Public record
Software health reportschema 0.15.0 · metrics 2.10.0 · 2026-07-19 11:41 UTC

meta-pytorch / torchft

Fault tolerance for PyTorch (HSDP, LocalSGD, DiLoCo, Streaming DiLoCo)

Python · RustCustom license★ 521 stars⑂ 70 forkssince Oct 2024View on GitHub ↗
KindNetwork serviceCommand-line toolhow this is determined
#fault-tolerance#ml#pytorch#torch#training#consensus#distributed-systems#fsdp#hsdp#localsgd#diloco

meta-pytorch/torchft holds a health index of 71 out of 100, placing it in the Good band. It scores highest on Engineering Quality (84/100) and lowest on Vitality (46/100). It was last updated 2 days ago. A single contributor accounts for most of its recent work.

71
overall / 100
Good

Software health index

Metrics are grouped into weighted categories on one standardized 1–100 scale. Overall starts as their weighted mean, calibrated against the distribution of the public record so bands carry percentile meaning; when public evidence triggers the High-Risk Jurisdiction Policy, the rating is adjusted and receives an At Risk ceiling of 34.

71
Exceptional93-100The record's top tier (≈ top 5%); essentially all checked criteria met
Excellent80-92Strong across the board; minor gaps
Good65-79Healthy; gaps are limited and manageable
Moderate50-64Acceptable with notable gaps; review recommended
Weak35-49Material weaknesses across several areas
At Risk20-34Significant weaknesses; adoption warrants caution
Critical1-19Severe problems (abandoned, single-maintainer, no hygiene)
VitalityCommunity &AdoptionSustainability &GovernanceEngineeringQualitySecurityAI Readiness

Score profile

Each axis is a category. The shape matters more than the average — a healthy subject fills the whole shape, while a spike-and-crater profile means strength in one dimension is masking risk in another.

The weighted overall 63 is calibrated to 71 on the published index scale (record calibration 2026-08-02).

Ownership

Meta PyTorchOrganization
1,389 followers58 public repossince Jun 2022

This repository is backed by an organization — shared, accountable stewardship that can outlive any single maintainer.

Package ecosystems

RegistryPackageVersionDownloads / moVersionsLast publish
PyPItorchftpoints to another repo — not scored0.1.13182576 days ago

Metrics by category

Vitality

Is the project alive — is code being written and are releases shipping?

46Weak · 21% of overall

Development activity

76Good
How it's scored
36/36Push recency — last push 2 days ago
16.6/36Commit cadence — 24/52 weeks with commits
16.3/18Commit volume — 64 commits in the last year
7/10OpenSSF Scorecard: Maintained — 8 commit(s) and 1 issue activity found in the last 90 days -- score normalized to 7
Inputs used
commits_last_year64
human_commit_share—
days_since_last_push2
active_weeks_last_year24

Release discipline

1Critical
How it's scored
0/27Ships releases — no releases published
0/36Release recency — no releases
0/27Release cadence — no releases
0/10OpenSSF Scorecard: Signed-Releases — no data
Inputs used
releases_count0
Excluded from scoring (no data or not applicable): OpenSSF Scorecard: Signed-Releases. Remaining weights renormalized.

Community & Adoption

Does the project have users, downloads, attention, and a welcoming setup for contributors?

72Good · 17% of overall

Popularity & adoption

66Good
How it's scored
44.1/60Stars — 521 stars
15.3/25Forks — 70 forks
6.7/15Watchers — 17 watchers
Inputs used
forks70
stars521
watchers17
growth_stateunverified
growth_factor_pct100
growth_unverified_reasonno_history

Community health

79Good
How it's scored
22.5/22.5README
16.9/22.5License — license file present, not a recognized license
18/18CONTRIBUTING guide
13.5/13.5Code of conduct
0/7.2Issue template
0/6.3PR template
Inputs used
has_readmeyes
has_licenseyes
readme_badges—
has_contributingyes
has_issue_templateno
has_code_of_conductyes
readme_badge_services—
has_pull_request_templateno

Sustainability & Governance

Will the project survive its people — bus factor, responsiveness, who backs it, and package upkeep?

