NVIDIA / NVIDIA/apex

test_affine_grid fails w/ PyTorch 1.3.x: linspace_cuda does not support Half

Open
#628 0 comments 2 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
9k
Forks
1.5k
Avg merge
2d 4h
Merged PRs (30d)
3

Description

Test test_affine_grid (in test_pyprof_nvtx.py) will fail with PyTorch 1.3.x with a complaint that linspace_node does not support Half:

RuntimeError: "linspace_cuda" not implemented for 'Half'

This is caused by a refactoring on PyTorch's affine_grid() that removed cuDNN support in favor of a native implementation.

Reported to PyTorch upstream as: https://github.com/pytorch/pytorch/issues/30404

Contributor guide

No contributing guide indexed for this repository

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

Read test_pyprof_nvtx.py and run test_affine_grid with PyTorch 1.3.x to reproduce the linspace_cuda Half failure. Review the linked PyTorch issue for the compatibility status and determine the project-supported resolution; done means the test no longer fails on the affected version.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
testing
Issue type
Bug
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.