NVIDIA / NVIDIA/apex

Disable AMP in particular layers

Open
#1,085 1 comment 0 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

Hi, thanks for your great work!
I notice that mixed precision is not very good for transpose convolution layers and up-sampling layers (https://github.com/NVIDIA/apex/issues/82#issuecomment-450304518).
However, I want to use AMP in U-Net 3D, so is there any way to off the mixed precision calculation in particular layers of my model?
Like this: set myUnet.transpose_layers into FP32 instead of mixed precision calculation.

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

No file, test, or entry point is named. Start by locating how Apex AMP applies mixed precision to model layers, then determine how per-layer FP32 selection should work for transpose convolution and up-sampling layers and define tests for the expected behavior.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
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.