64Moderate · 23% of overall

Maintainer resilience (bus factor)

42Weak
How it's scored
9/54Bus factor — 1 contributor(s) cover half of all commits
9.7/22.5Commit distribution — top contributor authored 57% of commits
13.5/13.5Contributor breadth — 30 contributors
10/10OpenSSF Scorecard: Contributors — project has 17 contributing companies or organizations
Inputs used
bus_factor1
contributors_sampled30
top_contributor_share0.569

Issue & PR responsiveness

61Moderate
How it's scored
14.8/42Issue resolution — 35% of issues closed
23.3/30PR acceptance — 194/250 decided PRs merged
0/13Newcomer PR acceptance — no first-time contributor's PR decided in 30d
15/15OpenSSF Scorecard: Code-Review — all changesets reviewed
Inputs used
merged_prs194
open_issues35
closed_issues19
prs_merged_7d—
prs_decided_7d—
prs_merged_30d—
prs_decided_30d—
issue_closed_ratio0.352
closed_unmerged_prs56
first_time_authors_30d—
first_time_prs_merged_30d—
first_time_prs_decided_30d—
Excluded from scoring (no data or not applicable): Newcomer PR acceptance. Remaining weights renormalized.

Ownership & stewardship

92Excellent
How it's scored
30/30Ownership backing — organization-owned
0/20Verified domain — verified-domain status not read for this organization
22.6/25Owner reach — 1,389 followers of meta-pytorch
21.1/25Track record — 58 public repos, account ~4 yr old
Inputs used
followers1,389
owner_typeOrganization
is_verified—
owner_loginmeta-pytorch
public_repos58
account_age_days1,500
Excluded from scoring (no data or not applicable): Verified domain. Remaining weights renormalized.

Engineering Quality

Are baseline engineering and documentation practices in place?

84Excellent · 19% of overall

Engineering practices

80Excellent
How it's scored
24/24CI workflows — 5 workflow(s)
24/24Tests present
16/16Linter config — .flake8
0/9.6Pre-commit hooks
0/6.4.editorconfig
0/20OpenSSF Scorecard: CI-Tests — no data
Inputs used
has_ciyes
has_testsyes
has_editorconfigno
has_linter_configyes
has_precommit_configno
Excluded from scoring (no data or not applicable): OpenSSF Scorecard: CI-Tests. Remaining weights renormalized.

Documentation

90Excellent
How it's scored
30/30README
25/25Documentation directory
15/15Documentation / homepage site — https://meta-pytorch.org/torchft/
10/10Repository description
10/10Topics — 11 topics
0/10Wiki
Inputs used
topicsfault-tolerance, ml, pytorch, torch, training, consensus, distributed-systems, fsdp, hsdp, localsgd, diloco
has_wikino
homepagehttps://meta-pytorch.org/torchft/
docs_sitehttps://meta-pytorch.org/torchft/
has_readmeyes
has_docs_diryes
has_descriptionyes

Security

Are visible security and supply-chain practices strong, without unresolved high-risk jurisdiction exposure?

55Moderate · 16% of overall

Security posture

55Moderate
How it's scored
7.5/7.5Binary-Artifacts — no binaries found in the repo
2.2/7.5Branch-Protection — branch protection is not maximal on development and all release branches
0/2.5CI-Tests — no data
0/2.5CII-Best-Practices — no effort to earn an OpenSSF best practices badge detected
7.5/7.5Code-Review — all changesets reviewed
2.5/2.5Contributors — project has 17 contributing companies or organizations
10/10Dangerous-Workflow — no dangerous workflow patterns detected
0/7.5Dependency-Update-Tool — no update tool detected
0/5Fuzzing — project is not fuzzed
2.2/2.5License — license file detected
5.2/7.5Maintained — 8 commit(s) and 1 issue activity found in the last 90 days -- score normalized to 7
0/5Packaging — no data
0/5Pinned-Dependencies — dependency not pinned by hash detected -- score normalized to 0
0/5SAST — no SAST tool detected
4.5/5Security-Policy — security policy file detected
0/7.5Signed-Releases — no data
0/7.5Token-Permissions — detected GitHub workflow tokens with excessive permissions
7.5/7.5Vulnerabilities — 0 existing vulnerabilities detected
Inputs used
sourceopenssf_scorecard
checks_evaluated15
scorecard_versionv5.5.0
checks_inconclusive3
scorecard_aggregate5.5
Excluded from scoring (no data or not applicable): CI-Tests, Packaging, Signed-Releases. Remaining weights renormalized.

AI Readiness

How well is the repo equipped to be developed and maintained with AI coding agents? Carries a deliberately small weight (4%): agent tooling is a real maintenance signal, but a repository with none can still reach 100/100.

50Moderate · 4% of overall

Agent context & guidance

1Critical
How it's scored
0/45Agent instructions — no CLAUDE.md / AGENTS.md / editor rules
0/15Machine-readable docs (llms.txt)
0/40Legible commit history — no data
Inputs used
has_llms_txtno
llms_txt_url—
legible_history_share—
agent_instruction_files—
agent_instruction_max_bytes—
Excluded from scoring (no data or not applicable): Legible commit history. Remaining weights renormalized.

