
+ opt_ae = torch.optim.Adam(list(self.encoder.parameters())+
+ list(self.decoder.parameters())+
+ list(self.quant_conv.parameters())+
+ list(self.post_quant_conv.parameters()),
+ lr=lr, betas=(0.5, 0.9))
+ opt_disc = torch.optim.Adam(self.loss.discriminator.parameters(),
+ lr=lr, betas=(0.5, 0.9))
+ return [opt_ae, opt_disc], []
\ No newline at end of file
diff --git a/OCR_earsing/latent_diffusion/taming/modules/__init__.py b/OCR_earsing/latent_diffusion/taming/modules/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/modules/diffusionmodules/__init__.py b/OCR_earsing/latent_diffusion/taming/modules/diffusionmodules/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/modules/diffusionmodules/model.py b/OCR_earsing/latent_diffusion/taming/modules/diffusionmodules/model.py
new file mode 100644
index 0000000..d3a5db6
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/diffusionmodules/model.py
@@ -0,0 +1,776 @@
+# pytorch_diffusion + derived encoder decoder
+import math
+import torch
+import torch.nn as nn
+import numpy as np
+
+
+def get_timestep_embedding(timesteps, embedding_dim):
+ """
+ This matches the implementation in Denoising Diffusion Probabilistic Models:
+ From Fairseq.
+ Build sinusoidal embeddings.
+ This matches the implementation in tensor2tensor, but differs slightly
+ from the description in Section 3.5 of "Attention Is All You Need".
+ """
+ assert len(timesteps.shape) == 1
+
+ half_dim = embedding_dim // 2
+ emb = math.log(10000) / (half_dim - 1)
+ emb = torch.exp(torch.arange(half_dim, dtype=torch.float32) * -emb)
+ emb = emb.to(device=timesteps.device)
+ emb = timesteps.float()[:, None] * emb[None, :]
+ emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
+ if embedding_dim % 2 == 1: # zero pad
+ emb = torch.nn.functional.pad(emb, (0,1,0,0))
+ return emb
+
+
+def nonlinearity(x):
+ # swish
+ return x*torch.sigmoid(x)
+
+
+def Normalize(in_channels):
+ return torch.nn.GroupNorm(num_groups=32, num_channels=in_channels, eps=1e-6, affine=True)
+
+
+class Upsample(nn.Module):
+ def __init__(self, in_channels, with_conv):
+ super().__init__()
+ self.with_conv = with_conv
+ if self.with_conv:
+ self.conv = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x):
+ x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest")
+ if self.with_conv:
+ x = self.conv(x)
+ return x
+
+
+class Downsample(nn.Module):
+ def __init__(self, in_channels, with_conv):
+ super().__init__()
+ self.with_conv = with_conv
+ if self.with_conv:
+ # no asymmetric padding in torch conv, must do it ourselves
+ self.conv = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=3,
+ stride=2,
+ padding=0)
+
+ def forward(self, x):
+ if self.with_conv:
+ pad = (0,1,0,1)
+ x = torch.nn.functional.pad(x, pad, mode="constant", value=0)
+ x = self.conv(x)
+ else:
+ x = torch.nn.functional.avg_pool2d(x, kernel_size=2, stride=2)
+ return x
+
+
+class ResnetBlock(nn.Module):
+ def __init__(self, *, in_channels, out_channels=None, conv_shortcut=False,
+ dropout, temb_channels=512):
+ super().__init__()
+ self.in_channels = in_channels
+ out_channels = in_channels if out_channels is None else out_channels
+ self.out_channels = out_channels
+ self.use_conv_shortcut = conv_shortcut
+
+ self.norm1 = Normalize(in_channels)
+ self.conv1 = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+ if temb_channels > 0:
+ self.temb_proj = torch.nn.Linear(temb_channels,
+ out_channels)
+ self.norm2 = Normalize(out_channels)
+ self.dropout = torch.nn.Dropout(dropout)
+ self.conv2 = torch.nn.Conv2d(out_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+ if self.in_channels != self.out_channels:
+ if self.use_conv_shortcut:
+ self.conv_shortcut = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+ else:
+ self.nin_shortcut = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+
+ def forward(self, x, temb):
+ h = x
+ h = self.norm1(h)
+ h = nonlinearity(h)
+ h = self.conv1(h)
+
+ if temb is not None:
+ h = h + self.temb_proj(nonlinearity(temb))[:,:,None,None]
+
+ h = self.norm2(h)
+ h = nonlinearity(h)
+ h = self.dropout(h)
+ h = self.conv2(h)
+
+ if self.in_channels != self.out_channels:
+ if self.use_conv_shortcut:
+ x = self.conv_shortcut(x)
+ else:
+ x = self.nin_shortcut(x)
+
+ return x+h
+
+
+class AttnBlock(nn.Module):
+ def __init__(self, in_channels):
+ super().__init__()
+ self.in_channels = in_channels
+
+ self.norm = Normalize(in_channels)
+ self.q = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.k = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.v = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ self.proj_out = torch.nn.Conv2d(in_channels,
+ in_channels,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+
+
+ def forward(self, x):
+ h_ = x
+ h_ = self.norm(h_)
+ q = self.q(h_)
+ k = self.k(h_)
+ v = self.v(h_)
+
+ # compute attention
+ b,c,h,w = q.shape
+ q = q.reshape(b,c,h*w)
+ q = q.permute(0,2,1) # b,hw,c
+ k = k.reshape(b,c,h*w) # b,c,hw
+ w_ = torch.bmm(q,k) # b,hw,hw w[b,i,j]=sum_c q[b,i,c]k[b,c,j]
+ w_ = w_ * (int(c)**(-0.5))
+ w_ = torch.nn.functional.softmax(w_, dim=2)
+
+ # attend to values
+ v = v.reshape(b,c,h*w)
+ w_ = w_.permute(0,2,1) # b,hw,hw (first hw of k, second of q)
+ h_ = torch.bmm(v,w_) # b, c,hw (hw of q) h_[b,c,j] = sum_i v[b,c,i] w_[b,i,j]
+ h_ = h_.reshape(b,c,h,w)
+
+ h_ = self.proj_out(h_)
+
+ return x+h_
+
+
+class Model(nn.Module):
+ def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+ resolution, use_timestep=True):
+ super().__init__()
+ self.ch = ch
+ self.temb_ch = self.ch*4
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ self.resolution = resolution
+ self.in_channels = in_channels
+
+ self.use_timestep = use_timestep
+ if self.use_timestep:
+ # timestep embedding
+ self.temb = nn.Module()
+ self.temb.dense = nn.ModuleList([
+ torch.nn.Linear(self.ch,
+ self.temb_ch),
+ torch.nn.Linear(self.temb_ch,
+ self.temb_ch),
+ ])
+
+ # downsampling
+ self.conv_in = torch.nn.Conv2d(in_channels,
+ self.ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ curr_res = resolution
+ in_ch_mult = (1,)+tuple(ch_mult)
+ self.down = nn.ModuleList()
+ for i_level in range(self.num_resolutions):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_in = ch*in_ch_mult[i_level]
+ block_out = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks):
+ block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(AttnBlock(block_in))
+ down = nn.Module()
+ down.block = block
+ down.attn = attn
+ if i_level != self.num_resolutions-1:
+ down.downsample = Downsample(block_in, resamp_with_conv)
+ curr_res = curr_res // 2
+ self.down.append(down)
+
+ # middle
+ self.mid = nn.Module()
+ self.mid.block_1 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+ self.mid.attn_1 = AttnBlock(block_in)
+ self.mid.block_2 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+
+ # upsampling
+ self.up = nn.ModuleList()
+ for i_level in reversed(range(self.num_resolutions)):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_out = ch*ch_mult[i_level]
+ skip_in = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks+1):
+ if i_block == self.num_res_blocks:
+ skip_in = ch*in_ch_mult[i_level]
+ block.append(ResnetBlock(in_channels=block_in+skip_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(AttnBlock(block_in))
+ up = nn.Module()
+ up.block = block
+ up.attn = attn
+ if i_level != 0:
+ up.upsample = Upsample(block_in, resamp_with_conv)
+ curr_res = curr_res * 2
+ self.up.insert(0, up) # prepend to get consistent order
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ out_ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+
+ def forward(self, x, t=None):
+ #assert x.shape[2] == x.shape[3] == self.resolution
+
+ if self.use_timestep:
+ # timestep embedding
+ assert t is not None
+ temb = get_timestep_embedding(t, self.ch)
+ temb = self.temb.dense[0](temb)
+ temb = nonlinearity(temb)
+ temb = self.temb.dense[1](temb)
+ else:
+ temb = None
+
+ # downsampling
+ hs = [self.conv_in(x)]
+ for i_level in range(self.num_resolutions):
+ for i_block in range(self.num_res_blocks):
+ h = self.down[i_level].block[i_block](hs[-1], temb)
+ if len(self.down[i_level].attn) > 0:
+ h = self.down[i_level].attn[i_block](h)
+ hs.append(h)
+ if i_level != self.num_resolutions-1:
+ hs.append(self.down[i_level].downsample(hs[-1]))
+
+ # middle
+ h = hs[-1]
+ h = self.mid.block_1(h, temb)
+ h = self.mid.attn_1(h)
+ h = self.mid.block_2(h, temb)
+
+ # upsampling
+ for i_level in reversed(range(self.num_resolutions)):
+ for i_block in range(self.num_res_blocks+1):
+ h = self.up[i_level].block[i_block](
+ torch.cat([h, hs.pop()], dim=1), temb)
+ if len(self.up[i_level].attn) > 0:
+ h = self.up[i_level].attn[i_block](h)
+ if i_level != 0:
+ h = self.up[i_level].upsample(h)
+
+ # end
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
+
+class Encoder(nn.Module):
+ def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+ resolution, z_channels, double_z=True, **ignore_kwargs):
+ super().__init__()
+ self.ch = ch
+ self.temb_ch = 0
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ self.resolution = resolution
+ self.in_channels = in_channels
+
+ # downsampling
+ self.conv_in = torch.nn.Conv2d(in_channels,
+ self.ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ curr_res = resolution
+ in_ch_mult = (1,)+tuple(ch_mult)
+ self.down = nn.ModuleList()
+ for i_level in range(self.num_resolutions):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_in = ch*in_ch_mult[i_level]
+ block_out = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks):
+ block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(AttnBlock(block_in))
+ down = nn.Module()
+ down.block = block
+ down.attn = attn
+ if i_level != self.num_resolutions-1:
+ down.downsample = Downsample(block_in, resamp_with_conv)
+ curr_res = curr_res // 2
+ self.down.append(down)
+
+ # middle
+ self.mid = nn.Module()
+ self.mid.block_1 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+ self.mid.attn_1 = AttnBlock(block_in)
+ self.mid.block_2 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ 2*z_channels if double_z else z_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+
+ def forward(self, x):
+ #assert x.shape[2] == x.shape[3] == self.resolution, "{}, {}, {}".format(x.shape[2], x.shape[3], self.resolution)
+
+ # timestep embedding
+ temb = None
+
+ # downsampling
+ hs = [self.conv_in(x)]
+ for i_level in range(self.num_resolutions):
+ for i_block in range(self.num_res_blocks):
+ h = self.down[i_level].block[i_block](hs[-1], temb)
+ if len(self.down[i_level].attn) > 0:
+ h = self.down[i_level].attn[i_block](h)
+ hs.append(h)
+ if i_level != self.num_resolutions-1:
+ hs.append(self.down[i_level].downsample(hs[-1]))
+
+ # middle
+ h = hs[-1]
+ h = self.mid.block_1(h, temb)
+ h = self.mid.attn_1(h)
+ h = self.mid.block_2(h, temb)
+
+ # end
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
+
+class Decoder(nn.Module):
+ def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+ resolution, z_channels, give_pre_end=False, **ignorekwargs):
