I had google and read various error of illegal memory access was encountered. But before I got this error, I was able to run just fine for weeks prior. All I am doing is a simple binary classification using Pytoruch and Cuda 12.6 (was 12.4 before) and encounter this error. I had contact ASUS support and they suspect this is an Pytorch issue. As such, I here looking for help.
Here is the output of the error with this fragment of code:
num_epochs = 20
# We use the pre-defined number of epochs to determine how many iterations to train the network on
for epoch in range(num_epochs):
print("Computing Epoch " + str(epoch) + "...")
# Load in the data in batches using the train_dataloader object
for i, (images, labels) in enumerate(train_dataloader):
# Move tensors to the configured device
images = images.to(device)
labels = labels.to(device)
# Forward pass
outputs = model(images)
loss = criterion(outputs, labels)
# Backward and optimize
optimizer.zero_grad()
loss.backward()
optimizer.step()
print('Epoch [{}/{}], Loss: {:.4f}'.format(epoch+1, num_epochs, loss.item()))
Error Output:
Any help would be appreciated. I can also provide the test.py if required as it is a basic binary classification with random generated image. And also the output from “compute-sanitizer python .\test.py”.