# davidtvs/pytorch-lr-finder

A learning rate range test implementation in PyTorch

Repository: https://github.com/davidtvs/pytorch-lr-finder
Canonical: https://ross.abutalabs.com/products/pytorch-lr-finder
Language: Python
License: MIT
License Family: permissive
Topics: pytorch, learning-rate
Last push: 2025-06-24T20:03:09+00:00

## Health v2 (maintenance only)
Score: 35/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 28, release rhythm 8, longevity 100
- inputs: {"age_days": 2843, "days_push": 435, "days_rel": 711, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1008, forks 122 (observed 2026-08-28T04:03:12.566505+00:00)

## What it is
A PyTorch library implementing the learning rate range test from Leslie Smith's cyclical learning rates paper, including the fastai-tweaked variant. It sweeps learning rates during a pre-training run and plots loss versus learning rate to help pick an optimal learning rate.

## Use cases
- find the optimal learning rate for training a PyTorch model
- run a learning rate range test before training
- determine start and end learning rates for cyclical learning rate schedules
- plot loss vs learning rate to pick a good static LR
- tune hyperparameters for neural network training

## When to choose
- you train PyTorch models and want a principled way to pick a learning rate
- you use cyclical learning rates and need boundary values
- you want a lightweight pip-installable LR finder with optional mixed-precision support

## When to avoid
- you use TensorFlow, JAX, or other non-PyTorch frameworks
- you need full hyperparameter search beyond learning rate
- you want an automatic tuner rather than a diagnostic plot

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, benchmarking, developer-tools
- domain: deep-learning, machine-learning, developer-tools
- platform: python, cross-platform
- tags: pytorch, learning-rate-finder, lr-range-test, hyperparameter-tuning, training

## Member repositories
- davidtvs/pytorch-lr-finder (main) score 35

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:12.566505+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:12:08.954634+00:00, confidence not recorded.
  - readme: https://github.com/davidtvs/pytorch-lr-finder (fetched 2026-08-28T04:03:12.566505+00:00, sha 0a88b5c4d5bf)
- Data as of 2026-08-30T08:39:29.467469+00:00.