+ super().__init__()
+ self.ch = ch
+ self.temb_ch = 0
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ self.resolution = resolution
+ self.in_channels = in_channels
+ self.give_pre_end = give_pre_end
+
+ # compute in_ch_mult, block_in and curr_res at lowest res
+ in_ch_mult = (1,)+tuple(ch_mult)
+ block_in = ch*ch_mult[self.num_resolutions-1]
+ curr_res = resolution // 2**(self.num_resolutions-1)
+ self.z_shape = (1,z_channels,curr_res,curr_res)
+ print("Working with z of shape {} = {} dimensions.".format(
+ self.z_shape, np.prod(self.z_shape)))
+
+ # z to block_in
+ self.conv_in = torch.nn.Conv2d(z_channels,
+ block_in,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ # middle
+ self.mid = nn.Module()
+ self.mid.block_1 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+ self.mid.attn_1 = AttnBlock(block_in)
+ self.mid.block_2 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+
+ # upsampling
+ self.up = nn.ModuleList()
+ for i_level in reversed(range(self.num_resolutions)):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_out = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks+1):
+ block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(AttnBlock(block_in))
+ up = nn.Module()
+ up.block = block
+ up.attn = attn
+ if i_level != 0:
+ up.upsample = Upsample(block_in, resamp_with_conv)
+ curr_res = curr_res * 2
+ self.up.insert(0, up) # prepend to get consistent order
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ out_ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, z):
+ #assert z.shape[1:] == self.z_shape[1:]
+ self.last_z_shape = z.shape
+
+ # timestep embedding
+ temb = None
+
+ # z to block_in
+ h = self.conv_in(z)
+
+ # middle
+ h = self.mid.block_1(h, temb)
+ h = self.mid.attn_1(h)
+ h = self.mid.block_2(h, temb)
+
+ # upsampling
+ for i_level in reversed(range(self.num_resolutions)):
+ for i_block in range(self.num_res_blocks+1):
+ h = self.up[i_level].block[i_block](h, temb)
+ if len(self.up[i_level].attn) > 0:
+ h = self.up[i_level].attn[i_block](h)
+ if i_level != 0:
+ h = self.up[i_level].upsample(h)
+
+ # end
+ if self.give_pre_end:
+ return h
+
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
+
+class VUNet(nn.Module):
+ def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+ attn_resolutions, dropout=0.0, resamp_with_conv=True,
+ in_channels, c_channels,
+ resolution, z_channels, use_timestep=False, **ignore_kwargs):
+ super().__init__()
+ self.ch = ch
+ self.temb_ch = self.ch*4
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ self.resolution = resolution
+
+ self.use_timestep = use_timestep
+ if self.use_timestep:
+ # timestep embedding
+ self.temb = nn.Module()
+ self.temb.dense = nn.ModuleList([
+ torch.nn.Linear(self.ch,
+ self.temb_ch),
+ torch.nn.Linear(self.temb_ch,
+ self.temb_ch),
+ ])
+
+ # downsampling
+ self.conv_in = torch.nn.Conv2d(c_channels,
+ self.ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ curr_res = resolution
+ in_ch_mult = (1,)+tuple(ch_mult)
+ self.down = nn.ModuleList()
+ for i_level in range(self.num_resolutions):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_in = ch*in_ch_mult[i_level]
+ block_out = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks):
+ block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(AttnBlock(block_in))
+ down = nn.Module()
+ down.block = block
+ down.attn = attn
+ if i_level != self.num_resolutions-1:
+ down.downsample = Downsample(block_in, resamp_with_conv)
+ curr_res = curr_res // 2
+ self.down.append(down)
+
+ self.z_in = torch.nn.Conv2d(z_channels,
+ block_in,
+ kernel_size=1,
+ stride=1,
+ padding=0)
+ # middle
+ self.mid = nn.Module()
+ self.mid.block_1 = ResnetBlock(in_channels=2*block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+ self.mid.attn_1 = AttnBlock(block_in)
+ self.mid.block_2 = ResnetBlock(in_channels=block_in,
+ out_channels=block_in,
+ temb_channels=self.temb_ch,
+ dropout=dropout)
+
+ # upsampling
+ self.up = nn.ModuleList()
+ for i_level in reversed(range(self.num_resolutions)):
+ block = nn.ModuleList()
+ attn = nn.ModuleList()
+ block_out = ch*ch_mult[i_level]
+ skip_in = ch*ch_mult[i_level]
+ for i_block in range(self.num_res_blocks+1):
+ if i_block == self.num_res_blocks:
+ skip_in = ch*in_ch_mult[i_level]
+ block.append(ResnetBlock(in_channels=block_in+skip_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ if curr_res in attn_resolutions:
+ attn.append(AttnBlock(block_in))
+ up = nn.Module()
+ up.block = block
+ up.attn = attn
+ if i_level != 0:
+ up.upsample = Upsample(block_in, resamp_with_conv)
+ curr_res = curr_res * 2
+ self.up.insert(0, up) # prepend to get consistent order
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ out_ch,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+
+ def forward(self, x, z):
+ #assert x.shape[2] == x.shape[3] == self.resolution
+
+ if self.use_timestep:
+ # timestep embedding
+ assert t is not None
+ temb = get_timestep_embedding(t, self.ch)
+ temb = self.temb.dense[0](temb)
+ temb = nonlinearity(temb)
+ temb = self.temb.dense[1](temb)
+ else:
+ temb = None
+
+ # downsampling
+ hs = [self.conv_in(x)]
+ for i_level in range(self.num_resolutions):
+ for i_block in range(self.num_res_blocks):
+ h = self.down[i_level].block[i_block](hs[-1], temb)
+ if len(self.down[i_level].attn) > 0:
+ h = self.down[i_level].attn[i_block](h)
+ hs.append(h)
+ if i_level != self.num_resolutions-1:
+ hs.append(self.down[i_level].downsample(hs[-1]))
+
+ # middle
+ h = hs[-1]
+ z = self.z_in(z)
+ h = torch.cat((h,z),dim=1)
+ h = self.mid.block_1(h, temb)
+ h = self.mid.attn_1(h)
+ h = self.mid.block_2(h, temb)
+
+ # upsampling
+ for i_level in reversed(range(self.num_resolutions)):
+ for i_block in range(self.num_res_blocks+1):
+ h = self.up[i_level].block[i_block](
+ torch.cat([h, hs.pop()], dim=1), temb)
+ if len(self.up[i_level].attn) > 0:
+ h = self.up[i_level].attn[i_block](h)
+ if i_level != 0:
+ h = self.up[i_level].upsample(h)
+
+ # end
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
+
+class SimpleDecoder(nn.Module):
+ def __init__(self, in_channels, out_channels, *args, **kwargs):
+ super().__init__()
+ self.model = nn.ModuleList([nn.Conv2d(in_channels, in_channels, 1),
+ ResnetBlock(in_channels=in_channels,
+ out_channels=2 * in_channels,
+ temb_channels=0, dropout=0.0),
+ ResnetBlock(in_channels=2 * in_channels,
+ out_channels=4 * in_channels,
+ temb_channels=0, dropout=0.0),
+ ResnetBlock(in_channels=4 * in_channels,
+ out_channels=2 * in_channels,
+ temb_channels=0, dropout=0.0),
+ nn.Conv2d(2*in_channels, in_channels, 1),
+ Upsample(in_channels, with_conv=True)])
+ # end
+ self.norm_out = Normalize(in_channels)
+ self.conv_out = torch.nn.Conv2d(in_channels,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x):
+ for i, layer in enumerate(self.model):
+ if i in [1,2,3]:
+ x = layer(x, None)
+ else:
+ x = layer(x)
+
+ h = self.norm_out(x)
+ h = nonlinearity(h)
+ x = self.conv_out(h)
+ return x
+
+
+class UpsampleDecoder(nn.Module):
+ def __init__(self, in_channels, out_channels, ch, num_res_blocks, resolution,
+ ch_mult=(2,2), dropout=0.0):
+ super().__init__()
+ # upsampling
+ self.temb_ch = 0
+ self.num_resolutions = len(ch_mult)
+ self.num_res_blocks = num_res_blocks
+ block_in = in_channels
+ curr_res = resolution // 2 ** (self.num_resolutions - 1)
+ self.res_blocks = nn.ModuleList()
+ self.upsample_blocks = nn.ModuleList()
+ for i_level in range(self.num_resolutions):
+ res_block = []
+ block_out = ch * ch_mult[i_level]
+ for i_block in range(self.num_res_blocks + 1):
+ res_block.append(ResnetBlock(in_channels=block_in,
+ out_channels=block_out,
+ temb_channels=self.temb_ch,
+ dropout=dropout))
+ block_in = block_out
+ self.res_blocks.append(nn.ModuleList(res_block))
+ if i_level != self.num_resolutions - 1:
+ self.upsample_blocks.append(Upsample(block_in, True))
+ curr_res = curr_res * 2
+
+ # end
+ self.norm_out = Normalize(block_in)
+ self.conv_out = torch.nn.Conv2d(block_in,
+ out_channels,
+ kernel_size=3,
+ stride=1,
+ padding=1)
+
+ def forward(self, x):
+ # upsampling
+ h = x
+ for k, i_level in enumerate(range(self.num_resolutions)):
+ for i_block in range(self.num_res_blocks + 1):
+ h = self.res_blocks[i_level][i_block](h, None)
+ if i_level != self.num_resolutions - 1:
+ h = self.upsample_blocks[k](h)
+ h = self.norm_out(h)
+ h = nonlinearity(h)
+ h = self.conv_out(h)
+ return h
+
diff --git a/OCR_earsing/latent_diffusion/taming/modules/discriminator/__init__.py b/OCR_earsing/latent_diffusion/taming/modules/discriminator/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/modules/discriminator/model.py b/OCR_earsing/latent_diffusion/taming/modules/discriminator/model.py
new file mode 100644
index 0000000..2aaa311
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/discriminator/model.py
@@ -0,0 +1,67 @@
+import functools
+import torch.nn as nn
+
+
+from taming.modules.util import ActNorm
+
+
+def weights_init(m):
+ classname = m.__class__.__name__
+ if classname.find('Conv') != -1:
+ nn.init.normal_(m.weight.data, 0.0, 0.02)
+ elif classname.find('BatchNorm') != -1:
+ nn.init.normal_(m.weight.data, 1.0, 0.02)
+ nn.init.constant_(m.bias.data, 0)
+
+
+class NLayerDiscriminator(nn.Module):
+ """Defines a PatchGAN discriminator as in Pix2Pix
+ --> see https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/blob/master/models/networks.py
+ """
+ def __init__(self, input_nc=3, ndf=64, n_layers=3, use_actnorm=False):
+ """Construct a PatchGAN discriminator
+ Parameters:
+ input_nc (int) -- the number of channels in input images
+ ndf (int) -- the number of filters in the last conv layer
+ n_layers (int) -- the number of conv layers in the discriminator
+ norm_layer -- normalization layer
+ """
+ super(NLayerDiscriminator, self).__init__()
+ if not use_actnorm:
+ norm_layer = nn.BatchNorm2d
+ else:
+ norm_layer = ActNorm
+ if type(norm_layer) == functools.partial: # no need to use bias as BatchNorm2d has affine parameters
+ use_bias = norm_layer.func != nn.BatchNorm2d
+ else:
+ use_bias = norm_layer != nn.BatchNorm2d
+
+ kw = 4
+ padw = 1
+ sequence = [nn.Conv2d(input_nc, ndf, kernel_size=kw, stride=2, padding=padw), nn.LeakyReLU(0.2, True)]
+ nf_mult = 1
+ nf_mult_prev = 1
+ for n in range(1, n_layers): # gradually increase the number of filters
+ nf_mult_prev = nf_mult
+ nf_mult = min(2 ** n, 8)
+ sequence += [
+ nn.Conv2d(ndf * nf_mult_prev, ndf * nf_mult, kernel_size=kw, stride=2, padding=padw, bias=use_bias),
+ norm_layer(ndf * nf_mult),
+ nn.LeakyReLU(0.2, True)
+ ]
+
+ nf_mult_prev = nf_mult
+ nf_mult = min(2 ** n_layers, 8)
+ sequence += [
+ nn.Conv2d(ndf * nf_mult_prev, ndf * nf_mult, kernel_size=kw, stride=1, padding=padw, bias=use_bias),
+ norm_layer(ndf * nf_mult),
+ nn.LeakyReLU(0.2, True)
+ ]
+
+ sequence += [
+ nn.Conv2d(ndf * nf_mult, 1, kernel_size=kw, stride=1, padding=padw)] # output 1 channel prediction map
+ self.main = nn.Sequential(*sequence)
+
+ def forward(self, input):
+ """Standard forward."""
