pytorch/ignite

More EpochMetric's compute_fn output types

Open

#1,757 opened on Mar 8, 2021

View on GitHub
 (5 comments) (0 reactions) (0 assignees)Python (602 forks)batch import
HacktoberfestPyDataGlobalenhancementhelp wantedmodule: metrics

Repository metrics

Stars
 (4,313 stars)
PR merge metrics
 (Avg merge 15d 11h) (17 merged PRs in 30d)

Description

🚀 Feature

Following https://github.com/pytorch/ignite/pull/1700#issuecomment-787151203 we would like to improve EpochMetric by supporting more output types returned by compute_fn:

  • EpochMetric should work with compute_fn outputs: scalars, tensors, tuple/list/mapping of tensors vs currently only scalars

A code snippet on possible implementation: https://github.com/pytorch/ignite/pull/1700#issuecomment-790413236 (It is not tested and maybe wrong, just an idea of how to do it).

It would be also helpful to raise a more user friendly error message if output type is not supported.

We need to update the docs as well.

If interested in fixing this issue, please leave a comment to assign the issue to you.

Contributor guide

Research direction
Review the existing implementation of EpochMetric in the ignite codebase (likely in ignite/metrics/epoch metric.py). Understand how compute fn is currently used and the output types it supports. Examine the code snippet in the linked comment for a potential approach. Then implement support for scalars, tensors, and tuple/list/mapping of tensors. Also add a more user friendly error message for unsupported types. Finally, update the documentation to reflect the new capabilities.
Tech stack
pythonpytorch
Domain
machine learning
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.
1-2 days
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.
Mostly clear
Prerequisites
Familiarity with PyTorchUnderstanding of EpochMetric class
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60