Update app.py

app.py

@@ -66,10 +66,10 @@ model_j = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to(device).eval()
 
-# Load medgemma-4b-it
-MODEL_ID_F = "google/medgemma-4b-it"
+# Load LMM-R1-MGT-PerceReason
+MODEL_ID_F = "VLM-Reasoner/LMM-R1-MGT-PerceReason"
 processor_f = AutoProcessor.from_pretrained(MODEL_ID_F, trust_remote_code=True)
-model_f = AutoModelForImageTextToText.from_pretrained(
+model_f = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_F,
     trust_remote_code=True,
     torch_dtype=torch.float16
@@ -118,7 +118,7 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     elif model_name == "Lumian-VLR-7B-Thinking":
         processor = processor_j
         model = model_j
-    elif model_name == "medgemma-4b-it":
+    elif model_name == "LMM-R1-MGT-PerceReason":
         processor = processor_f
         model = model_f
     else:
@@ -177,7 +177,7 @@ def generate_video(model_name: str, text: str, video_path: str,
     elif model_name == "Lumian-VLR-7B-Thinking":
         processor = processor_j
         model = model_j
-    elif model_name == "medgemma-4b-it":
+    elif model_name == "LMM-R1-MGT-PerceReason":
         processor = processor_f
         model = model_f
     else:
@@ -302,7 +302,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
                     markdown_output = gr.Markdown(label="(Result.Md)")
         
             model_choice = gr.Radio(
-                choices=["Lumian-VLR-7B-Thinking", "DREX-062225-7B-exp", "olmOCR-7B-0225-preview", "medgemma-4b-it", "Typhoon-OCR-3B"],
+                choices=["Lumian-VLR-7B-Thinking", "DREX-062225-7B-exp", "olmOCR-7B-0225-preview", "LMM-R1-MGT-PerceReason", "Typhoon-OCR-3B"],
                 label="Select Model",
                 value="Lumian-VLR-7B-Thinking"
             )