AD Teaching Wiki:

A short overview over the PyTorch operations most relevant for this course. The full documentation can be accessed here.

Operation

Description

torch.empty

Create a tensor of a specific shape with uninitialized data

torch.zeros

Create a tensor of a specific shape and fill it with zeros

torch.ones

Create a tensor of a specific shape and fill it with ones

torch.tensor

Create a tensor with the given data

torch.sparse_coo_tensor

Create a sparse tensor with the given indices and values

torch.matmul

Perform matrix multiplication

torch.nonzero

Find the indices where a tensor contains non-zero values

torch.sort

Return the sorted values and sorting permutation of a tensor

torch.stack

Concatenate a list of tensors along a new dimension

torch.cat

Concatenate a list of tensors along an existing dimension

torch.sum

Sum a tensor along one or more dimensions

torch.mean

Average a tensor along one or more dimensions

torch.norm

Calculate the norm of tensor along the given dimensions

torch.where

Return a new tensor containing values from one of two sources depending on a condition

torch.topk

Get top k largest or smallest values and their indices from a tensor

torch.sigmoid

Apply the sigmoid function to all values of a tensor

torch.softmax

Perform softmax normalization

torch.index_select

Select indices from a tensor along a specific dimension

torch.arange

Create a tensor from a range

torch.multinomial

Sample from a multinomial distribution

AD Teaching Wiki: PyTorchCheatSheet (last edited 2024-01-30 10:53:54 by adpult)