update app

app.py

@@ -41,10 +41,10 @@ model_t = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to(device).eval()
 
-# Load GLM-4.1V-9B-Thinking
-MODEL_ID_S = "zai-org/GLM-4.1V-9B-Thinking"
+# Load Video-MTR
+MODEL_ID_S = "Phoebe13/Video-MTR"
 processor_s = AutoProcessor.from_pretrained(MODEL_ID_S, trust_remote_code=True)
-model_s = Glm4vForConditionalGeneration.from_pretrained(
+model_s = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_S,
     trust_remote_code=True,
     torch_dtype=torch.float16
@@ -95,7 +95,7 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     elif model_name == "Megalodon-OCR-Sync-0713":
         processor = processor_t
         model = model_t
-    elif model_name == "GLM-4.1V-9B-Thinking":
+    elif model_name == "Video-MTR":
         processor = processor_s
         model = model_s
     elif model_name == "ViLaSR-7B":
@@ -151,7 +151,7 @@ def generate_video(model_name: str, text: str, video_path: str,
     elif model_name == "Megalodon-OCR-Sync-0713":
         processor = processor_t
         model = model_t
-    elif model_name == "GLM-4.1V-9B-Thinking":
+    elif model_name == "Video-MTR":
         processor = processor_s
         model = model_s
     elif model_name == "ViLaSR-7B":
@@ -268,15 +268,18 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
                 with gr.Accordion("(Result.md)", open=False):
                     markdown_output = gr.Markdown(label="(Result.md)")
             model_choice = gr.Radio(
-                choices=["Camel-Doc-OCR-062825", "GLM-4.1V-9B-Thinking", "Megalodon-OCR-Sync-0713", "ViLaSR-7B"],
+                choices=["Camel-Doc-OCR-062825", "Video-MTR", "Megalodon-OCR-Sync-0713", "ViLaSR-7B"],
                 label="Select Model",
                 value="Camel-Doc-OCR-062825"
             )
             gr.Markdown("**Model Info 💻** | [Report Bug](https://huggingface.co/spaces/prithivMLmods/Multimodal-VLM-v1.0/discussions)")
-            gr.Markdown("> [Camel-Doc-OCR-062825](https://huggingface.co/prithivMLmods/Camel-Doc-OCR-062825), [GLM-4.1V-9B-Thinking](https://huggingface.co/zai-org/GLM-4.1V-9B-Thinking), [Megalodon-OCR-Sync-0713](https://huggingface.co/prithivMLmods/Megalodon-OCR-Sync-0713), and [ViLaSR-7B](https://huggingface.co/inclusionAI/ViLaSR) are recent vision-language models excelling in document intelligence and multimodal understanding. Camel-Doc-OCR-062825 is a Qwen2.5-VL-7B-Instruct finetune, highly optimized for document retrieval, structured extraction, analysis, and direct Markdown generation from images and PDFs. GLM-4.1V-9B-Thinking offers next-level multimodal reasoning, bringing visual and textual comprehension together for advanced question answering.")
-            gr.Markdown("> Megalodon-OCR-Sync-0713, finetuned from Qwen2.5-VL-3B-Instruct, specializes in context-aware multimodal document extraction and analysis, excelling at retrieval, layout parsing, math, and chart/table recognition, with robust video and long-form comprehension capabilities. ViLaSR-7B focuses on reinforcing spatial reasoning in visual-language tasks by combining interwoven thinking with visual drawing, making it especially suited for spatial reasoning and complex tip-based queries.")
-            gr.Markdown("> ✋ ViLaSR-7B - demo only supports text-only reasoning, which doesn't reflect the full behavior of the model and may underrepresent its capabilities.")            
 
+            gr.Markdown("> [Camel-Doc-OCR-062825](https://huggingface.co/prithivMLmods/Camel-Doc-OCR-062825) is a Qwen2.5-VL-7B-Instruct finetune, highly optimized for document retrieval, structured extraction, analysis, and direct Markdown generation from images and PDFs.")
+            gr.Markdown("> [Megalodon-OCR-Sync-0713](https://huggingface.co/prithivMLmods/Megalodon-OCR-Sync-0713), finetuned from Qwen2.5-VL-3B-Instruct, specializes in context-aware multimodal document extraction and analysis, excelling at retrieval, layout parsing, math, and chart/table recognition.")
+            gr.Markdown("> [ViLaSR-7B](https://huggingface.co/inclusionAI/ViLaSR) focuses on reinforcing spatial reasoning in visual-language tasks by combining interwoven thinking with visual drawing, making it especially suited for spatial reasoning and complex tip-based queries.")
+            gr.Markdown("> [Video-MTR](https://huggingface.co/Phoebe13/Video-MTR) introduces reinforced multi-turn reasoning for long-form video understanding, enabling iterative key segment selection and deeper question comprehension.")
+       
+            gr.Markdown("> ✋ ViLaSR-7B - demo only supports text-only reasoning, which doesn't reflect the full behavior of the model and may underrepresent its capabilities.")            
             gr.Markdown("> ⚠️ Note: Models in this space may not perform well on video inference tasks.")
     # Define the submit button actions
     image_submit.click(fn=generate_image,