Update app.py

app.py

@@ -69,17 +69,17 @@ qwen_model = Qwen2VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to("cuda").eval()
 
-ROLMOCR_MODEL_ID = "reducto/RolmOCR"
-rolmocr_processor = AutoProcessor.from_pretrained(ROLMOCR_MODEL_ID, trust_remote_code=True)
-rolmocr_model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
-    ROLMOCR_MODEL_ID,
+DOCSCOPEOCR_MODEL_ID = "prithivMLmods/docscopeOCR-7B-050425-exp"
+docscopeocr_processor = AutoProcessor.from_pretrained(DOCSCOPEOCR_MODEL_ID, trust_remote_code=True)
+docscopeocr_model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
+    DOCSCOPEOCR_MODEL_ID,
     trust_remote_code=True,
     torch_dtype=torch.bfloat16
 ).to("cuda").eval()
 
 # Main Inference Function
 @spaces.GPU
-def model_inference(message, history, use_rolmocr):
+def model_inference(message, history, use_docscopeocr):
     text = message["text"].strip()
     files = message.get("files", [])
 
@@ -125,10 +125,10 @@ def model_inference(message, history, use_rolmocr):
     messages = [{"role": "user", "content": content}]
 
     # Select processor and model
-    if use_rolmocr:
-        processor = rolmocr_processor
-        model = rolmocr_model
-        model_name = "RolmOCR"
+    if use_docscopeocr:
+        processor = docscopeocr_processor
+        model = docscopeocr_model
+        model_name = "DocScopeOCR"
     else:
         processor = qwen_processor
         model = qwen_model
@@ -176,7 +176,7 @@ demo = gr.ChatInterface(
     stop_btn="Stop Generation",
     multimodal=True,
     cache_examples=False,
-    additional_inputs=[gr.Checkbox(label="Use RolmOCR", value=True)],
+    additional_inputs=[gr.Checkbox(label="Use DocScopeOCR", value=True, info="Check to use DocScopeOCR, uncheck to use Qwen2VL OCR")],
 )
 
-demo.launch(debug=True)
+demo.launch(debug=True, ssr_mode=False)