-
Notifications
You must be signed in to change notification settings - Fork 443
Open
Description
I found two problems in the following official website documents
Document address: https://learn.microsoft.com/en-us/windows/ai/windows-ml/tutorials/pytorch-train-model
The first problem is that the code in the following picture example does not call the model.train() method to train the model in each epoch, because the train() method calls the testAccuracy() method and the testAccuracy() method calls the model.eval() method, which will cause the model to be trained through model.eval() in subsequent model training. In this case, if there is a Dropout layer or a BatchNorm layer in the CNN network, this will cause both of them to not work.

The second problem is that the text in the following picture shows that CNN is written as CCN.

Metadata
Metadata
Assignees
Labels
No labels