pytorch/ignite

Unify metrics output type

Open

#1,635 opened on Feb 12, 2021

View on GitHub
 (12 comments) (0 reactions) (0 assignees)Python (602 forks)batch import
HacktoberfestPyDataGlobaldocsenhancementhelp wantedmodule: metricsneeds-discussion

Repository metrics

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

Description

🚀 Feature

The idea is to verify the output type for all metrics (output of compute function) and update the docs accordingly.

In general, metric's output should be a float number. In some particular cases, like Recall/Precision with average=False, the output is a torch tensor. So, let's see and decide if the output of compute() method can be :

  • def compute() -> float
  • def compute() -> Union[float, torch.Tensor] and tensor is on CPU
  • def compute() -> torch.Tensor with tensor on CPU

To address this FR, we have to make sure for each metric what kind of type it supposes to return and update the docs accordingly.

Contributor guide

Research direction
Start by listing all metrics in the repository under ignite/metrics/. For each metric, examine the compute method's return type and relevant tests. Then decide on the unified type based on the options provided. Update the documentation in docs/source/metrics.rst accordingly.
Tech stack
pythonpytorch
Domain
apibackend
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.
3-5 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
PyTorch basicsPython type hints
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
35