update app

app.py

@@ -222,6 +222,15 @@ model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to(device).eval()
 
+# Load Nanonets-OCR2-1.5B-exp
+MODEL_ID_Y = "nanonets/Nanonets-OCR2-1.5B-exp"
+processor_y = AutoProcessor.from_pretrained(MODEL_ID_Y, trust_remote_code=True)
+model_y = Qwen2VLForConditionalGeneration.from_pretrained(
+    MODEL_ID_Y,
+    trust_remote_code=True,
+    torch_dtype=torch.float16
+).to(device).eval()
+
 def downsample_video(video_path):
     """
     Downsamples the video to evenly spaced frames.
@@ -270,6 +279,9 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     elif model_name == "olmOCR-7B-0725":
         processor = processor_w
         model = model_w
+    elif model_name == "Nanonets-OCR2-1.5B-exp":
+        processor = processor_y
+        model = model_y
     else:
         yield "Invalid model selected.", "Invalid model selected."
         return
@@ -332,6 +344,9 @@ def generate_video(model_name: str, text: str, video_path: str,
     elif model_name == "olmOCR-7B-0725":
         processor = processor_w
         model = model_w
+    elif model_name == "Nanonets-OCR2-1.5B-exp":
+        processor = processor_y
+        model = model_y
     else:
         yield "Invalid model selected.", "Invalid model selected."
         return
@@ -424,7 +439,7 @@ with gr.Blocks(css=css, theme=thistle_theme) as demo:
                     markdown_output = gr.Markdown(label="(Result.Md)")
                                    
                 model_choice = gr.Radio(
-                    choices=["Nanonets-OCR2-3B", "olmOCR-7B-0725", "RolmOCR-7B", 
+                    choices=["Nanonets-OCR2-3B", "olmOCR-7B-0725", "RolmOCR-7B", "Nanonets-OCR2-1.5B-exp",
                          "Aya-Vision-8B", "Qwen2-VL-OCR-2B"],
                     label="Select Model",
                     value="Nanonets-OCR2-3B"