Dive Into Deep Learning - Lecture 5: Parameter Access, Initialization, and storage in PyTorch
In this video, we discuss how to extract the parameters of a neural network or deep learning model, i.e., weights and biases, to reuse them in some other context or to save them to disk. We use torch.nn to define neural network models in PyTorch and we discuss the use of state_dict, named_parameters, nn.init for initialization, and torch.save/torch.load for saving and loading model parameters in PyTorch.
#PyTorch #Parameters #Initialization