init commit

configs/instruction_refinement_api.json

@@ -0,0 +1,14 @@
+{
+  "job_type": "instruction_refinement_api",
+  "dataset": {
+    "input_path": "./train.json",
+    "output_path": "./train_refined.json"
+  },
+  "inference":{
+    "base_url": "ENDPOINT",
+    "api_key": "TOKEN",
+    "stream": true,
+    "prompt" : "Assume you are a prompt re-writing expert. Given an instruction as input, you should generate a new instruction semantically similar to the input to support the training of large language models. Transform the input raw prompt into a detailed prompt that comprehensively captures the user’s request. Make sure to maintain the original intent while significantly enhancing clarity and depth. You should place your answer enclosed within <answer></answer> tags. The input prompt is as follows:",
+    "max_new_tokens": 512
+  }
+}