pytorch/ignite

enable automatic mixed precision for xla

Open

#1,931 opened on Apr 12, 2021

View on GitHub
 (8 comments) (1 reaction) (0 assignees)Python (4,313 stars) (602 forks)batch import
enhancementhelp wanted

Description

Feature

Automatic mixed precision for xla has landed in pytorch 1.8.1 and torch/xla nightly. We should enable it in create_supervised_* helper functions.

Suggested solution

Remove xla and amp checks in _check_arg().

  • For create_supervised_trainer, update supervised_training_step_tpu() function to accept scaler argument just like supervised_training_step_amp().
  • For create_supervised_evaluator, just removing xla and amp checks in _check_arg() should work.
  • For tests, we could remove xla checks and only run with pytorch 1.8.1.

Additional context

This feature should not be included in ignite release until the next torch and xla comes out.

Contributor guide

Tech stack
pythonpytorch
Domain
machine learningperformance
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
PyTorchIgniteAMPXLA
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50
Research direction
Implement the changes as described in the issue: remove XLA and AMP checks in ` check arg()` in `ignite/engine/ init .py`. Update `supervised training step tpu()` to accept a `scaler` argument similar to `supervised training step amp()`. For the evaluator, just removing checks should suffice. Also update tests to run only with PyTorch 1.8.1.
enable automatic mixed precision for xla · pytorch/ignite#1931 | Good First Issue