Update main.py

main.py

@@ -23,37 +23,38 @@ from utils import (
 )
 import torch
 
-#yolo_model = get_yolo_model(model_path='best.pt')
-#caption_model_processor = get_caption_model_processor(model_name="florence2", model_name_or_path="icon_caption_florence")
+# yolo_model = get_yolo_model(model_path='/data/icon_detect/best.pt')
+# caption_model_processor = get_caption_model_processor(model_name="florence2", model_name_or_path="/data/icon_caption_florence")
 
 from ultralytics import YOLO
 
-#if not os.path.exists("NewParser/best.pt"):
-    #os.makedirs("NewParser/best.pt")
+# if not os.path.exists("/data/icon_detect"):
+#     os.makedirs("/data/icon_detect")
+
+try:
+    yolo_model = YOLO("weights/icon_detect/best.pt").to("cuda")
+except:
+    yolo_model = YOLO("weights/icon_detect/best.pt")
 
-#try:
-    #yolo_model = YOLO("best.pt").to("cuda")
-#except:
-    #yolo_model = YOLO("best.pt")
 from transformers import AutoProcessor, AutoModelForCausalLM
-from transformers import AutoProcessor, Blip2ForConditionalGeneration
-# Correctly load the processor and model for Blip-2
+
+processor = AutoProcessor.from_pretrained(
+    "microsoft/Florence-2-base", trust_remote_code=True
+)
+
 try:
-    processor = AutoProcessor.from_pretrained("microsoft/Florence-2-base", trust_remote_code=True)
-    model = Blip2ForConditionalGeneration.from_pretrained(
+    model = AutoModelForCausalLM.from_pretrained(
         "microsoft/OmniParser",
-        torch_dtype=torch.float16,  # Assuming you're using a GPU
-        trust_remote_code=True
+        torch_dtype=torch.float16,
+        trust_remote_code=True,
     ).to("cuda")
-except Exception as e:
-    print(f"Error loading caption model: {e}")
-    processor = AutoProcessor.from_pretrained("microsoft/Florence-2-base", trust_remote_code=True)
-    model = Blip2ForConditionalGeneration.from_pretrained(
+except:
+    model = AutoModelForCausalLM.from_pretrained(
         "microsoft/OmniParser",
         torch_dtype=torch.float16,
-        trust_remote_code=True
-    ).to("cpu")  # Fallback to CPU if CUDA fails
-
+        trust_remote_code=True,
+    )
+caption_model_processor = {"processor": processor, "model": model}
 print("finish loading model!!!")
 
 app = FastAPI()