+ return self.main(input)
diff --git a/OCR_earsing/latent_diffusion/taming/modules/losses/__init__.py b/OCR_earsing/latent_diffusion/taming/modules/losses/__init__.py
new file mode 100644
index 0000000..d09caf9
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/losses/__init__.py
@@ -0,0 +1,2 @@
+from taming.modules.losses.vqperceptual import DummyLoss
+
diff --git a/OCR_earsing/latent_diffusion/taming/modules/losses/lpips.py b/OCR_earsing/latent_diffusion/taming/modules/losses/lpips.py
new file mode 100644
index 0000000..a728044
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/losses/lpips.py
@@ -0,0 +1,123 @@
+"""Stripped version of https://github.com/richzhang/PerceptualSimilarity/tree/master/models"""
+
+import torch
+import torch.nn as nn
+from torchvision import models
+from collections import namedtuple
+
+from taming.util import get_ckpt_path
+
+
+class LPIPS(nn.Module):
+ # Learned perceptual metric
+ def __init__(self, use_dropout=True):
+ super().__init__()
+ self.scaling_layer = ScalingLayer()
+ self.chns = [64, 128, 256, 512, 512] # vg16 features
+ self.net = vgg16(pretrained=True, requires_grad=False)
+ self.lin0 = NetLinLayer(self.chns[0], use_dropout=use_dropout)
+ self.lin1 = NetLinLayer(self.chns[1], use_dropout=use_dropout)
+ self.lin2 = NetLinLayer(self.chns[2], use_dropout=use_dropout)
+ self.lin3 = NetLinLayer(self.chns[3], use_dropout=use_dropout)
+ self.lin4 = NetLinLayer(self.chns[4], use_dropout=use_dropout)
+ self.load_from_pretrained()
+ for param in self.parameters():
+ param.requires_grad = False
+
+ def load_from_pretrained(self, name="vgg_lpips"):
+ ckpt = get_ckpt_path(name, "taming/modules/autoencoder/lpips")
+ self.load_state_dict(torch.load(ckpt, map_location=torch.device("cpu")), strict=False)
+ print("loaded pretrained LPIPS loss from {}".format(ckpt))
+
+ @classmethod
+ def from_pretrained(cls, name="vgg_lpips"):
+ if name != "vgg_lpips":
+ raise NotImplementedError
+ model = cls()
+ ckpt = get_ckpt_path(name)
+ model.load_state_dict(torch.load(ckpt, map_location=torch.device("cpu")), strict=False)
+ return model
+
+ def forward(self, input, target):
+ in0_input, in1_input = (self.scaling_layer(input), self.scaling_layer(target))
+ outs0, outs1 = self.net(in0_input), self.net(in1_input)
+ feats0, feats1, diffs = {}, {}, {}
+ lins = [self.lin0, self.lin1, self.lin2, self.lin3, self.lin4]
+ for kk in range(len(self.chns)):
+ feats0[kk], feats1[kk] = normalize_tensor(outs0[kk]), normalize_tensor(outs1[kk])
+ diffs[kk] = (feats0[kk] - feats1[kk]) ** 2
+
+ res = [spatial_average(lins[kk].model(diffs[kk]), keepdim=True) for kk in range(len(self.chns))]
+ val = res[0]
+ for l in range(1, len(self.chns)):
+ val += res[l]
+ return val
+
+
+class ScalingLayer(nn.Module):
+ def __init__(self):
+ super(ScalingLayer, self).__init__()
+ self.register_buffer('shift', torch.Tensor([-.030, -.088, -.188])[None, :, None, None])
+ self.register_buffer('scale', torch.Tensor([.458, .448, .450])[None, :, None, None])
+
+ def forward(self, inp):
+ return (inp - self.shift) / self.scale
+
+
+class NetLinLayer(nn.Module):
+ """ A single linear layer which does a 1x1 conv """
+ def __init__(self, chn_in, chn_out=1, use_dropout=False):
+ super(NetLinLayer, self).__init__()
+ layers = [nn.Dropout(), ] if (use_dropout) else []
+ layers += [nn.Conv2d(chn_in, chn_out, 1, stride=1, padding=0, bias=False), ]
+ self.model = nn.Sequential(*layers)
+
+
+class vgg16(torch.nn.Module):
+ def __init__(self, requires_grad=False, pretrained=True):
+ super(vgg16, self).__init__()
+ vgg_pretrained_features = models.vgg16(pretrained=pretrained).features
+ self.slice1 = torch.nn.Sequential()
+ self.slice2 = torch.nn.Sequential()
+ self.slice3 = torch.nn.Sequential()
+ self.slice4 = torch.nn.Sequential()
+ self.slice5 = torch.nn.Sequential()
+ self.N_slices = 5
+ for x in range(4):
+ self.slice1.add_module(str(x), vgg_pretrained_features[x])
+ for x in range(4, 9):
+ self.slice2.add_module(str(x), vgg_pretrained_features[x])
+ for x in range(9, 16):
+ self.slice3.add_module(str(x), vgg_pretrained_features[x])
+ for x in range(16, 23):
+ self.slice4.add_module(str(x), vgg_pretrained_features[x])
+ for x in range(23, 30):
+ self.slice5.add_module(str(x), vgg_pretrained_features[x])
+ if not requires_grad:
+ for param in self.parameters():
+ param.requires_grad = False
+
+ def forward(self, X):
+ h = self.slice1(X)
+ h_relu1_2 = h
+ h = self.slice2(h)
+ h_relu2_2 = h
+ h = self.slice3(h)
+ h_relu3_3 = h
+ h = self.slice4(h)
+ h_relu4_3 = h
+ h = self.slice5(h)
+ h_relu5_3 = h
+ vgg_outputs = namedtuple("VggOutputs", ['relu1_2', 'relu2_2', 'relu3_3', 'relu4_3', 'relu5_3'])
+ out = vgg_outputs(h_relu1_2, h_relu2_2, h_relu3_3, h_relu4_3, h_relu5_3)
+ return out
+
+
+def normalize_tensor(x,eps=1e-10):
+ norm_factor = torch.sqrt(torch.sum(x**2,dim=1,keepdim=True))
+ return x/(norm_factor+eps)
+
+
+def spatial_average(x, keepdim=True):
+ return x.mean([2,3],keepdim=keepdim)
+
diff --git a/OCR_earsing/latent_diffusion/taming/modules/losses/segmentation.py b/OCR_earsing/latent_diffusion/taming/modules/losses/segmentation.py
new file mode 100644
index 0000000..4ba77de
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/losses/segmentation.py
@@ -0,0 +1,22 @@
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class BCELoss(nn.Module):
+ def forward(self, prediction, target):
+ loss = F.binary_cross_entropy_with_logits(prediction,target)
+ return loss, {}
+
+
+class BCELossWithQuant(nn.Module):
+ def __init__(self, codebook_weight=1.):
+ super().__init__()
+ self.codebook_weight = codebook_weight
+
+ def forward(self, qloss, target, prediction, split):
+ bce_loss = F.binary_cross_entropy_with_logits(prediction,target)
+ loss = bce_loss + self.codebook_weight*qloss
+ return loss, {"{}/total_loss".format(split): loss.clone().detach().mean(),
+ "{}/bce_loss".format(split): bce_loss.detach().mean(),
+ "{}/quant_loss".format(split): qloss.detach().mean()
+ }
diff --git a/OCR_earsing/latent_diffusion/taming/modules/losses/vqperceptual.py b/OCR_earsing/latent_diffusion/taming/modules/losses/vqperceptual.py
new file mode 100644
index 0000000..c2febd4
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/losses/vqperceptual.py
@@ -0,0 +1,136 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from taming.modules.losses.lpips import LPIPS
+from taming.modules.discriminator.model import NLayerDiscriminator, weights_init
+
+
+class DummyLoss(nn.Module):
+ def __init__(self):
+ super().__init__()
+
+
+def adopt_weight(weight, global_step, threshold=0, value=0.):
+ if global_step < threshold:
+ weight = value
+ return weight
+
+
+def hinge_d_loss(logits_real, logits_fake):
+ loss_real = torch.mean(F.relu(1. - logits_real))
+ loss_fake = torch.mean(F.relu(1. + logits_fake))
+ d_loss = 0.5 * (loss_real + loss_fake)
+ return d_loss
+
+
+def vanilla_d_loss(logits_real, logits_fake):
+ d_loss = 0.5 * (
+ torch.mean(torch.nn.functional.softplus(-logits_real)) +
+ torch.mean(torch.nn.functional.softplus(logits_fake)))
+ return d_loss
+
+
+class VQLPIPSWithDiscriminator(nn.Module):
+ def __init__(self, disc_start, codebook_weight=1.0, pixelloss_weight=1.0,
+ disc_num_layers=3, disc_in_channels=3, disc_factor=1.0, disc_weight=1.0,
+ perceptual_weight=1.0, use_actnorm=False, disc_conditional=False,
+ disc_ndf=64, disc_loss="hinge"):
+ super().__init__()
+ assert disc_loss in ["hinge", "vanilla"]
+ self.codebook_weight = codebook_weight
+ self.pixel_weight = pixelloss_weight
+ self.perceptual_loss = LPIPS().eval()
+ self.perceptual_weight = perceptual_weight
+
+ self.discriminator = NLayerDiscriminator(input_nc=disc_in_channels,
+ n_layers=disc_num_layers,
+ use_actnorm=use_actnorm,
+ ndf=disc_ndf
+ ).apply(weights_init)
+ self.discriminator_iter_start = disc_start
+ if disc_loss == "hinge":
+ self.disc_loss = hinge_d_loss
+ elif disc_loss == "vanilla":
+ self.disc_loss = vanilla_d_loss
+ else:
+ raise ValueError(f"Unknown GAN loss '{disc_loss}'.")
+ print(f"VQLPIPSWithDiscriminator running with {disc_loss} loss.")
+ self.disc_factor = disc_factor
+ self.discriminator_weight = disc_weight
+ self.disc_conditional = disc_conditional
+
+ def calculate_adaptive_weight(self, nll_loss, g_loss, last_layer=None):
+ if last_layer is not None:
+ nll_grads = torch.autograd.grad(nll_loss, last_layer, retain_graph=True)[0]
+ g_grads = torch.autograd.grad(g_loss, last_layer, retain_graph=True)[0]
+ else:
+ nll_grads = torch.autograd.grad(nll_loss, self.last_layer[0], retain_graph=True)[0]
+ g_grads = torch.autograd.grad(g_loss, self.last_layer[0], retain_graph=True)[0]
+
+ d_weight = torch.norm(nll_grads) / (torch.norm(g_grads) + 1e-4)
+ d_weight = torch.clamp(d_weight, 0.0, 1e4).detach()
+ d_weight = d_weight * self.discriminator_weight
+ return d_weight
+
+ def forward(self, codebook_loss, inputs, reconstructions, optimizer_idx,
+ global_step, last_layer=None, cond=None, split="train"):
+ rec_loss = torch.abs(inputs.contiguous() - reconstructions.contiguous())
+ if self.perceptual_weight > 0:
+ p_loss = self.perceptual_loss(inputs.contiguous(), reconstructions.contiguous())
+ rec_loss = rec_loss + self.perceptual_weight * p_loss
+ else:
+ p_loss = torch.tensor([0.0])
+
+ nll_loss = rec_loss
+ #nll_loss = torch.sum(nll_loss) / nll_loss.shape[0]
+ nll_loss = torch.mean(nll_loss)
+
+ # now the GAN part
+ if optimizer_idx == 0:
+ # generator update
+ if cond is None:
+ assert not self.disc_conditional
+ logits_fake = self.discriminator(reconstructions.contiguous())
+ else:
+ assert self.disc_conditional
+ logits_fake = self.discriminator(torch.cat((reconstructions.contiguous(), cond), dim=1))
+ g_loss = -torch.mean(logits_fake)
+
+ try:
+ d_weight = self.calculate_adaptive_weight(nll_loss, g_loss, last_layer=last_layer)
+ except RuntimeError:
+ assert not self.training
+ d_weight = torch.tensor(0.0)
+
+ disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+ loss = nll_loss + d_weight * disc_factor * g_loss + self.codebook_weight * codebook_loss.mean()
+
+ log = {"{}/total_loss".format(split): loss.clone().detach().mean(),
+ "{}/quant_loss".format(split): codebook_loss.detach().mean(),
+ "{}/nll_loss".format(split): nll_loss.detach().mean(),
+ "{}/rec_loss".format(split): rec_loss.detach().mean(),
+ "{}/p_loss".format(split): p_loss.detach().mean(),
+ "{}/d_weight".format(split): d_weight.detach(),
+ "{}/disc_factor".format(split): torch.tensor(disc_factor),
+ "{}/g_loss".format(split): g_loss.detach().mean(),
+ }
+ return loss, log
+
+ if optimizer_idx == 1:
+ # second pass for discriminator update
+ if cond is None:
+ logits_real = self.discriminator(inputs.contiguous().detach())
+ logits_fake = self.discriminator(reconstructions.contiguous().detach())
+ else:
+ logits_real = self.discriminator(torch.cat((inputs.contiguous().detach(), cond), dim=1))
+ logits_fake = self.discriminator(torch.cat((reconstructions.contiguous().detach(), cond), dim=1))
+
+ disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+ d_loss = disc_factor * self.disc_loss(logits_real, logits_fake)
+
+ log = {"{}/disc_loss".format(split): d_loss.clone().detach().mean(),
+ "{}/logits_real".format(split): logits_real.detach().mean(),
+ "{}/logits_fake".format(split): logits_fake.detach().mean()
+ }
+ return d_loss, log
diff --git a/OCR_earsing/latent_diffusion/taming/modules/misc/__init__.py b/OCR_earsing/latent_diffusion/taming/modules/misc/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/modules/misc/coord.py b/OCR_earsing/latent_diffusion/taming/modules/misc/coord.py
new file mode 100644
index 0000000..ee69b0c
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/misc/coord.py
@@ -0,0 +1,31 @@
+import torch
+
+class CoordStage(object):
+ def __init__(self, n_embed, down_factor):
+ self.n_embed = n_embed
+ self.down_factor = down_factor
+
+ def eval(self):
+ return self
+
+ def encode(self, c):
+ """fake vqmodel interface"""
+ assert 0.0 <= c.min() and c.max() <= 1.0
+ b,ch,h,w = c.shape
+ assert ch == 1
+
+ c = torch.nn.functional.interpolate(c, scale_factor=1/self.down_factor,
+ mode="area")
+ c = c.clamp(0.0, 1.0)
+ c = self.n_embed*c
+ c_quant = c.round()
+ c_ind = c_quant.to(dtype=torch.long)
+
+ info = None, None, c_ind
+ return c_quant, None, info
+
+ def decode(self, c):
+ c = c/self.n_embed
+ c = torch.nn.functional.interpolate(c, scale_factor=self.down_factor,
+ mode="nearest")
+ return c
diff --git a/OCR_earsing/latent_diffusion/taming/modules/transformer/__init__.py b/OCR_earsing/latent_diffusion/taming/modules/transformer/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/modules/transformer/mingpt.py b/OCR_earsing/latent_diffusion/taming/modules/transformer/mingpt.py
new file mode 100644
index 0000000..d14b7b6
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/transformer/mingpt.py
@@ -0,0 +1,415 @@
+"""
+taken from: https://github.com/karpathy/minGPT/
+GPT model:
+- the initial stem consists of a combination of token encoding and a positional encoding
+- the meat of it is a uniform sequence of Transformer blocks
+ - each Transformer is a sequential combination of a 1-hidden-layer MLP block and a self-attention block
+ - all blocks feed into a central residual pathway similar to resnets
+- the final decoder is a linear projection into a vanilla Softmax classifier
+"""
+
+import math
+import logging
+
+import torch
+import torch.nn as nn
+from torch.nn import functional as F
+from transformers import top_k_top_p_filtering
+
+logger = logging.getLogger(__name__)
+
+
+class GPTConfig:
+ """ base GPT config, params common to all GPT versions """
+ embd_pdrop = 0.1
+ resid_pdrop = 0.1
+ attn_pdrop = 0.1
+
+ def __init__(self, vocab_size, block_size, **kwargs):
+ self.vocab_size = vocab_size
+ self.block_size = block_size
+ for k,v in kwargs.items():
+ setattr(self, k, v)
+
+
+class GPT1Config(GPTConfig):
+ """ GPT-1 like network roughly 125M params """
+ n_layer = 12
+ n_head = 12
+ n_embd = 768
+
+
+class CausalSelfAttention(nn.Module):
+ """
+ A vanilla multi-head masked self-attention layer with a projection at the end.
+ It is possible to use torch.nn.MultiheadAttention here but I am including an
+ explicit implementation here to show that there is nothing too scary here.
