Update app.py

app.py

@@ -16,7 +16,6 @@ import cv2
 from transformers import (
     Qwen2_5_VLForConditionalGeneration,
     Qwen2VLForConditionalGeneration,
-    Llama4ForConditionalGeneration,
     AutoProcessor,
     TextIteratorStreamer,
 )
@@ -29,10 +28,10 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
-# Load meta-llama/Llama-Guard-4-12B
-MODEL_ID_M = "meta-llama/Llama-Guard-4-12B"
+# Load typhoon
+MODEL_ID_M = "scb10x/typhoon-ocr-7b"
 processor_m = AutoProcessor.from_pretrained(MODEL_ID_M, trust_remote_code=True)
-model_m = Llama4ForConditionalGeneration.from_pretrained(
+model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_M,
     trust_remote_code=True,
     torch_dtype=torch.float16
@@ -263,9 +262,9 @@ 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=["Llama-4", "coreOCR-7B-050325-preview", "Imgscope-OCR-2B", "Qwen2.5-VL-3B"],
+                choices=["coreOCR-7B-050325-preview", "typhoon-ocr-7b", "Qwen2.5-VL-3B", "Imgscope-OCR-2B"],
                 label="Select Model",
-                value="Llama-4"
+                value="coreOCR-7B-050325-preview"
             )
             
     image_submit.click(