Update app.py

app.py

@@ -56,8 +56,8 @@ model_o = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16).to(device).eval()
 
 #-----------------------------subfolder-----------------------------#
-# Load MonkeyOCR-1.2B-0709
-MODEL_ID_W = "echo840/MonkeyOCR-1.2B-0709"
+# Load MonkeyOCR-pro-1.2B
+MODEL_ID_W = "echo840/MonkeyOCR-pro-1.2B"
 SUBFOLDER = "Recognition"
 processor_w = AutoProcessor.from_pretrained(MODEL_ID_W, trust_remote_code=True, subfolder=SUBFOLDER)
 model_w = Qwen2_5_VLForConditionalGeneration.from_pretrained(
@@ -113,7 +113,7 @@ def generate_image(model_name: str,
     elif model_name == "R1-Onevision-7B":
         processor = processor_t
         model = model_t
-    elif model_name == "MonkeyOCR-1.2B-0709":
+    elif model_name == "MonkeyOCR-pro-1.2B":
         processor = processor_w
         model = model_w
     else:
@@ -180,7 +180,7 @@ def generate_video(model_name: str,
     elif model_name == "R1-Onevision-7B":
         processor = processor_t
         model = model_t
-    elif model_name == "MonkeyOCR-1.2B-0709":
+    elif model_name == "MonkeyOCR-pro-1.2B":
         processor = processor_w
         model = model_w
     else:
@@ -332,13 +332,13 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
 
             model_choice = gr.Radio(choices=[
                  "Vision-Matters-7B", "R1-Onevision-7B",
-                 "ViGaL-7B", "MonkeyOCR-1.2B-0709", "Visionary-R1-3B"
+                 "ViGaL-7B", "MonkeyOCR-pro-1.2B", "Visionary-R1-3B"
             ],
                                     label="Select Model",
                                     value="Vision-Matters-7B")
 
             gr.Markdown("**Model Info 💻** | [Report Bug](https://huggingface.co/spaces/prithivMLmods/Multimodal-VLMs-5x/discussions)")          
-            gr.Markdown("> [MonkeyOCR-1.2B-0709](https://huggingface.co/echo840/MonkeyOCR-1.2B-0709): MonkeyOCR adopts a structure-recognition-relation (SRR) triplet paradigm, which simplifies the multi-tool pipeline of modular approaches while avoiding the inefficiency of using large multimodal models for full-page document processing.")
+            gr.Markdown("> [MonkeyOCR-pro-1.2B](https://huggingface.co/echo840/MonkeyOCR-pro-1.2B): MonkeyOCR adopts a structure-recognition-relation (SRR) triplet paradigm, which simplifies the multi-tool pipeline of modular approaches while avoiding the inefficiency of using large multimodal models for full-page document processing.")
             gr.Markdown("> [Vision Matters 7B](https://huggingface.co/Yuting6/Vision-Matters-7B): vision-matters is a simple visual perturbation framework that can be easily integrated into existing post-training pipelines including sft, dpo, and grpo. our findings highlight the critical role of visual perturbation: better reasoning begins with better seeing.")
             gr.Markdown("> [ViGaL 7B](https://huggingface.co/yunfeixie/ViGaL-7B): vigal-7b shows that training a 7b mllm on simple games like snake using reinforcement learning boosts performance on benchmarks like mathvista and mmmu without needing worked solutions or diagrams indicating transferable reasoning skills.")
             gr.Markdown("> [Visionary-R1](https://huggingface.co/maifoundations/Visionary-R1): visionary-r1 is a novel framework for training visual language models (vlms) to perform robust visual reasoning using reinforcement learning (rl). unlike traditional approaches that rely heavily on (sft) or (cot) annotations, visionary-r1 leverages only visual question-answer pairs and rl, making the process more scalable and accessible.")