Update app.py

app.py

@@ -54,6 +54,15 @@ 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.
@@ -94,6 +103,9 @@ 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
@@ -125,7 +137,6 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     buffer = ""
     for new_text in streamer:
         buffer += new_text
-        buffer = buffer.replace("<|im_end|>", "")
         time.sleep(0.01)
         yield buffer
 
@@ -148,6 +159,9 @@ 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
@@ -190,7 +204,6 @@ def generate_video(model_name: str, text: str, video_path: str,
     buffer = ""
     for new_text in streamer:
         buffer += new_text
-        buffer = buffer.replace("<|im_end|>", "")
         time.sleep(0.01)
         yield buffer
 
@@ -246,7 +259,7 @@ 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", "Captioner-7B"],
+                choices=["Cosmos-Reason1-7B", "docscopeOCR-7B-050425-exp", "Mimo-7B-RL", "Captioner-7B"],
                 label="Select Model",
                 value="Cosmos-Reason1-7B"
             )
@@ -254,8 +267,10 @@ 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],
@@ -268,4 +283,4 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
     )
 
 if __name__ == "__main__":
-    demo.queue(max_size=30).launch(share=True, ssr_mode=False, show_error=True)
+    demo.queue(max_size=40).launch(share=True, mcp_server=True, ssr_mode=False, show_error=True)