update app

app.py

@@ -99,6 +99,7 @@ MODEL_ID_M = "prithivMLmods/docscopeOCR-7B-050425-exp"
 processor_m = AutoProcessor.from_pretrained(MODEL_ID_M, trust_remote_code=True)
 model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_M,
+    attn_implementation="flash_attention_2",
     trust_remote_code=True,
     torch_dtype=torch.float16
 ).to(device).eval()
@@ -108,6 +109,7 @@ MODEL_ID_X = "prithivMLmods/coreOCR-7B-050325-preview"
 processor_x = AutoProcessor.from_pretrained(MODEL_ID_X, trust_remote_code=True)
 model_x = Qwen2VLForConditionalGeneration.from_pretrained(
     MODEL_ID_X,
+    attn_implementation="flash_attention_2",
     trust_remote_code=True,
     torch_dtype=torch.float16
 ).to(device).eval()
@@ -125,6 +127,7 @@ processor_g = AutoProcessor.from_pretrained(
 
 model_g = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_G,
+    attn_implementation="flash_attention_2",
     trust_remote_code=True,
     subfolder=SUBFOLDER,
     torch_dtype=torch.float16
@@ -136,6 +139,7 @@ MODEL_ID_O = "prithivMLmods/Camel-Doc-OCR-080125"
 processor_o = AutoProcessor.from_pretrained(MODEL_ID_O, trust_remote_code=True)
 model_o = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_O,
+    attn_implementation="flash_attention_2",
     trust_remote_code=True,
     torch_dtype=torch.float16
 ).to(device).eval()
@@ -214,7 +218,7 @@ css = """
 """
 
 # Create the Gradio Interface
-with gr.Blocks(css=css, theme=steel_blue_theme) as demo:
+with gr.Blocks() as demo:
     gr.Markdown("# **core [OCR](https://huggingface.co/collections/prithivMLmods/multimodal-implementations)**", elem_id="main-title")
     with gr.Row():
         with gr.Column(scale=2):
@@ -249,4 +253,4 @@ with gr.Blocks(css=css, theme=steel_blue_theme) as demo:
     )
 
 if __name__ == "__main__":
-    demo.queue(max_size=50).launch(mcp_server=True, ssr_mode=False, show_error=True)
+    demo.queue(max_size=50).launch(css=css, theme=steel_blue_theme, mcp_server=True, ssr_mode=False, show_error=True)