+ """
+
+ def __init__(self, config):
+ super().__init__()
+ assert config.n_embd % config.n_head == 0
+ # key, query, value projections for all heads
+ self.key = nn.Linear(config.n_embd, config.n_embd)
+ self.query = nn.Linear(config.n_embd, config.n_embd)
+ self.value = nn.Linear(config.n_embd, config.n_embd)
+ # regularization
+ self.attn_drop = nn.Dropout(config.attn_pdrop)
+ self.resid_drop = nn.Dropout(config.resid_pdrop)
+ # output projection
+ self.proj = nn.Linear(config.n_embd, config.n_embd)
+ # causal mask to ensure that attention is only applied to the left in the input sequence
+ mask = torch.tril(torch.ones(config.block_size,
+ config.block_size))
+ if hasattr(config, "n_unmasked"):
+ mask[:config.n_unmasked, :config.n_unmasked] = 1
+ self.register_buffer("mask", mask.view(1, 1, config.block_size, config.block_size))
+ self.n_head = config.n_head
+
+ def forward(self, x, layer_past=None):
+ B, T, C = x.size()
+
+ # calculate query, key, values for all heads in batch and move head forward to be the batch dim
+ k = self.key(x).view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs)
+ q = self.query(x).view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs)
+ v = self.value(x).view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs)
+
+ present = torch.stack((k, v))
+ if layer_past is not None:
+ past_key, past_value = layer_past
+ k = torch.cat((past_key, k), dim=-2)
+ v = torch.cat((past_value, v), dim=-2)
+
+ # causal self-attention; Self-attend: (B, nh, T, hs) x (B, nh, hs, T) -> (B, nh, T, T)
+ att = (q @ k.transpose(-2, -1)) * (1.0 / math.sqrt(k.size(-1)))
+ if layer_past is None:
+ att = att.masked_fill(self.mask[:,:,:T,:T] == 0, float('-inf'))
+
+ att = F.softmax(att, dim=-1)
+ att = self.attn_drop(att)
+ y = att @ v # (B, nh, T, T) x (B, nh, T, hs) -> (B, nh, T, hs)
+ y = y.transpose(1, 2).contiguous().view(B, T, C) # re-assemble all head outputs side by side
+
+ # output projection
+ y = self.resid_drop(self.proj(y))
+ return y, present # TODO: check that this does not break anything
+
+
+class Block(nn.Module):
+ """ an unassuming Transformer block """
+ def __init__(self, config):
+ super().__init__()
+ self.ln1 = nn.LayerNorm(config.n_embd)
+ self.ln2 = nn.LayerNorm(config.n_embd)
+ self.attn = CausalSelfAttention(config)
+ self.mlp = nn.Sequential(
+ nn.Linear(config.n_embd, 4 * config.n_embd),
+ nn.GELU(), # nice
+ nn.Linear(4 * config.n_embd, config.n_embd),
+ nn.Dropout(config.resid_pdrop),
+ )
+
+ def forward(self, x, layer_past=None, return_present=False):
+ # TODO: check that training still works
+ if return_present: assert not self.training
+ # layer past: tuple of length two with B, nh, T, hs
+ attn, present = self.attn(self.ln1(x), layer_past=layer_past)
+
+ x = x + attn
+ x = x + self.mlp(self.ln2(x))
+ if layer_past is not None or return_present:
+ return x, present
+ return x
+
+
+class GPT(nn.Module):
+ """ the full GPT language model, with a context size of block_size """
+ def __init__(self, vocab_size, block_size, n_layer=12, n_head=8, n_embd=256,
+ embd_pdrop=0., resid_pdrop=0., attn_pdrop=0., n_unmasked=0):
+ super().__init__()
+ config = GPTConfig(vocab_size=vocab_size, block_size=block_size,
+ embd_pdrop=embd_pdrop, resid_pdrop=resid_pdrop, attn_pdrop=attn_pdrop,
+ n_layer=n_layer, n_head=n_head, n_embd=n_embd,
+ n_unmasked=n_unmasked)
+ # input embedding stem
+ self.tok_emb = nn.Embedding(config.vocab_size, config.n_embd)
+ self.pos_emb = nn.Parameter(torch.zeros(1, config.block_size, config.n_embd))
+ self.drop = nn.Dropout(config.embd_pdrop)
+ # transformer
+ self.blocks = nn.Sequential(*[Block(config) for _ in range(config.n_layer)])
+ # decoder head
+ self.ln_f = nn.LayerNorm(config.n_embd)
+ self.head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
+ self.block_size = config.block_size
+ self.apply(self._init_weights)
+ self.config = config
+ logger.info("number of parameters: %e", sum(p.numel() for p in self.parameters()))
+
+ def get_block_size(self):
+ return self.block_size
+
+ def _init_weights(self, module):
+ if isinstance(module, (nn.Linear, nn.Embedding)):
+ module.weight.data.normal_(mean=0.0, std=0.02)
+ if isinstance(module, nn.Linear) and module.bias is not None:
+ module.bias.data.zero_()
+ elif isinstance(module, nn.LayerNorm):
+ module.bias.data.zero_()
+ module.weight.data.fill_(1.0)
+
+ def forward(self, idx, embeddings=None, targets=None):
+ # forward the GPT model
+ token_embeddings = self.tok_emb(idx) # each index maps to a (learnable) vector
+
+ if embeddings is not None: # prepend explicit embeddings
+ token_embeddings = torch.cat((embeddings, token_embeddings), dim=1)
+
+ t = token_embeddings.shape[1]
+ assert t <= self.block_size, "Cannot forward, model block size is exhausted."
+ position_embeddings = self.pos_emb[:, :t, :] # each position maps to a (learnable) vector
+ x = self.drop(token_embeddings + position_embeddings)
+ x = self.blocks(x)
+ x = self.ln_f(x)
+ logits = self.head(x)
+
+ # if we are given some desired targets also calculate the loss
+ loss = None
+ if targets is not None:
+ loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1))
+
+ return logits, loss
+
+ def forward_with_past(self, idx, embeddings=None, targets=None, past=None, past_length=None):
+ # inference only
+ assert not self.training
+ token_embeddings = self.tok_emb(idx) # each index maps to a (learnable) vector
+ if embeddings is not None: # prepend explicit embeddings
+ token_embeddings = torch.cat((embeddings, token_embeddings), dim=1)
+
+ if past is not None:
+ assert past_length is not None
+ past = torch.cat(past, dim=-2) # n_layer, 2, b, nh, len_past, dim_head
+ past_shape = list(past.shape)
+ expected_shape = [self.config.n_layer, 2, idx.shape[0], self.config.n_head, past_length, self.config.n_embd//self.config.n_head]
+ assert past_shape == expected_shape, f"{past_shape} =/= {expected_shape}"
+ position_embeddings = self.pos_emb[:, past_length, :] # each position maps to a (learnable) vector
+ else:
+ position_embeddings = self.pos_emb[:, :token_embeddings.shape[1], :]
+
+ x = self.drop(token_embeddings + position_embeddings)
+ presents = [] # accumulate over layers
+ for i, block in enumerate(self.blocks):
+ x, present = block(x, layer_past=past[i, ...] if past is not None else None, return_present=True)
+ presents.append(present)
+
+ x = self.ln_f(x)
+ logits = self.head(x)
+ # if we are given some desired targets also calculate the loss
+ loss = None
+ if targets is not None:
+ loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1))
+
+ return logits, loss, torch.stack(presents) # _, _, n_layer, 2, b, nh, 1, dim_head
+
+
+class DummyGPT(nn.Module):
+ # for debugging
+ def __init__(self, add_value=1):
+ super().__init__()
+ self.add_value = add_value
+
+ def forward(self, idx):
+ return idx + self.add_value, None
+
+
+class CodeGPT(nn.Module):
+ """Takes in semi-embeddings"""
+ def __init__(self, vocab_size, block_size, in_channels, n_layer=12, n_head=8, n_embd=256,
+ embd_pdrop=0., resid_pdrop=0., attn_pdrop=0., n_unmasked=0):
+ super().__init__()
+ config = GPTConfig(vocab_size=vocab_size, block_size=block_size,
+ embd_pdrop=embd_pdrop, resid_pdrop=resid_pdrop, attn_pdrop=attn_pdrop,
+ n_layer=n_layer, n_head=n_head, n_embd=n_embd,
+ n_unmasked=n_unmasked)
+ # input embedding stem
+ self.tok_emb = nn.Linear(in_channels, config.n_embd)
+ self.pos_emb = nn.Parameter(torch.zeros(1, config.block_size, config.n_embd))
+ self.drop = nn.Dropout(config.embd_pdrop)
+ # transformer
+ self.blocks = nn.Sequential(*[Block(config) for _ in range(config.n_layer)])
+ # decoder head
+ self.ln_f = nn.LayerNorm(config.n_embd)
+ self.head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
+ self.block_size = config.block_size
+ self.apply(self._init_weights)
+ self.config = config
+ logger.info("number of parameters: %e", sum(p.numel() for p in self.parameters()))
+
+ def get_block_size(self):
+ return self.block_size
+
+ def _init_weights(self, module):
+ if isinstance(module, (nn.Linear, nn.Embedding)):
+ module.weight.data.normal_(mean=0.0, std=0.02)
+ if isinstance(module, nn.Linear) and module.bias is not None:
+ module.bias.data.zero_()
+ elif isinstance(module, nn.LayerNorm):
+ module.bias.data.zero_()
+ module.weight.data.fill_(1.0)
+
+ def forward(self, idx, embeddings=None, targets=None):
+ # forward the GPT model
+ token_embeddings = self.tok_emb(idx) # each index maps to a (learnable) vector
+
+ if embeddings is not None: # prepend explicit embeddings
+ token_embeddings = torch.cat((embeddings, token_embeddings), dim=1)
+
+ t = token_embeddings.shape[1]
+ assert t <= self.block_size, "Cannot forward, model block size is exhausted."
+ position_embeddings = self.pos_emb[:, :t, :] # each position maps to a (learnable) vector
+ x = self.drop(token_embeddings + position_embeddings)
+ x = self.blocks(x)
+ x = self.taming_cinln_f(x)
+ logits = self.head(x)
+
+ # if we are given some desired targets also calculate the loss
+ loss = None
+ if targets is not None:
+ loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1))
+
+ return logits, loss
+
+
+
+#### sampling utils
+
+def top_k_logits(logits, k):
+ v, ix = torch.topk(logits, k)
+ out = logits.clone()
+ out[out < v[:, [-1]]] = -float('Inf')
+ return out
+
+@torch.no_grad()
+def sample(model, x, steps, temperature=1.0, sample=False, top_k=None):
+ """
+ take a conditioning sequence of indices in x (of shape (b,t)) and predict the next token in
+ the sequence, feeding the predictions back into the model each time. Clearly the sampling
+ has quadratic complexity unlike an RNN that is only linear, and has a finite context window
+ of block_size, unlike an RNN that has an infinite context window.
