Update app.py

app.py

@@ -16,6 +16,7 @@ import cv2
 
 from transformers import (
     Qwen2_5_VLForConditionalGeneration,
+    Qwen2VLForConditionalGeneration,
     AutoProcessor,
     AutoTokenizer,
     TextIteratorStreamer,
@@ -51,12 +52,22 @@ model_z = Qwen2_5_VLForConditionalGeneration.from_pretrained(
 # Load coreOCR-7B-050325-preview
 MODEL_ID_K = "prithivMLmods/coreOCR-7B-050325-preview"
 processor_k = AutoProcessor.from_pretrained(MODEL_ID_K, trust_remote_code=True)
-model_k = Qwen2_5_VLForConditionalGeneration.from_pretrained(
+model_k = Qwen2VLForConditionalGeneration.from_pretrained(
     MODEL_ID_K,
     trust_remote_code=True,
     torch_dtype=torch.float16
 ).to(device).eval()
 
+# Load Imgscope-OCR-2B-0527
+MODEL_ID_Y = "prithivMLmods/Imgscope-OCR-2B-0527"
+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.
@@ -97,6 +108,9 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     elif model_name == "coreOCR-7B-050325-preview":
         processor = processor_k
         model = model_k
+    elif model_name == "Imgscope-OCR-2B-0527":
+        processor = processor_y
+        model = model_y
     else:
         yield "Invalid model selected."
         return
@@ -150,6 +164,9 @@ def generate_video(model_name: str, text: str, video_path: str,
     elif model_name == "coreOCR-7B-050325-preview":
         processor = processor_k
         model = model_k
+    elif model_name == "Imgscope-OCR-2B-0527":
+        processor = processor_y
+        model = model_y
     else:
         yield "Invalid model selected."
         return
@@ -250,7 +267,7 @@ 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=["SkyCaptioner-V1", "SpaceThinker-3B", "coreOCR-7B-050325-preview"],
+                choices=["SkyCaptioner-V1", "SpaceThinker-3B", "coreOCR-7B-050325-preview", "Imgscope-OCR-2B-0527"],
                 label="Select Model",
                 value="SkyCaptioner-V1"
             )