facebookresearch/metaseq

Remove apex dependency after PyTorch 2.0?

Open

#617 opened on Jan 24, 2023

View on GitHub
 (3 comments) (0 reactions) (1 assignee)Python (6,195 stars) (701 forks)batch import
better-enggood first issue

Description

This is to look into whether or not we still need to use apex for speedups if out-of-the-box PyTorch 2.0 may "just work". Will require benchmarking at a few different scales to confirm.

Contributor guide

Tech stack
pythonpytorch
Domain
performance
Issue type
research
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.
1-2 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
blocked
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
PythonPyTorchapex librarybenchmarking
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
15
Research direction
This issue investigates whether the apex library is still necessary with PyTorch 2.0. The investigation requires benchmarking training speed and memory usage at multiple scales (e.g., small, medium, large models). Relevant files include training scripts and configuration files in the repository. The assignee may have started work, so check recent branches or commits.