+ """
+ block_size = model.get_block_size()
+ model.eval()
+ for k in range(steps):
+ x_cond = x if x.size(1) <= block_size else x[:, -block_size:] # crop context if needed
+ logits, _ = model(x_cond)
+ # pluck the logits at the final step and scale by temperature
+ logits = logits[:, -1, :] / temperature
+ # optionally crop probabilities to only the top k options
+ if top_k is not None:
+ logits = top_k_logits(logits, top_k)
+ # apply softmax to convert to probabilities
+ probs = F.softmax(logits, dim=-1)
+ # sample from the distribution or take the most likely
+ if sample:
+ ix = torch.multinomial(probs, num_samples=1)
+ else:
+ _, ix = torch.topk(probs, k=1, dim=-1)
+ # append to the sequence and continue
+ x = torch.cat((x, ix), dim=1)
+
+ return x
+
+
+@torch.no_grad()
+def sample_with_past(x, model, steps, temperature=1., sample_logits=True,
+ top_k=None, top_p=None, callback=None):
+ # x is conditioning
+ sample = x
+ cond_len = x.shape[1]
+ past = None
+ for n in range(steps):
+ if callback is not None:
+ callback(n)
+ logits, _, present = model.forward_with_past(x, past=past, past_length=(n+cond_len-1))
+ if past is None:
+ past = [present]
+ else:
+ past.append(present)
+ logits = logits[:, -1, :] / temperature
+ if top_k is not None:
+ logits = top_k_top_p_filtering(logits, top_k=top_k, top_p=top_p)
+
+ probs = F.softmax(logits, dim=-1)
+ if not sample_logits:
+ _, x = torch.topk(probs, k=1, dim=-1)
+ else:
+ x = torch.multinomial(probs, num_samples=1)
+ # append to the sequence and continue
+ sample = torch.cat((sample, x), dim=1)
+ del past
+ sample = sample[:, cond_len:] # cut conditioning off
+ return sample
+
+
+#### clustering utils
+
+class KMeans(nn.Module):
+ def __init__(self, ncluster=512, nc=3, niter=10):
+ super().__init__()
+ self.ncluster = ncluster
+ self.nc = nc
+ self.niter = niter
+ self.shape = (3,32,32)
+ self.register_buffer("C", torch.zeros(self.ncluster,nc))
+ self.register_buffer('initialized', torch.tensor(0, dtype=torch.uint8))
+
+ def is_initialized(self):
+ return self.initialized.item() == 1
+
+ @torch.no_grad()
+ def initialize(self, x):
+ N, D = x.shape
+ assert D == self.nc, D
+ c = x[torch.randperm(N)[:self.ncluster]] # init clusters at random
+ for i in range(self.niter):
+ # assign all pixels to the closest codebook element
+ a = ((x[:, None, :] - c[None, :, :])**2).sum(-1).argmin(1)
+ # move each codebook element to be the mean of the pixels that assigned to it
+ c = torch.stack([x[a==k].mean(0) for k in range(self.ncluster)])
+ # re-assign any poorly positioned codebook elements
+ nanix = torch.any(torch.isnan(c), dim=1)
+ ndead = nanix.sum().item()
+ print('done step %d/%d, re-initialized %d dead clusters' % (i+1, self.niter, ndead))
+ c[nanix] = x[torch.randperm(N)[:ndead]] # re-init dead clusters
+
+ self.C.copy_(c)
+ self.initialized.fill_(1)
+
+
+ def forward(self, x, reverse=False, shape=None):
+ if not reverse:
+ # flatten
+ bs,c,h,w = x.shape
+ assert c == self.nc
+ x = x.reshape(bs,c,h*w,1)
+ C = self.C.permute(1,0)
+ C = C.reshape(1,c,1,self.ncluster)
+ a = ((x-C)**2).sum(1).argmin(-1) # bs, h*w indices
+ return a
+ else:
+ # flatten
+ bs, HW = x.shape
+ """
+ c = self.C.reshape( 1, self.nc, 1, self.ncluster)
+ c = c[bs*[0],:,:,:]
+ c = c[:,:,HW*[0],:]
+ x = x.reshape(bs, 1, HW, 1)
+ x = x[:,3*[0],:,:]
+ x = torch.gather(c, dim=3, index=x)
+ """
+ x = self.C[x]
+ x = x.permute(0,2,1)
+ shape = shape if shape is not None else self.shape
+ x = x.reshape(bs, *shape)
+
+ return x
diff --git a/OCR_earsing/latent_diffusion/taming/modules/transformer/permuter.py b/OCR_earsing/latent_diffusion/taming/modules/transformer/permuter.py
new file mode 100644
index 0000000..0d43bb1
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/transformer/permuter.py
@@ -0,0 +1,248 @@
+import torch
+import torch.nn as nn
+import numpy as np
+
+
+class AbstractPermuter(nn.Module):
+ def __init__(self, *args, **kwargs):
+ super().__init__()
+ def forward(self, x, reverse=False):
+ raise NotImplementedError
+
+
+class Identity(AbstractPermuter):
+ def __init__(self):
+ super().__init__()
+
+ def forward(self, x, reverse=False):
+ return x
+
+
+class Subsample(AbstractPermuter):
+ def __init__(self, H, W):
+ super().__init__()
+ C = 1
+ indices = np.arange(H*W).reshape(C,H,W)
+ while min(H, W) > 1:
+ indices = indices.reshape(C,H//2,2,W//2,2)
+ indices = indices.transpose(0,2,4,1,3)
+ indices = indices.reshape(C*4,H//2, W//2)
+ H = H//2
+ W = W//2
+ C = C*4
+ assert H == W == 1
+ idx = torch.tensor(indices.ravel())
+ self.register_buffer('forward_shuffle_idx',
+ nn.Parameter(idx, requires_grad=False))
+ self.register_buffer('backward_shuffle_idx',
+ nn.Parameter(torch.argsort(idx), requires_grad=False))
+
+ def forward(self, x, reverse=False):
+ if not reverse:
+ return x[:, self.forward_shuffle_idx]
+ else:
+ return x[:, self.backward_shuffle_idx]
+
+
+def mortonify(i, j):
+ """(i,j) index to linear morton code"""
+ i = np.uint64(i)
+ j = np.uint64(j)
+
+ z = np.uint(0)
+
+ for pos in range(32):
+ z = (z |
+ ((j & (np.uint64(1) << np.uint64(pos))) << np.uint64(pos)) |
+ ((i & (np.uint64(1) << np.uint64(pos))) << np.uint64(pos+1))
+ )
+ return z
+
+
+class ZCurve(AbstractPermuter):
+ def __init__(self, H, W):
+ super().__init__()
+ reverseidx = [np.int64(mortonify(i,j)) for i in range(H) for j in range(W)]
+ idx = np.argsort(reverseidx)
+ idx = torch.tensor(idx)
+ reverseidx = torch.tensor(reverseidx)
+ self.register_buffer('forward_shuffle_idx',
+ idx)
+ self.register_buffer('backward_shuffle_idx',
+ reverseidx)
+
+ def forward(self, x, reverse=False):
+ if not reverse:
+ return x[:, self.forward_shuffle_idx]
+ else:
+ return x[:, self.backward_shuffle_idx]
+
+
+class SpiralOut(AbstractPermuter):
+ def __init__(self, H, W):
+ super().__init__()
+ assert H == W
+ size = W
+ indices = np.arange(size*size).reshape(size,size)
+
+ i0 = size//2
+ j0 = size//2-1
+
+ i = i0
+ j = j0
+
+ idx = [indices[i0, j0]]
+ step_mult = 0
+ for c in range(1, size//2+1):
+ step_mult += 1
+ # steps left
+ for k in range(step_mult):
+ i = i - 1
+ j = j
+ idx.append(indices[i, j])
+
+ # step down
+ for k in range(step_mult):
+ i = i
+ j = j + 1
+ idx.append(indices[i, j])
+
+ step_mult += 1
+ if c < size//2:
+ # step right
+ for k in range(step_mult):
+ i = i + 1
+ j = j
+ idx.append(indices[i, j])
+
+ # step up
+ for k in range(step_mult):
+ i = i
+ j = j - 1
+ idx.append(indices[i, j])
+ else:
+ # end reached
+ for k in range(step_mult-1):
+ i = i + 1
+ idx.append(indices[i, j])
+
+ assert len(idx) == size*size
+ idx = torch.tensor(idx)
+ self.register_buffer('forward_shuffle_idx', idx)
+ self.register_buffer('backward_shuffle_idx', torch.argsort(idx))
+
+ def forward(self, x, reverse=False):
+ if not reverse:
+ return x[:, self.forward_shuffle_idx]
+ else:
+ return x[:, self.backward_shuffle_idx]
+
+
+class SpiralIn(AbstractPermuter):
+ def __init__(self, H, W):
+ super().__init__()
+ assert H == W
+ size = W
+ indices = np.arange(size*size).reshape(size,size)
+
+ i0 = size//2
+ j0 = size//2-1
+
+ i = i0
+ j = j0
+
+ idx = [indices[i0, j0]]
+ step_mult = 0
+ for c in range(1, size//2+1):
+ step_mult += 1
+ # steps left
+ for k in range(step_mult):
+ i = i - 1
+ j = j
+ idx.append(indices[i, j])
+
+ # step down
+ for k in range(step_mult):
+ i = i
+ j = j + 1
+ idx.append(indices[i, j])
+
+ step_mult += 1
+ if c < size//2:
+ # step right
+ for k in range(step_mult):
+ i = i + 1
+ j = j
+ idx.append(indices[i, j])
+
+ # step up
+ for k in range(step_mult):
+ i = i
+ j = j - 1
+ idx.append(indices[i, j])
+ else:
+ # end reached
+ for k in range(step_mult-1):
+ i = i + 1
+ idx.append(indices[i, j])
+
+ assert len(idx) == size*size
+ idx = idx[::-1]
+ idx = torch.tensor(idx)
+ self.register_buffer('forward_shuffle_idx', idx)
+ self.register_buffer('backward_shuffle_idx', torch.argsort(idx))
+
+ def forward(self, x, reverse=False):
+ if not reverse:
+ return x[:, self.forward_shuffle_idx]
+ else:
+ return x[:, self.backward_shuffle_idx]
+
+
+class Random(nn.Module):
+ def __init__(self, H, W):
+ super().__init__()
+ indices = np.random.RandomState(1).permutation(H*W)
+ idx = torch.tensor(indices.ravel())
+ self.register_buffer('forward_shuffle_idx', idx)
+ self.register_buffer('backward_shuffle_idx', torch.argsort(idx))
+
+ def forward(self, x, reverse=False):
+ if not reverse:
+ return x[:, self.forward_shuffle_idx]
+ else:
+ return x[:, self.backward_shuffle_idx]
+
+
+class AlternateParsing(AbstractPermuter):
+ def __init__(self, H, W):
+ super().__init__()
+ indices = np.arange(W*H).reshape(H,W)
+ for i in range(1, H, 2):
+ indices[i, :] = indices[i, ::-1]
+ idx = indices.flatten()
+ assert len(idx) == H*W
+ idx = torch.tensor(idx)
+ self.register_buffer('forward_shuffle_idx', idx)
+ self.register_buffer('backward_shuffle_idx', torch.argsort(idx))
+
+ def forward(self, x, reverse=False):
+ if not reverse:
+ return x[:, self.forward_shuffle_idx]
+ else:
+ return x[:, self.backward_shuffle_idx]
+
+
+if __name__ == "__main__":
+ p0 = AlternateParsing(16, 16)
+ print(p0.forward_shuffle_idx)
+ print(p0.backward_shuffle_idx)
+
+ x = torch.randint(0, 768, size=(11, 256))
+ y = p0(x)
+ xre = p0(y, reverse=True)
+ assert torch.equal(x, xre)
+
+ p1 = SpiralOut(2, 2)
+ print(p1.forward_shuffle_idx)
+ print(p1.backward_shuffle_idx)
diff --git a/OCR_earsing/latent_diffusion/taming/modules/util.py b/OCR_earsing/latent_diffusion/taming/modules/util.py
new file mode 100644
index 0000000..9ee1638
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/util.py
@@ -0,0 +1,130 @@
+import torch
+import torch.nn as nn
+
+
+def count_params(model):
+ total_params = sum(p.numel() for p in model.parameters())
+ return total_params
+
+
+class ActNorm(nn.Module):
+ def __init__(self, num_features, logdet=False, affine=True,
+ allow_reverse_init=False):
+ assert affine
+ super().__init__()
+ self.logdet = logdet
+ self.loc = nn.Parameter(torch.zeros(1, num_features, 1, 1))
+ self.scale = nn.Parameter(torch.ones(1, num_features, 1, 1))
+ self.allow_reverse_init = allow_reverse_init
+
+ self.register_buffer('initialized', torch.tensor(0, dtype=torch.uint8))
+
+ def initialize(self, input):
+ with torch.no_grad():
+ flatten = input.permute(1, 0, 2, 3).contiguous().view(input.shape[1], -1)
+ mean = (
+ flatten.mean(1)
+ .unsqueeze(1)
+ .unsqueeze(2)
+ .unsqueeze(3)
+ .permute(1, 0, 2, 3)
+ )
+ std = (
+ flatten.std(1)
+ .unsqueeze(1)
+ .unsqueeze(2)
+ .unsqueeze(3)
+ .permute(1, 0, 2, 3)
+ )
+
+ self.loc.data.copy_(-mean)
+ self.scale.data.copy_(1 / (std + 1e-6))
+
+ def forward(self, input, reverse=False):
+ if reverse:
+ return self.reverse(input)
+ if len(input.shape) == 2:
+ input = input[:,:,None,None]
+ squeeze = True
+ else:
+ squeeze = False
+
+ _, _, height, width = input.shape
+
+ if self.training and self.initialized.item() == 0:
+ self.initialize(input)
+ self.initialized.fill_(1)
+
+ h = self.scale * (input + self.loc)
+
+ if squeeze:
+ h = h.squeeze(-1).squeeze(-1)
+
+ if self.logdet:
+ log_abs = torch.log(torch.abs(self.scale))
+ logdet = height*width*torch.sum(log_abs)
+ logdet = logdet * torch.ones(input.shape[0]).to(input)
+ return h, logdet
+
+ return h
+
+ def reverse(self, output):
+ if self.training and self.initialized.item() == 0:
+ if not self.allow_reverse_init:
+ raise RuntimeError(
+ "Initializing ActNorm in reverse direction is "
+ "disabled by default. Use allow_reverse_init=True to enable."
+ )
+ else:
+ self.initialize(output)
+ self.initialized.fill_(1)
+
+ if len(output.shape) == 2:
+ output = output[:,:,None,None]
+ squeeze = True
+ else:
+ squeeze = False
+
+ h = output / self.scale - self.loc
+
+ if squeeze:
+ h = h.squeeze(-1).squeeze(-1)
+ return h
+
+
+class AbstractEncoder(nn.Module):
+ def __init__(self):
+ super().__init__()
+
+ def encode(self, *args, **kwargs):
+ raise NotImplementedError
+
+
+class Labelator(AbstractEncoder):
+ """Net2Net Interface for Class-Conditional Model"""
+ def __init__(self, n_classes, quantize_interface=True):
+ super().__init__()
+ self.n_classes = n_classes
+ self.quantize_interface = quantize_interface
+
+ def encode(self, c):
+ c = c[:,None]
+ if self.quantize_interface:
+ return c, None, [None, None, c.long()]
+ return c
+
+
+class SOSProvider(AbstractEncoder):
+ # for unconditional training
+ def __init__(self, sos_token, quantize_interface=True):
+ super().__init__()
+ self.sos_token = sos_token
+ self.quantize_interface = quantize_interface
+
+ def encode(self, x):
+ # get batch size from data and replicate sos_token
+ c = torch.ones(x.shape[0], 1)*self.sos_token
+ c = c.long().to(x.device)
+ if self.quantize_interface:
+ return c, None, [None, None, c]
+ return c
diff --git a/OCR_earsing/latent_diffusion/taming/modules/vqvae/__init__.py b/OCR_earsing/latent_diffusion/taming/modules/vqvae/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/OCR_earsing/latent_diffusion/taming/modules/vqvae/quantize.py b/OCR_earsing/latent_diffusion/taming/modules/vqvae/quantize.py
new file mode 100644
index 0000000..d75544e
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/modules/vqvae/quantize.py
@@ -0,0 +1,445 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import numpy as np
+from torch import einsum
+from einops import rearrange
+
+
+class VectorQuantizer(nn.Module):
+ """
+ see https://github.com/MishaLaskin/vqvae/blob/d761a999e2267766400dc646d82d3ac3657771d4/models/quantizer.py
+ ____________________________________________
+ Discretization bottleneck part of the VQ-VAE.
