pytorch/ignite

Improve distributed tests for Confusion Matrix (CM and MLCM)

Open

#1,657 opened on Feb 20, 2021

View on GitHub
 (6 comments) (0 reactions) (0 assignees)Python (4,313 stars) (602 forks)batch import
HacktoberfestPyDataGlobalenhancementhelp wantedmodule: metrics

Description

Currently, the testing data is generated without taking into account distributed context: https://github.com/pytorch/ignite/blob/e17acc740d38909ddbbed6648423750abe593e58/tests/ignite/metrics/test_confusion_matrix.py#L559-L565

same for MultiLabelConfusionMatrix tests.

We would like to rewrite the tests in a similar way as here : https://github.com/pytorch/ignite/blob/e17acc740d38909ddbbed6648423750abe593e58/tests/ignite/metrics/test_recall.py#L778-L785

See https://github.com/pytorch/ignite/issues#issuecomment-782750122

cc @touqir14

Contributor guide

Tech stack
pythonpytorch
Domain
testingmachine learning
Issue type
test
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
PyTorch basicsIgnite test structuredistributed testing knowledgeconfusion matrix metrics
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60
Research direction
The issue requires rewriting the test data generation for ConfusionMatrix and MultiLabelConfusionMatrix to properly simulate distributed context. The existing tests in tests/ignite/metrics/test confusion matrix.py (lines 559-565) need to be replaced with a pattern similar to tests/ignite/metrics/test recall.py (lines 778-785). The contributor should study the distributed test setup in the linked example and ensure the metrics compute correctly across multiple processes.