support gsam2 image predictor model

grounding_dino/docker_test.py

@@ -0,0 +1,8 @@
+from groundingdino.util.inference import load_model, load_image, predict, annotate
+import torch
+import cv2
+
+model = load_model("groundingdino/config/GroundingDINO_SwinT_OGC.pyy", "weights/groundingdino_swint_ogc.pth")
+model = model.to('cuda:0')
+print(torch.cuda.is_available())
+print('DONE!')