+ Inputs:
+ - n_e : number of embeddings
+ - e_dim : dimension of embedding
+ - beta : commitment cost used in loss term, beta * ||z_e(x)-sg[e]||^2
+ _____________________________________________
+ """
+
+ # NOTE: this class contains a bug regarding beta; see VectorQuantizer2 for
+ # a fix and use legacy=False to apply that fix. VectorQuantizer2 can be
+ # used wherever VectorQuantizer has been used before and is additionally
+ # more efficient.
+ def __init__(self, n_e, e_dim, beta):
+ super(VectorQuantizer, self).__init__()
+ self.n_e = n_e
+ self.e_dim = e_dim
+ self.beta = beta
+
+ self.embedding = nn.Embedding(self.n_e, self.e_dim)
+ self.embedding.weight.data.uniform_(-1.0 / self.n_e, 1.0 / self.n_e)
+
+ def forward(self, z):
+ """
+ Inputs the output of the encoder network z and maps it to a discrete
+ one-hot vector that is the index of the closest embedding vector e_j
+ z (continuous) -> z_q (discrete)
+ z.shape = (batch, channel, height, width)
+ quantization pipeline:
+ 1. get encoder input (B,C,H,W)
+ 2. flatten input to (B*H*W,C)
+ """
+ # reshape z -> (batch, height, width, channel) and flatten
+ z = z.permute(0, 2, 3, 1).contiguous()
+ z_flattened = z.view(-1, self.e_dim)
+ # distances from z to embeddings e_j (z - e)^2 = z^2 + e^2 - 2 e * z
+
+ d = torch.sum(z_flattened ** 2, dim=1, keepdim=True) + \
+ torch.sum(self.embedding.weight**2, dim=1) - 2 * \
+ torch.matmul(z_flattened, self.embedding.weight.t())
+
+ ## could possible replace this here
+ # #\start...
+ # find closest encodings
+ min_encoding_indices = torch.argmin(d, dim=1).unsqueeze(1)
+
+ min_encodings = torch.zeros(
+ min_encoding_indices.shape[0], self.n_e).to(z)
+ min_encodings.scatter_(1, min_encoding_indices, 1)
+
+ # dtype min encodings: torch.float32
+ # min_encodings shape: torch.Size([2048, 512])
+ # min_encoding_indices.shape: torch.Size([2048, 1])
+
+ # get quantized latent vectors
+ z_q = torch.matmul(min_encodings, self.embedding.weight).view(z.shape)
+ #.........\end
+
+ # with:
+ # .........\start
+ #min_encoding_indices = torch.argmin(d, dim=1)
+ #z_q = self.embedding(min_encoding_indices)
+ # ......\end......... (TODO)
+
+ # compute loss for embedding
+ loss = torch.mean((z_q.detach()-z)**2) + self.beta * \
+ torch.mean((z_q - z.detach()) ** 2)
+
+ # preserve gradients
+ z_q = z + (z_q - z).detach()
+
+ # perplexity
+ e_mean = torch.mean(min_encodings, dim=0)
+ perplexity = torch.exp(-torch.sum(e_mean * torch.log(e_mean + 1e-10)))
+
+ # reshape back to match original input shape
+ z_q = z_q.permute(0, 3, 1, 2).contiguous()
+
+ return z_q, loss, (perplexity, min_encodings, min_encoding_indices)
+
+ def get_codebook_entry(self, indices, shape):
+ # shape specifying (batch, height, width, channel)
+ # TODO: check for more easy handling with nn.Embedding
+ min_encodings = torch.zeros(indices.shape[0], self.n_e).to(indices)
+ min_encodings.scatter_(1, indices[:,None], 1)
+
+ # get quantized latent vectors
+ z_q = torch.matmul(min_encodings.float(), self.embedding.weight)
+
+ if shape is not None:
+ z_q = z_q.view(shape)
+
+ # reshape back to match original input shape
+ z_q = z_q.permute(0, 3, 1, 2).contiguous()
+
+ return z_q
+
+
+class GumbelQuantize(nn.Module):
+ """
+ credit to @karpathy: https://github.com/karpathy/deep-vector-quantization/blob/main/model.py (thanks!)
+ Gumbel Softmax trick quantizer
+ Categorical Reparameterization with Gumbel-Softmax, Jang et al. 2016
+ https://arxiv.org/abs/1611.01144
+ """
+ def __init__(self, num_hiddens, embedding_dim, n_embed, straight_through=True,
+ kl_weight=5e-4, temp_init=1.0, use_vqinterface=True,
+ remap=None, unknown_index="random"):
+ super().__init__()
+
+ self.embedding_dim = embedding_dim
+ self.n_embed = n_embed
+
+ self.straight_through = straight_through
+ self.temperature = temp_init
+ self.kl_weight = kl_weight
+
+ self.proj = nn.Conv2d(num_hiddens, n_embed, 1)
+ self.embed = nn.Embedding(n_embed, embedding_dim)
+
+ self.use_vqinterface = use_vqinterface
+
+ self.remap = remap
+ if self.remap is not None:
+ self.register_buffer("used", torch.tensor(np.load(self.remap)))
+ self.re_embed = self.used.shape[0]
+ self.unknown_index = unknown_index # "random" or "extra" or integer
+ if self.unknown_index == "extra":
+ self.unknown_index = self.re_embed
+ self.re_embed = self.re_embed+1
+ print(f"Remapping {self.n_embed} indices to {self.re_embed} indices. "
+ f"Using {self.unknown_index} for unknown indices.")
+ else:
+ self.re_embed = n_embed
+
+ def remap_to_used(self, inds):
+ ishape = inds.shape
+ assert len(ishape)>1
+ inds = inds.reshape(ishape[0],-1)
+ used = self.used.to(inds)
+ match = (inds[:,:,None]==used[None,None,...]).long()
+ new = match.argmax(-1)
+ unknown = match.sum(2)<1
+ if self.unknown_index == "random":
+ new[unknown]=torch.randint(0,self.re_embed,size=new[unknown].shape).to(device=new.device)
+ else:
+ new[unknown] = self.unknown_index
+ return new.reshape(ishape)
+
+ def unmap_to_all(self, inds):
+ ishape = inds.shape
+ assert len(ishape)>1
+ inds = inds.reshape(ishape[0],-1)
+ used = self.used.to(inds)
+ if self.re_embed > self.used.shape[0]: # extra token
+ inds[inds>=self.used.shape[0]] = 0 # simply set to zero
+ back=torch.gather(used[None,:][inds.shape[0]*[0],:], 1, inds)
+ return back.reshape(ishape)
+
+ def forward(self, z, temp=None, return_logits=False):
+ # force hard = True when we are in eval mode, as we must quantize. actually, always true seems to work
+ hard = self.straight_through if self.training else True
+ temp = self.temperature if temp is None else temp
+
+ logits = self.proj(z)
+ if self.remap is not None:
+ # continue only with used logits
+ full_zeros = torch.zeros_like(logits)
+ logits = logits[:,self.used,...]
+
+ soft_one_hot = F.gumbel_softmax(logits, tau=temp, dim=1, hard=hard)
+ if self.remap is not None:
+ # go back to all entries but unused set to zero
+ full_zeros[:,self.used,...] = soft_one_hot
+ soft_one_hot = full_zeros
+ z_q = einsum('b n h w, n d -> b d h w', soft_one_hot, self.embed.weight)
+
+ # + kl divergence to the prior loss
+ qy = F.softmax(logits, dim=1)
+ diff = self.kl_weight * torch.sum(qy * torch.log(qy * self.n_embed + 1e-10), dim=1).mean()
+
+ ind = soft_one_hot.argmax(dim=1)
+ if self.remap is not None:
+ ind = self.remap_to_used(ind)
+ if self.use_vqinterface:
+ if return_logits:
+ return z_q, diff, (None, None, ind), logits
+ return z_q, diff, (None, None, ind)
+ return z_q, diff, ind
+
+ def get_codebook_entry(self, indices, shape):
+ b, h, w, c = shape
+ assert b*h*w == indices.shape[0]
+ indices = rearrange(indices, '(b h w) -> b h w', b=b, h=h, w=w)
+ if self.remap is not None:
+ indices = self.unmap_to_all(indices)
+ one_hot = F.one_hot(indices, num_classes=self.n_embed).permute(0, 3, 1, 2).float()
+ z_q = einsum('b n h w, n d -> b d h w', one_hot, self.embed.weight)
+ return z_q
+
+
+class VectorQuantizer2(nn.Module):
+ """
+ Improved version over VectorQuantizer, can be used as a drop-in replacement. Mostly
+ avoids costly matrix multiplications and allows for post-hoc remapping of indices.
+ """
+ # NOTE: due to a bug the beta term was applied to the wrong term. for
+ # backwards compatibility we use the buggy version by default, but you can
+ # specify legacy=False to fix it.
+ def __init__(self, n_e, e_dim, beta, remap=None, unknown_index="random",
+ sane_index_shape=False, legacy=True):
+ super().__init__()
+ self.n_e = n_e
+ self.e_dim = e_dim
+ self.beta = beta
+ self.legacy = legacy
+
+ self.embedding = nn.Embedding(self.n_e, self.e_dim)
+ self.embedding.weight.data.uniform_(-1.0 / self.n_e, 1.0 / self.n_e)
+
+ self.remap = remap
+ if self.remap is not None:
+ self.register_buffer("used", torch.tensor(np.load(self.remap)))
+ self.re_embed = self.used.shape[0]
+ self.unknown_index = unknown_index # "random" or "extra" or integer
+ if self.unknown_index == "extra":
+ self.unknown_index = self.re_embed
+ self.re_embed = self.re_embed+1
+ print(f"Remapping {self.n_e} indices to {self.re_embed} indices. "
+ f"Using {self.unknown_index} for unknown indices.")
+ else:
+ self.re_embed = n_e
+
+ self.sane_index_shape = sane_index_shape
+
+ def remap_to_used(self, inds):
+ ishape = inds.shape
+ assert len(ishape)>1
+ inds = inds.reshape(ishape[0],-1)
+ used = self.used.to(inds)
+ match = (inds[:,:,None]==used[None,None,...]).long()
+ new = match.argmax(-1)
+ unknown = match.sum(2)<1
+ if self.unknown_index == "random":
+ new[unknown]=torch.randint(0,self.re_embed,size=new[unknown].shape).to(device=new.device)
+ else:
+ new[unknown] = self.unknown_index
+ return new.reshape(ishape)
+
+ def unmap_to_all(self, inds):
+ ishape = inds.shape
+ assert len(ishape)>1
+ inds = inds.reshape(ishape[0],-1)
+ used = self.used.to(inds)
+ if self.re_embed > self.used.shape[0]: # extra token
+ inds[inds>=self.used.shape[0]] = 0 # simply set to zero
+ back=torch.gather(used[None,:][inds.shape[0]*[0],:], 1, inds)
+ return back.reshape(ishape)
+
+ def forward(self, z, temp=None, rescale_logits=False, return_logits=False):
+ assert temp is None or temp==1.0, "Only for interface compatible with Gumbel"
+ assert rescale_logits==False, "Only for interface compatible with Gumbel"
+ assert return_logits==False, "Only for interface compatible with Gumbel"
+ # reshape z -> (batch, height, width, channel) and flatten
+ z = rearrange(z, 'b c h w -> b h w c').contiguous()
+ z_flattened = z.view(-1, self.e_dim)
+ # distances from z to embeddings e_j (z - e)^2 = z^2 + e^2 - 2 e * z
+
+ d = torch.sum(z_flattened ** 2, dim=1, keepdim=True) + \
+ torch.sum(self.embedding.weight**2, dim=1) - 2 * \
+ torch.einsum('bd,dn->bn', z_flattened, rearrange(self.embedding.weight, 'n d -> d n'))
+
+ min_encoding_indices = torch.argmin(d, dim=1)
+ z_q = self.embedding(min_encoding_indices).view(z.shape)
+ perplexity = None
+ min_encodings = None
+
+ # compute loss for embedding
+ if not self.legacy:
+ loss = self.beta * torch.mean((z_q.detach()-z)**2) + \
+ torch.mean((z_q - z.detach()) ** 2)
+ else:
+ loss = torch.mean((z_q.detach()-z)**2) + self.beta * \
+ torch.mean((z_q - z.detach()) ** 2)
+
+ # preserve gradients
+ z_q = z + (z_q - z).detach()
+
+ # reshape back to match original input shape
+ z_q = rearrange(z_q, 'b h w c -> b c h w').contiguous()
+
+ if self.remap is not None:
+ min_encoding_indices = min_encoding_indices.reshape(z.shape[0],-1) # add batch axis
+ min_encoding_indices = self.remap_to_used(min_encoding_indices)
+ min_encoding_indices = min_encoding_indices.reshape(-1,1) # flatten
+
+ if self.sane_index_shape:
+ min_encoding_indices = min_encoding_indices.reshape(
+ z_q.shape[0], z_q.shape[2], z_q.shape[3])
+
+ return z_q, loss, (perplexity, min_encodings, min_encoding_indices)
+
+ def get_codebook_entry(self, indices, shape):
+ # shape specifying (batch, height, width, channel)
+ if self.remap is not None:
+ indices = indices.reshape(shape[0],-1) # add batch axis
+ indices = self.unmap_to_all(indices)
+ indices = indices.reshape(-1) # flatten again
+
+ # get quantized latent vectors
+ z_q = self.embedding(indices)
+
+ if shape is not None:
+ z_q = z_q.view(shape)
+ # reshape back to match original input shape
+ z_q = z_q.permute(0, 3, 1, 2).contiguous()
+
+ return z_q
+
+class EmbeddingEMA(nn.Module):
+ def __init__(self, num_tokens, codebook_dim, decay=0.99, eps=1e-5):
+ super().__init__()
+ self.decay = decay
+ self.eps = eps
+ weight = torch.randn(num_tokens, codebook_dim)
+ self.weight = nn.Parameter(weight, requires_grad = False)
+ self.cluster_size = nn.Parameter(torch.zeros(num_tokens), requires_grad = False)
+ self.embed_avg = nn.Parameter(weight.clone(), requires_grad = False)
+ self.update = True
+
+ def forward(self, embed_id):
+ return F.embedding(embed_id, self.weight)
+
+ def cluster_size_ema_update(self, new_cluster_size):
+ self.cluster_size.data.mul_(self.decay).add_(new_cluster_size, alpha=1 - self.decay)
+
+ def embed_avg_ema_update(self, new_embed_avg):
+ self.embed_avg.data.mul_(self.decay).add_(new_embed_avg, alpha=1 - self.decay)
+
+ def weight_update(self, num_tokens):
+ n = self.cluster_size.sum()
+ smoothed_cluster_size = (
+ (self.cluster_size + self.eps) / (n + num_tokens * self.eps) * n
+ )
+ #normalize embedding average with smoothed cluster size
+ embed_normalized = self.embed_avg / smoothed_cluster_size.unsqueeze(1)
+ self.weight.data.copy_(embed_normalized)
+
+
+class EMAVectorQuantizer(nn.Module):
+ def __init__(self, n_embed, embedding_dim, beta, decay=0.99, eps=1e-5,
+ remap=None, unknown_index="random"):
+ super().__init__()
+ self.codebook_dim = codebook_dim
+ self.num_tokens = num_tokens
+ self.beta = beta
+ self.embedding = EmbeddingEMA(self.num_tokens, self.codebook_dim, decay, eps)
+
+ self.remap = remap
+ if self.remap is not None:
+ self.register_buffer("used", torch.tensor(np.load(self.remap)))
+ self.re_embed = self.used.shape[0]
+ self.unknown_index = unknown_index # "random" or "extra" or integer
+ if self.unknown_index == "extra":
+ self.unknown_index = self.re_embed
+ self.re_embed = self.re_embed+1
+ print(f"Remapping {self.n_embed} indices to {self.re_embed} indices. "
+ f"Using {self.unknown_index} for unknown indices.")
