# facebookresearch/fvcore

Collection of common code that's shared among different research projects in FAIR computer vision team.

Repository: https://github.com/facebookresearch/fvcore
Canonical: https://ross.abutalabs.com/products/fvcore
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-19T05:36:55+00:00

## Health v2 (maintenance only)
Score: 76/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 35, longevity 100
- inputs: {"age_days": 2534, "days_push": 14, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2250, forks 237 (observed 2026-08-28T04:06:30.948024+00:00)

## What it is
fvcore is a lightweight Python core library providing common functionality shared across FAIR's computer vision frameworks such as Detectron2, PySlowFast, and ClassyVision. It includes PyTorch layers, losses, hierarchical per-operator FLOP counting, parameter counting, BatchNorm statistics recomputation, and a scale-invariant hyperparameter scheduler.

## Use cases
- count FLOPs of a PyTorch model per operator
- count model parameters recursively
- recompute BatchNorm population statistics
- schedule learning rate hyperparameters in a scale-invariant way
- share common layers and losses across computer vision projects

## When to choose
- you use Detectron2, PySlowFast, or ClassyVision and need their shared core utilities
- you need to benchmark model compute cost (FLOPs) and parameter counts in PyTorch
- you want a tested, type-annotated utility library for vision research

## When to avoid
- you need a full computer vision framework rather than shared utilities
- your project does not use PyTorch
- you need general-purpose ML tooling outside computer vision

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, benchmarking, developer-tools
- domain: computer-vision, deep-learning, machine-learning, developer-tools
- platform: python, cross-platform
- tags: pytorch, flop-counting, parameter-counting, fair, detectron2, model-analysis, utilities, gpu

## Member repositories
- facebookresearch/fvcore (main) score 76

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:30.948024+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 2026-08-30T02:43:51.918915+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/fvcore (fetched 2026-08-28T04:06:30.948024+00:00, sha 4a88a3064d7e)
  - registry_pypi: https://pypi.org/pypi/fvcore/json (fetched 2026-08-29T10:24:03.613607+00:00, sha 0217a6086654)
- Data as of 2026-08-30T08:39:29.467469+00:00.