Update app.py

app.py

@@ -1,30 +1,26 @@
-import torch
-import torch.nn as nn
 from config import MLP
-import gradio as gr
+from mmcv import Config
+from mmdet.models import build_detector
+import torch
 
-# Load the mode
-model = MLP()
-model.load_state_dict(torch.load("pytorch_model.pth", map_location=torch.device("cpu")))
-model.eval()
+def main():
+    # Print model type from config
+    print(f"Model type: {MLP['type']}")
 
-# Define a prediction function
-def predict(inputs):
-    with torch.no_grad():
-        inputs = torch.tensor(inputs).float().unsqueeze(0)  # Add batch dimension
-        output = model(inputs)
-        if isinstance(output, torch.Tensor):
-            return output.squeeze().tolist()
-        return output  # fallback
+    # Build the model from the config dict
+    model = build_detector(MLP, train_cfg=MLP.get('train_cfg'), test_cfg=MLP.get('test_cfg'))
 
-# Create the Gradio interface
-demo = gr.Interface(
-    fn=predict,
-    inputs=gr.Textbox(label="Enter comma-separated input values (e.g., 1.2, 3.4, 5.6)"),
-    outputs=gr.Textbox(label="Model Output"),
-    title="PyTorch MLP Classifier"
-)
+    # Set model to evaluation mode
+    model.eval()
+
+    # Print model architecture summary
+    print(model)
+
+    # Optional: dummy input test (batch of 1 image with 3 channels, 800x1333)
+    dummy_input = torch.randn(1, 3, 800, 1333)
+    with torch.no_grad():
+        result = model.forward_dummy(dummy_input)
+        print("Forward pass output:", result)
 
-# Launch
 if __name__ == "__main__":
-    demo.launch()
+    main()