+ else:
+ self.re_embed = n_embed
+
+ def remap_to_used(self, inds):
+ ishape = inds.shape
+ assert len(ishape)>1
+ inds = inds.reshape(ishape[0],-1)
+ used = self.used.to(inds)
+ match = (inds[:,:,None]==used[None,None,...]).long()
+ new = match.argmax(-1)
+ unknown = match.sum(2)<1
+ if self.unknown_index == "random":
+ new[unknown]=torch.randint(0,self.re_embed,size=new[unknown].shape).to(device=new.device)
+ else:
+ new[unknown] = self.unknown_index
+ return new.reshape(ishape)
+
+ def unmap_to_all(self, inds):
+ ishape = inds.shape
+ assert len(ishape)>1
+ inds = inds.reshape(ishape[0],-1)
+ used = self.used.to(inds)
+ if self.re_embed > self.used.shape[0]: # extra token
+ inds[inds>=self.used.shape[0]] = 0 # simply set to zero
+ back=torch.gather(used[None,:][inds.shape[0]*[0],:], 1, inds)
+ return back.reshape(ishape)
+
+ def forward(self, z):
+ # reshape z -> (batch, height, width, channel) and flatten
+ #z, 'b c h w -> b h w c'
+ z = rearrange(z, 'b c h w -> b h w c')
+ z_flattened = z.reshape(-1, self.codebook_dim)
+
+ # distances from z to embeddings e_j (z - e)^2 = z^2 + e^2 - 2 e * z
+ d = z_flattened.pow(2).sum(dim=1, keepdim=True) + \
+ self.embedding.weight.pow(2).sum(dim=1) - 2 * \
+ torch.einsum('bd,nd->bn', z_flattened, self.embedding.weight) # 'n d -> d n'
+
+
+ encoding_indices = torch.argmin(d, dim=1)
+
+ z_q = self.embedding(encoding_indices).view(z.shape)
+ encodings = F.one_hot(encoding_indices, self.num_tokens).type(z.dtype)
+ avg_probs = torch.mean(encodings, dim=0)
+ perplexity = torch.exp(-torch.sum(avg_probs * torch.log(avg_probs + 1e-10)))
+
+ if self.training and self.embedding.update:
+ #EMA cluster size
+ encodings_sum = encodings.sum(0)
+ self.embedding.cluster_size_ema_update(encodings_sum)
+ #EMA embedding average
+ embed_sum = encodings.transpose(0,1) @ z_flattened
+ self.embedding.embed_avg_ema_update(embed_sum)
+ #normalize embed_avg and update weight
+ self.embedding.weight_update(self.num_tokens)
+
+ # compute loss for embedding
+ loss = self.beta * F.mse_loss(z_q.detach(), z)
+
+ # preserve gradients
+ z_q = z + (z_q - z).detach()
+
+ # reshape back to match original input shape
+ #z_q, 'b h w c -> b c h w'
+ z_q = rearrange(z_q, 'b h w c -> b c h w')
+ return z_q, loss, (perplexity, encodings, encoding_indices)
diff --git a/OCR_earsing/latent_diffusion/taming/util.py b/OCR_earsing/latent_diffusion/taming/util.py
new file mode 100644
index 0000000..06053e5
--- /dev/null
+++ b/OCR_earsing/latent_diffusion/taming/util.py
@@ -0,0 +1,157 @@
+import os, hashlib
+import requests
+from tqdm import tqdm
+
+URL_MAP = {
+ "vgg_lpips": "https://heibox.uni-heidelberg.de/f/607503859c864bc1b30b/?dl=1"
+}
+
+CKPT_MAP = {
+ "vgg_lpips": "vgg.pth"
+}
+
+MD5_MAP = {
+ "vgg_lpips": "d507d7349b931f0638a25a48a722f98a"
+}
+
+
+def download(url, local_path, chunk_size=1024):
+ os.makedirs(os.path.split(local_path)[0], exist_ok=True)
+ with requests.get(url, stream=True) as r:
+ total_size = int(r.headers.get("content-length", 0))
+ with tqdm(total=total_size, unit="B", unit_scale=True) as pbar:
+ with open(local_path, "wb") as f:
+ for data in r.iter_content(chunk_size=chunk_size):
+ if data:
+ f.write(data)
+ pbar.update(chunk_size)
+
+
+def md5_hash(path):
+ with open(path, "rb") as f:
+ content = f.read()
+ return hashlib.md5(content).hexdigest()
+
+
+def get_ckpt_path(name, root, check=False):
+ assert name in URL_MAP
+ path = os.path.join(root, CKPT_MAP[name])
+ if not os.path.exists(path) or (check and not md5_hash(path) == MD5_MAP[name]):
+ print("Downloading {} model from {} to {}".format(name, URL_MAP[name], path))
+ download(URL_MAP[name], path)
+ md5 = md5_hash(path)
+ assert md5 == MD5_MAP[name], md5
+ return path
+
+
+class KeyNotFoundError(Exception):
+ def __init__(self, cause, keys=None, visited=None):
+ self.cause = cause
+ self.keys = keys
+ self.visited = visited
+ messages = list()
+ if keys is not None:
+ messages.append("Key not found: {}".format(keys))
+ if visited is not None:
+ messages.append("Visited: {}".format(visited))
+ messages.append("Cause:\n{}".format(cause))
+ message = "\n".join(messages)
+ super().__init__(message)
+
+
+def retrieve(
+ list_or_dict, key, splitval="/", default=None, expand=True, pass_success=False
+):
+ """Given a nested list or dict return the desired value at key expanding
+ callable nodes if necessary and :attr:`expand` is ``True``. The expansion
+ is done in-place.
+
+ Parameters
+ ----------
+ list_or_dict : list or dict
+ Possibly nested list or dictionary.
+ key : str
+ key/to/value, path like string describing all keys necessary to
+ consider to get to the desired value. List indices can also be
+ passed here.
+ splitval : str
+ String that defines the delimiter between keys of the
+ different depth levels in `key`.
+ default : obj
+ Value returned if :attr:`key` is not found.
+ expand : bool
+ Whether to expand callable nodes on the path or not.
+
+ Returns
+ -------
+ The desired value or if :attr:`default` is not ``None`` and the
+ :attr:`key` is not found returns ``default``.
+
+ Raises
+ ------
+ Exception if ``key`` not in ``list_or_dict`` and :attr:`default` is
+ ``None``.
+ """
+
+ keys = key.split(splitval)
+
+ success = True
+ try:
+ visited = []
+ parent = None
+ last_key = None
+ for key in keys:
+ if callable(list_or_dict):
+ if not expand:
+ raise KeyNotFoundError(
+ ValueError(
+ "Trying to get past callable node with expand=False."
+ ),
+ keys=keys,
+ visited=visited,
+ )
+ list_or_dict = list_or_dict()
+ parent[last_key] = list_or_dict
+
+ last_key = key
+ parent = list_or_dict
+
+ try:
+ if isinstance(list_or_dict, dict):
+ list_or_dict = list_or_dict[key]
+ else:
+ list_or_dict = list_or_dict[int(key)]
+ except (KeyError, IndexError, ValueError) as e:
+ raise KeyNotFoundError(e, keys=keys, visited=visited)
+
+ visited += [key]
+ # final expansion of retrieved value
+ if expand and callable(list_or_dict):
+ list_or_dict = list_or_dict()
+ parent[last_key] = list_or_dict
+ except KeyNotFoundError as e:
+ if default is None:
+ raise e
+ else:
+ list_or_dict = default
+ success = False
+
+ if not pass_success:
+ return list_or_dict
+ else:
+ return list_or_dict, success
+
+
+if __name__ == "__main__":
+ config = {"keya": "a",
+ "keyb": "b",
+ "keyc":
+ {"cc1": 1,
+ "cc2": 2,
+ }
+ }
+ from omegaconf import OmegaConf
+ config = OmegaConf.create(config)
+ print(config)
+ retrieve(config, "keya")
+
diff --git a/OCR_earsing/model_checkpoint/det_latin_words/1/model.onnx b/OCR_earsing/model_checkpoint/det_latin_words/1/model.onnx
new file mode 100644
index 0000000..23fb0fd
Binary files /dev/null and b/OCR_earsing/model_checkpoint/det_latin_words/1/model.onnx differ
diff --git a/OCR_earsing/model_checkpoint/latin_dict.txt b/OCR_earsing/model_checkpoint/latin_dict.txt
new file mode 100644
index 0000000..e166bf3
--- /dev/null
+++ b/OCR_earsing/model_checkpoint/latin_dict.txt
@@ -0,0 +1,185 @@
+
+!
+"
+#
+$
+%
+&
+'
+(
+)
+*
++
+,
+-
+.
+/
+0
+1
+2
+3
+4
+5
+6
+7
+8
+9
+:
+;
+<
+=
+>
+?
