debugging

app.py

@@ -9,39 +9,39 @@ import matplotlib.pyplot as plt
 import os
 
 # Load the model
-checkpoints_dir = Path("saved_checkpoints")
-if not checkpoints_dir.exists():
-	os.system("aws s3 --no-sign-request cp --recursive s3://dataforgood-fb-data/forests/v1/models/ .")
-checkpoint = "SSLhuge_satellite.pth"
-device = "cpu"
-ckpt_path = checkpoints_dir / checkpoint
-model = SSLModule(ssl_path=str(ckpt_path))
-model.to(device)
-model = model.eval()
+#checkpoints_dir = Path("saved_checkpoints")
+#if not checkpoints_dir.exists():
+#	os.system("aws s3 --no-sign-request cp --recursive s3://dataforgood-fb-data/forests/v1/models/ .")
+#checkpoint = "SSLhuge_satellite.pth"
+#device = "cpu"
+#ckpt_path = checkpoints_dir / checkpoint
+#model = SSLModule(ssl_path=str(ckpt_path))
+#model.to(device)
+#model = model.eval()
 
 # Define the normalization transform
-norm = T.Normalize((0.420, 0.411, 0.296), (0.213, 0.156, 0.143))
-norm = norm.to(device)
+#norm = T.Normalize((0.420, 0.411, 0.296), (0.213, 0.156, 0.143))
+#norm = norm.to(device)
 
 
 # Define a function to make predictions
 def predict(image):
     # Convert PIL Image to tensor
-    image_t = torch.tensor(image).permute(2, 0, 1)[:3].float().to(device) / 255
-    # Normalize the image
-    with torch.no_grad():
-        pred = model(norm(image_t.unsqueeze(0)))
-        pred = pred.cpu().detach().relu()
-    # Convert tensor to numpy array
-    pred_np = pred[0, 0].numpy()
-    # Save the image to an in-memory buffer
-    buffer = BytesIO()
-    plt.imsave(buffer, pred_np, cmap="Greens")
-    buffer.seek(0)  # Rewind the buffer to the beginning
-    # Read the image back from the buffer
-    image_from_buffer = Image.open(buffer)
-    return image_from_buffer
-
+#    image_t = torch.tensor(image).permute(2, 0, 1)[:3].float().to(device) / 255
+#    # Normalize the image
+#    with torch.no_grad():
+#        pred = model(norm(image_t.unsqueeze(0)))
+#        pred = pred.cpu().detach().relu()
+#    # Convert tensor to numpy array
+#    pred_np = pred[0, 0].numpy()
+#    # Save the image to an in-memory buffer
+#    buffer = BytesIO()
+#    plt.imsave(buffer, pred_np, cmap="Greens")
+#    buffer.seek(0)  # Rewind the buffer to the beginning
+#    # Read the image back from the buffer
+#    image_from_buffer = Image.open(buffer)
+#    return image_from_buffer
+    return image
 
 # create a Gradio interface
 demo = gr.Interface(