Update app.py

app.py

@@ -45,6 +45,16 @@ model_x = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     trust_remote_code=True,
     torch_dtype=torch.float16
 ).to(device).eval()
+
+# Load olmOCR-7B-0225-preview
+MODEL_ID_O = "allenai/olmOCR-7B-0225-preview"
+processor_o = AutoProcessor.from_pretrained(MODEL_ID_O, trust_remote_code=True)
+model_o = Qwen2VLForConditionalGeneration.from_pretrained(
+    MODEL_ID_O,
+    trust_remote_code=True,
+    torch_dtype=torch.float16
+).to(device).eval()
+
 def downsample_video(video_path):
     """
     Downsamples the video to evenly spaced frames.
@@ -82,6 +92,9 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     elif model_name == "DREX-062225-exp":
         processor = processor_x
         model = model_x
+    elif model_name == "olmOCR-7B-0225":
+        processor = processor_o
+        model = model_o
     else:
         yield "Invalid model selected."
         return
@@ -133,6 +146,9 @@ def generate_video(model_name: str, text: str, video_path: str,
     elif model_name == "DREX-062225-exp":
         processor = processor_x
         model = model_x
+    elif model_name == "olmOCR-7B-0225":
+        processor = processor_o
+        model = model_o
     else:
         yield "Invalid model selected."
         return
@@ -231,7 +247,7 @@ 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=["DREX-062225-exp", "VIREX-062225-exp"],
+                choices=["DREX-062225-exp", "VIREX-062225-exp", "olmOCR-7B-0225"],
                 label="Select Model",
                 value="DREX-062225-exp"
             )