Update app.py

app.py

@@ -13,15 +13,13 @@ print("Loading model...")
 model = AutoModel.from_pretrained(
     model_name,
     trust_remote_code=True,
-    _attn_implementation="flash_attention_2",
-    use_safetensors=True
-).cuda().to(torch.bfloat16).eval()
+    use_safetensors=True,
+    low_cpu_mem_usage=True
+).cuda().eval()
 
 def ocr_infer(image):
-    # Save temporary image
     image.save("input.png")
 
-    # Standard OCR prompt
     prompt = "<image>\nFree OCR."
 
     result = model.infer(
@@ -29,20 +27,18 @@ def ocr_infer(image):
         prompt=prompt,
         image_file="input.png",
         output_path=".",
-        base_size=1024,
         image_size=640,
-        crop_mode=True,
-        save_results=False
+        crop_mode=True
     )
-    
+
     return result["text"]
 
 demo = gr.Interface(
     fn=ocr_infer,
     inputs=gr.Image(type="pil"),
     outputs="text",
-    title="DeepSeek-OCR API (HuggingFace)",
-    description="Upload an image and get OCR text using DeepSeek-OCR"
+    title="DeepSeek-OCR (HF Space)",
+    description="GPU OCR API"
 )
 
 demo.launch()