pytorch / pytorch/vision

Add `channels_last` argument to `transforms.ToTensor()`

Open
#7,090 13 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
17.9k
Forks
7.3k
Avg merge
1d 15h
Merged PRs (30d)
13

Description

🚀 The feature

Currently, for using channels_last:

  • During data loading, torchvision.transforms.ToTensor() converts the input images into CHW.
  • Then during training, the images are converted back to HWC: images = images.to(memory_format=torch.channels_last).

This is highly inefficient. I suggest to add an optional boolean argument to ToTensor() that skips the permutation and outputs directly HWC:

torchvision.transforms.ToTensor(channels_last=True)

Motivation, pitch

To avoid unnecessary axis permutations and improve efficiency.

Alternatives

I had to write my custom functional.to_tensor to add the suggested argument.

Additional context

No response

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reading torchvision.transforms.ToTensor() and the functional.to_tensor path, including the existing behavior that converts images to CHW. Determine how the optional channels_last setting should affect the output and verify the result against the requested efficiency goal; the issue does not name a test file.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.