Update app.py

app.py

@@ -46,7 +46,7 @@ model_x = Qwen2_5_VLForConditionalGeneration.from_pretrained(
 ).to(device).eval()
 
 # Load Relaxed
-MODEL_ID_Z = "Ertugrul/Qwen2.5-VL-7B-Captioner-Relaxed"
+MODEL_ID_Z = "XiaomiMiMo/MiMo-VL-7B-RL"
 processor_z = AutoProcessor.from_pretrained(MODEL_ID_Z, trust_remote_code=True)
 model_z = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_Z,
@@ -54,15 +54,6 @@ model_z = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to(device).eval()
 
-# Load Mimo
-MODEL_ID_Mimo = "XiaomiMiMo/MiMo-VL-7B-RL"
-processor_Mimo = AutoProcessor.from_pretrained(MODEL_ID_Mimo, trust_remote_code=True)
-model_Mimo = Qwen2_5_VLForConditionalGeneration.from_pretrained(
-    MODEL_ID_Mimo,
-    trust_remote_code=True,
-    torch_dtype=torch.float16
-).to(device).eval()
-
 def downsample_video(video_path):
     """
     Downsamples the video to evenly spaced frames.
@@ -103,9 +94,6 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     elif model_name == "Captioner-7B":
         processor = processor_z
         model = model_z
-    elif model_name == "Mimo-7B-RL":
-        processor = processor_Mimo
-        model = model_Mimo
     else:
         yield "Invalid model selected."
         return
@@ -159,9 +147,6 @@ def generate_video(model_name: str, text: str, video_path: str,
     elif model_name == "Captioner-7B":
         processor = processor_z
         model = model_z
-    elif model_name == "Mimo-7B-RL":
-        processor = processor_Mimo
-        model = model_Mimo
     else:
         yield "Invalid model selected."
         return
@@ -230,7 +215,7 @@ css = """
 
 # Create the Gradio Interface
 with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
-    gr.Markdown("# **DocScope-R1**")
+    gr.Markdown("# **Cosmos-x-DocScope**")
     with gr.Row():
         with gr.Column():
             with gr.Tabs():
@@ -259,7 +244,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
         with gr.Column():
             output = gr.Textbox(label="Output", interactive=False, lines=2, scale=2)
             model_choice = gr.Radio(
-                choices=["Cosmos-Reason1-7B", "docscopeOCR-7B-050425-exp", "Mimo-7B-RL", "Captioner-7B"],
+                choices=["Cosmos-Reason1-7B", "docscopeOCR-7B-050425-exp", "Captioner-7B"],
                 label="Select Model",
                 value="Cosmos-Reason1-7B"
             )
@@ -267,10 +252,8 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
             gr.Markdown("**Model Info**")
             gr.Markdown("⤷ [Cosmos-Reason1-7B](https://huggingface.co/nvidia/Cosmos-Reason1-7B): understand physical common sense and generate appropriate embodied decisions.")
             gr.Markdown("⤷ [docscopeOCR-7B-050425-exp](https://huggingface.co/prithivMLmods/docscopeOCR-7B-050425-exp): optimized for document-level optical character recognition, long-context vision-language understanding.")
-            gr.Markdown("⤷ [MiMo-VL-7B-RL](https://huggingface.co/XiaomiMiMo/MiMo-VL-7B-RL): MiMo-7B language model, specifically optimized for complex reasoning tasks. Mixed On-policy Reinforcement Learning (MORL), a novel framework that seamlessly integrates diverse reward signals spanning perception accuracy.")
             gr.Markdown("⤷ [Captioner-Relaxed-7B](https://huggingface.co/Ertugrul/Qwen2.5-VL-7B-Captioner-Relaxed): build with hand-curated dataset for text-to-image models, providing significantly more detailed descriptions or captions of given images.")
 
-
     image_submit.click(
         fn=generate_image,
         inputs=[model_choice, image_query, image_upload, max_new_tokens, temperature, top_p, top_k, repetition_penalty],
@@ -283,4 +266,4 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
     )
 
 if __name__ == "__main__":
-    demo.queue(max_size=40).launch(share=True, mcp_server=True, ssr_mode=False, show_error=True)
+    demo.queue(max_size=30).launch(share=True, mcp_server=True, ssr_mode=False, show_error=True)