+@
+A
+B
+C
+D
+E
+F
+G
+H
+I
+J
+K
+L
+M
+N
+O
+P
+Q
+R
+S
+T
+U
+V
+W
+X
+Y
+Z
+[
+]
+_
+`
+a
+b
+c
+d
+e
+f
+g
+h
+i
+j
+k
+l
+m
+n
+o
+p
+q
+r
+s
+t
+u
+v
+w
+x
+y
+z
+{
+}
+¡
+£
+§
+ª
+«
+
+°
+²
+³
+´
+µ
+·
+º
+»
+¿
+À
+Á
+Â
+Ä
+Å
+Ç
+È
+É
+Ê
+Ë
+Ì
+Í
+Î
+Ï
+Ò
+Ó
+Ô
+Õ
+Ö
+Ú
+Ü
+Ý
+ß
+à
+á
+â
+ã
+ä
+å
+æ
+ç
+è
+é
+ê
+ë
+ì
+í
+î
+ï
+ñ
+ò
+ó
+ô
+õ
+ö
+ø
+ù
+ú
+û
+ü
+ý
+ą
+Ć
+ć
+Č
+č
+Đ
+đ
+ę
+ı
+Ł
+ł
+ō
+Œ
+œ
+Š
+š
+Ÿ
+Ž
+ž
+ʒ
+β
+δ
+ε
+з
+Ṡ
+‘
+€
+™
diff --git a/OCR_earsing/model_checkpoint/rec_general_words_fr/1/model.onnx b/OCR_earsing/model_checkpoint/rec_general_words_fr/1/model.onnx
new file mode 100644
index 0000000..8d9123d
Binary files /dev/null and b/OCR_earsing/model_checkpoint/rec_general_words_fr/1/model.onnx differ
diff --git a/OCR_earsing/ocr_eraser.py b/OCR_earsing/ocr_eraser.py
new file mode 100644
index 0000000..8e7a377
--- /dev/null
+++ b/OCR_earsing/ocr_eraser.py
@@ -0,0 +1,125 @@
+import logging
+import os
+import sys
+import torch
+
+# Download checkpoints
+os.system("pip install git+https://github.com/facebookresearch/segment-anything.git")
+os.system('wget https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth')
+os.system('wget -O last.ckpt https://heibox.uni-heidelberg.de/f/4d9ac7ea40c64582b7c9/?dl=1')
+
+from PIL import Image
+import numpy as np
+import cv2
+
+# OCR
+from utils.ocr_utils import ocr_extraction
+from utils.easy_ocr_utils import easy_ocr_extraction
+
+# SAM
+from segment_anything import SamPredictor, sam_model_registry
+
+# Diffusion model
+sys.path.append('latent_diffusion')
+from latent_diffusion.ldm_erase_text import erase_text_from_image, instantiate_from_config, OmegaConf
+
+logger = logging.getLogger(__name__)
+
+
+def multi_mask2one_mask(masks):
+ _, _, h, w = masks.shape
+ for i, mask in enumerate(masks):
+ mask_image = mask.reshape(h, w, 1)
+ whole_mask = mask_image if i == 0 else whole_mask + mask_image
+ whole_mask = np.where(whole_mask == False, 0, 255)
+ return whole_mask
+
+
+def numpy2PIL(numpy_image):
+ out = Image.fromarray(numpy_image.astype(np.uint8))
+ return out
+
+
+def show_mask(mask, ax, random_color=False):
+ if random_color:
+ color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
+ else:
+ color = np.array([30 / 255, 144 / 255, 255 / 255, 0.6])
+ h, w = mask.shape[-2:]
+ mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
+ ax.imshow(mask_image)
+
+
+def run_earse(img_path, sam_type, sam_checkpoint, config_path, model_checkpoint, device="cpu", img_size=(512, 512),
+ steps=50, use_easy_ocr=False):
+ img = cv2.imread(img_path)
+ # h, w, c = img.shape
+
+ # SAM
+ sam = sam_model_registry[sam_type](checkpoint=sam_checkpoint)
+ sam = sam.to(device)
+ sam_predictor = SamPredictor(sam)
+
+ # Diffusion model
+ config = OmegaConf.load(config_path)
+ model = instantiate_from_config(config.model)
+ model.load_state_dict(
+ torch.load(model_checkpoint)["state_dict"],
+ strict=False
+ )
+ model = model.to(device)
+
+ if use_easy_ocr:
+ word_info = easy_ocr_extraction(img_path)
+ else:
+ word_info = ocr_extraction(img_path)
+ det_bboxes = [bbox[:4] for bbox in word_info]
+ # convert to torch tensor
+ det_bboxes = torch.tensor(det_bboxes)
+ transformed_boxes = sam_predictor.transform.apply_boxes_torch(
+ det_bboxes, img.shape[:2]
+ )
+ sam_predictor.set_image(img, image_format='BGR')
+ masks, _, _ = sam_predictor.predict_torch(
+ point_coords=None,
+ point_labels=None,
+ boxes=transformed_boxes,
+ multimask_output=False,
+ )
+ ori_mask = multi_mask2one_mask(masks=masks)
+ mask_img = ori_mask[:, :, 0].astype('uint8')
+ kernel = np.ones((5, 5), np.int8)
+ whole_mask = cv2.dilate(
+ mask_img, kernel, iterations=2
+ )
+
+ mask_pil_image = numpy2PIL(numpy_image=whole_mask)
+ result_img = erase_text_from_image(
+ img_path=img_path,
+ mask_pil_img=mask_pil_image,
+ model=model,
+ device=device,
+ opt=None,
+ img_size=img_size,
+ steps=steps
+ )
+
+ result_img = cv2.cvtColor(np.array(result_img), cv2.COLOR_RGB2BGR)
+ cv2.namedWindow("Result Image", cv2.WINDOW_NORMAL)
+ cv2.imshow("Result Image", result_img)
+ cv2.waitKey(0) # Wait until a key is pressed
+ cv2.destroyAllWindows() # Close the image window
+
+
+if __name__ == "__main__":
+ run_earse(
+ img_path="Facture médecine douce-27746732_0.jpg",
+ sam_type="vit_h",
+ sam_checkpoint="sam_vit_h_4b8939.pth",
+ config_path="latent_diffusion/inpainting_big/config.yaml",
+ model_checkpoint="last.ckpt",
+ device="cpu",
+ img_size=(512, 512),
+ steps=50,
+ use_easy_ocr=True
+ )
diff --git a/OCR_earsing/requirements.txt b/OCR_earsing/requirements.txt
new file mode 100644
index 0000000..2c4dc54
--- /dev/null
+++ b/OCR_earsing/requirements.txt
@@ -0,0 +1,6 @@
+numpy
+omegaconf
+einops
+transformers
+pytorch-lightning
+diffusers
\ No newline at end of file
diff --git a/OCR_earsing/utils/easy_ocr_utils.py b/OCR_earsing/utils/easy_ocr_utils.py
new file mode 100644
index 0000000..45b92ae
--- /dev/null
+++ b/OCR_earsing/utils/easy_ocr_utils.py
@@ -0,0 +1,17 @@
+import cv2
+import easyocr
+
+
+def easy_ocr_extraction(
+ img_path: str
+) -> dict:
+ org_img_bgr = cv2.imread(img_path)
+ reader = easyocr.Reader(['fr', 'en'])
+ result = reader.readtext(org_img_bgr)
+ word_info = []
+ for (bbox, text, _) in result:
+ word_info.append(
+ [bbox[0][0], bbox[0][1], bbox[2][0], bbox[2][1], text]
+ )
+
+ return word_info
diff --git a/OCR_earsing/utils/ocr_utils.py b/OCR_earsing/utils/ocr_utils.py
new file mode 100644
index 0000000..a161656
--- /dev/null
+++ b/OCR_earsing/utils/ocr_utils.py
@@ -0,0 +1,212 @@
+from typing import List, Optional, Tuple, Union
+import logging
+import sys
+import math
+
+import cv2
+from PIL import Image
+import torch
+import numpy as np
+import numpy.typing as npt
+from cv_base.postprocess.db_postprocess import DBPostProcess
+from cv_base.postprocess.pp_rec_postprocess import CTCLabelDecode
+from cv_base.preprocess.db_preprocess import DBPreprocess
+from cv_base.utils.perspective import four_point_transform
+from infer_client.adapters import InferenceAdapter
+from infer_client.adapters.onnx import OnnxInferenceAdapter
+from infer_client.inference import Inference
+
+logger = logging.getLogger(__name__)
+
+DET_MODEL = "model_checkpoint/det_latin_words"
+REC_MODEL = "model_checkpoint/rec_general_words_fr"
+VOCAB_DICT = "model_checkpoint/latin_dict.txt"
+
+
+def get_vocab_str(vocab_dict_path: str):
+ with open(vocab_dict_path, "r") as rf:
+ vocab = rf.read()
+ return vocab.replace("\n", "")
+
+
+def box4point_to_box2point(box4point):
+ # bounding box = [x0, y0, x1, y1, x2, y2, x3, y3]
+ all_x = [box4point[2 * i] for i in range(4)]
+ all_y = [box4point[2 * i + 1] for i in range(4)]
+ box2point = [min(all_x), min(all_y), max(all_x), max(all_y)]
+ return box2point
+
+
+def minimum_bounding_rectangle(points):
+ from scipy.spatial import ConvexHull
+
+ hull = ConvexHull(points)
+ hull_points = points[hull.vertices]
+
+ # Find the minimum area rectangle
+ rect = cv2.minAreaRect(hull_points.astype(np.float32))
+ box = cv2.boxPoints(rect)
+ box = np.int0(box)
+
+ return box
+
+
+class DetTextDocument(object):
+ def __init__(
+ self,
+ adapter: InferenceAdapter,
+ img_size: Union[int, Tuple[int, int]] = 736,
+ thresh: float = 0.2,
+ bbox_thresh: float = 0.5,
+ unclip_ratio: float = 2.0,
+ ):
+ self.infer_obj = Inference(adapter)
+
+ if self.infer_obj.health():
+ logger.info(f"Init {self.__class__.__name__} done")
+ else:
+ logger.error(f"Init {self.__class__.__name__} fail")
+ sys.exit(1)
+
+ self.pre_process = DBPreprocess(img_size=img_size)
+ self.post_process = DBPostProcess(
+ thresh=thresh,
+ bbox_thresh=bbox_thresh,
+ unclip_ratio=unclip_ratio,
+ score_mode="fast",
+ )
+
+ def inference(
+ self,
+ img: npt.NDArray[np.uint8],
+ ) -> Tuple[List[Optional[npt.NDArray[np.int32]]], List[Optional[float]]]:
+ pre_processed_img, _ = self.pre_process(img)
+ outs_dbnet = self.infer_obj.inference({"input": pre_processed_img}, ["output"])
+
+ if not outs_dbnet:
+ return [], []
+
+ _, bbox_lst, score_lst = self.post_process(
+ outs_dbnet[0][0], img_shape=(img.shape[0], img.shape[1])
+ )
+ return [bbox.astype(np.int32) for bbox in bbox_lst], score_lst
+
+
+class RecPreprocessImage(object):
+ def __init__(self, rec_image_shape):
+ self.rec_image_shape = rec_image_shape
+
+ def __call__(self, img):
+ imgC, imgH, imgW = self.rec_image_shape[:3]
+ max_wh_ratio = imgW / imgH
+ h, w = img.shape[:2]
+ ratio = w * 1.0 / h
+ max_wh_ratio = max(max_wh_ratio, ratio)
+ imgW = int((imgH * max_wh_ratio))
+ if math.ceil(imgH * ratio) > imgW:
+ resized_w = imgW
+ else:
+ resized_w = int(math.ceil(imgH * ratio))
+ resized_image = cv2.resize(img, (resized_w, imgH))
+ resized_image = resized_image.astype("float32")
+ resized_image = resized_image.transpose((2, 0, 1)) / 255
+ resized_image -= 0.5
+ resized_image /= 0.5
+ padding_im = np.zeros((imgC, imgH, imgW), dtype=np.float32)
+ padding_im[:, :, 0:resized_w] = resized_image
+ return padding_im[np.newaxis, :]
+
+
+class RecTextDocument(object):
+ def __init__(
+ self,
+ adapter: InferenceAdapter,
+ vocab_dict: str,
+ ):
+ self.pre_process = RecPreprocessImage(rec_image_shape=[3, 48, 320])
+ self.infer_obj = Inference(adapter)
+
+ if self.infer_obj.health():
+ logger.info(f"Init {self.__class__.__name__} done")
+ else:
+ logger.error(f"Init {self.__class__.__name__} fail")
+ sys.exit(1)
+ vocab_dict += " "
+ self.post_process = CTCLabelDecode(vocab_dict, use_space_char=True)
+
+ def inference(self, img: npt.NDArray[np.uint8]) -> Tuple[str, float]:
+ preprocessed_img = self.pre_process(img)
+ if res := self.infer_obj.inference(
+ {"x": preprocessed_img}, ["softmax_2.tmp_0"]
+ ):
+ sent, probs = self.post_process(res)[0]
+ probs = [probs]
+ return sent, round(sum(probs) / len(probs), 5)
+ return "", 0.0
+
+
+det_model = DetTextDocument(
+ OnnxInferenceAdapter(
+ model_name=DET_MODEL,
+ version="1"
+ ),
+ # TritonInferenceAdapter(
+ # triton_server=TRITON_SERVER_URL,
+ # model_name=os.path.basename(DET_MODEL),
+ # ssl=False,
+ # ),
+ img_size=736,
+ unclip_ratio=1.7,
+)
+rec_model = RecTextDocument(
+ OnnxInferenceAdapter(
+ model_name=REC_MODEL,
+ version="1"
+ ),
+ # TritonInferenceAdapter(
+ # triton_server=TRITON_SERVER_URL,
+ # model_name=os.path.basename(REC_MODEL),
+ # ssl=False,
+ # ),
+ vocab_dict=get_vocab_str(VOCAB_DICT),
+)
+
+
+def ocr_extraction(
+ img_path: str
+) -> dict:
+ org_img_bgr = cv2.imread(img_path)
+
+ # text detection
+ bbox_lst, _ = det_model.inference(org_img_bgr)
+
+ word_info = []
+ for bbox in bbox_lst:
+ # text recognition
+ text_img = four_point_transform(org_img_bgr, bbox)
+ word, conf_text = rec_model.inference(text_img)
+ word_info.append(
+ [min(bbox[:, 0]), min(bbox[:, 1]), max(bbox[:, 0]), max(bbox[:, 1]), word]
+ )
+
+ return word_info
+
+
+def visualize_ocr(img_path: str, word_info):
+ org_img_bgr = cv2.imread(img_path)
+ for i, (x1, y1, x2, y2, word) in enumerate(word_info):
+ cv2.rectangle(org_img_bgr, (x1, y1), (x2, y2), (0, 255, 0), 2)
+ cv2.putText(
+ org_img_bgr,
+ word,
+ (x1, y1 - 10),
+ cv2.FONT_HERSHEY_SIMPLEX,
+ 0.5,
+ (0, 200, 0),
+ 2,
+ )
+
+ cv2.namedWindow("Image with Annotations", cv2.WINDOW_NORMAL)
+ cv2.imshow("Image with Annotations", org_img_bgr)
+ cv2.waitKey(0) # Wait until a key is pressed
+ cv2.destroyAllWindows() # Close the image window
\ No newline at end of file
diff --git a/README.md b/README.md
index 1e36b2f..e8792aa 100644
--- a/README.md
+++ b/README.md
@@ -1,2 +1,2 @@
-# synthetics_handwritten_OCR
+# Synthetics_handwritten_OCR
diff --git a/awesome-SynthHandwritten/README.md b/awesome-SynthHandwritten/README.md
new file mode 100644
index 0000000..eaf7126
--- /dev/null
+++ b/awesome-SynthHandwritten/README.md
@@ -0,0 +1,7 @@
+A curated list of awesome synthetic handwritten text datasets.
+
+- https://github.com/Belval/TextRecognitionDataGenerator \
+- https://github.com/oh-my-ocr/text_renderer \
+- https://github.com/Grzego/handwriting-generation \
+- https://github.com/sleep3r/Diffusion-Handwriting-Generation.pytorch/tree/main/diffusion_handwriting_generation \
+- https://github.com/ankanbhunia/Handwriting-Transformers?tab=readme-ov-file \
\ No newline at end of file
diff --git a/synthetics_handwritten/README.md b/synthetics_handwritten/README.md
new file mode 100644
index 0000000..e69de29