Fix app.py 3

app.py

@@ -1,5 +1,4 @@
 import torch
-import onnx
 import onnxruntime as rt
 from torchvision import transforms as T
 from PIL import Image
@@ -15,26 +14,20 @@ charset = r"0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ!\"#$%
 tokenizer_base = Tokenizer(charset)
 
 def get_transform(img_size):
-    transforms = [
+    return T.Compose([
         T.Resize(img_size, T.InterpolationMode.BICUBIC),
         T.ToTensor(),
         T.Normalize(0.5, 0.5)
-    ]
-    return T.Compose(transforms)
+    ])
 
-def to_numpy(tensor):
-    return tensor.detach().cpu().numpy() if tensor.requires_grad else tensor.cpu().numpy()
+transform = get_transform(img_size)
+ort_session = rt.InferenceSession(model_file)
 
-def initialize_model(model_file):
-    transform = get_transform(img_size)
-    ort_session = rt.InferenceSession(model_file)
-    return transform, ort_session 
-
-def get_text(img_org):
+def predict(img):
     try:
-        if img_org is None: return ""
-        x = transform(img_org.convert('RGB')).unsqueeze(0)
-        ort_inputs = {ort_session.get_inputs()[0].name: to_numpy(x)}
+        if img is None: return "Sube una imagen"
+        x = transform(img.convert('RGB')).unsqueeze(0)
+        ort_inputs = {ort_session.get_inputs()[0].name: x.detach().cpu().numpy()}
         logits = ort_session.run(None, ort_inputs)[0]
         probs = torch.tensor(logits).softmax(-1)
         preds, _ = tokenizer_base.decode(probs)
@@ -42,15 +35,14 @@ def get_text(img_org):
     except Exception as e:
         return str(e)
 
-transform, ort_session = initialize_model(model_file=model_file)
-
-# INTERFAZ MÍNIMA (Sin ejemplos para evitar el bug del log)
-demo = gr.Interface(
-    fn=get_text,
-    inputs=gr.Image(type="pil"),
-    outputs=gr.Textbox(),
-    title="Captcha Reader API"
-)
+# Usamos Blocks para evitar el generador automático de API que está roto
+with gr.Blocks() as demo:
+    gr.Markdown("# Captcha Solver")
+    with gr.Row():
+        input_img = gr.Image(type="pil")
+        output_text = gr.Textbox()
+    btn = gr.Button("Resolver")
+    btn.click(fn=predict, inputs=input_img, outputs=output_text)
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860, show_api=False)
+    demo.launch(server_name="0.0.0.0", server_port=7860)