Ben Chuanlong Du's Blog

It is never too late to learn.

Make Your Model Training Reproducible in PyTorch

Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!

The PyTorch doc Reproducibility has very detailed instructions on how to make your model training reproducible. Basically, you need the following code.

torch.manual_seed(args.seed)
np.random.seed(args.seed …

Delete a Layer in a Pretrained Model in PyTorch

It is common to customize a pretrained model by delete the output layer or replace it to the output layer that suits your use case. There are several ways to achieve this in PyTorch.

Replace the Fully Connected Layer with an Identity Layer

  1. Define an identity layer.

  2. Replace the fully connected layer with an Identity Layer (using ResNet18 as an example).

Save and Load PyTorch Models

Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!

  1. PyTorch uses pickle to serialize and deserialize objects.

  2. The PyTorch convention is to use the file extension .pt or .pth for saving model (or its parameters) and use the file extension …