Verify loop (build / test / typecheck)

74Good
How it's scored
18/18One-command bootstrap — docs/Makefile
22/22Automated tests
11/11Lint / format config — .flake8
0/11Static type checking
10/10Reproducible environment — Dockerfile
0/10Demonstrated agent practice — no data
0/8Automated maintenance — no data
0/10OpenSSF Scorecard: Pinned-Dependencies — dependency not pinned by hash detected -- score normalized to 0
Inputs used
has_nixno
has_testsyes
lockfiles—
has_dockerfileyes
typed_languageno
bootstrap_filesdocs/Makefile
has_devcontainerno
has_linter_configyes
typecheck_configs—
agent_commit_share—
toolchain_manifests—
dependency_bot_commit_share—
Excluded from scoring (no data or not applicable): Demonstrated agent practice, Automated maintenance. Remaining weights renormalized.

Code legibility for models

54Moderate
How it's scored
0/45Type-checkable code — Python without a type-check config
54.2/55Manageable file sizes — 1/73 source files over 60KB
Inputs used
primary_languagePython
largest_source_bytes71,104
source_files_sampled73
oversized_source_files1

Machine-readable interfaces

80Excellent
How it's scored
40/40API schema (OpenAPI/GraphQL/proto) — proto/torchft.proto
0/20MCP server
40/40Runnable examples — examples
Inputs used
example_dirsexamples
has_mcp_signalno
api_schema_filesproto/torchft.proto
interfaces_expected_ofnetwork-service

Key facts

521GitHub stars
30contributors
64commits, last 12 months
2days since last push
0releases
1bus factor
35open issues
crates.io, PyPIpackage ecosystems

Data collection warnings

  • Could not fetch crates package 'torchft' from its registry
  • pypi package 'torchft' points at a different repository (https://github.com/pytorch-labs/torchft); excluded from ecosystem scoring

More detail

OpenSSF Scorecard 5.5 / 10
5.5aggregate

Independent, tool-agnostic security assessment from the open-source OpenSSF Scorecard. Each check rewards a security practice, not a specific vendor's tool. Checks Scorecard could not determine are marked n/a and excluded from the security score (never counted as zero).Scorecard v5.5.0 · 2026-07-19 11:41 UTC

10Binary-Artifactsno binaries found in the repo
3Branch-Protectionbranch protection is not maximal on development and all release branches
n/aCI-Testsno pull request found
0CII-Best-Practicesno effort to earn an OpenSSF best practices badge detected
10Code-Reviewall changesets reviewed
10Contributorsproject has 17 contributing companies or organizations
10Dangerous-Workflowno dangerous workflow patterns detected
0Dependency-Update-Toolno update tool detected
0Fuzzingproject is not fuzzed
9Licenselicense file detected
7Maintained8 commit(s) and 1 issue activity found in the last 90 days -- score normalized to 7
n/aPackagingpackaging workflow not detected
0Pinned-Dependenciesdependency not pinned by hash detected -- score normalized to 0
0SASTno SAST tool detected
9Security-Policysecurity policy file detected
n/aSigned-Releasesno releases found
0Token-Permissionsdetected GitHub workflow tokens with excessive permissions
10Vulnerabilities0 existing vulnerabilities detected
Direct dependencies 22
RegistryPackageVersion constraintManifest
crates.ioanyhow1.0.89Cargo.toml
crates.ioaskama0.12.1Cargo.toml
crates.ioatty0.2.14Cargo.toml
crates.ioaxum0.7.7Cargo.toml
crates.iochrono0.4.40Cargo.toml
crates.iofern0.7.1Cargo.toml
crates.iogethostname0.5.0Cargo.toml
crates.iolog0.4.22Cargo.toml
crates.ioprost0.13.3Cargo.toml
crates.ioprost-types0.13.3Cargo.toml
crates.iopyo30.24Cargo.toml
crates.iorand0.8.5Cargo.toml
crates.ioslog2.7.0Cargo.toml
crates.ioslog-stdlog4.1.1Cargo.toml
crates.iostderrlog0.6.0Cargo.toml
crates.iostructopt0.3.26Cargo.toml
crates.iotokio1.40.0Cargo.toml
crates.iotonic0.12.2Cargo.toml
PyPItorch>=2.7pyproject.toml
PyPIopentelemetry-exporter-otlp-proto-http>=1.39.0pyproject.toml
PyPIopentelemetry-sdk>=1.39.0pyproject.toml
PyPIopentelemetry-api>=1.39.0pyproject.toml
All dependencies 34

Full resolved dependency set from the GitHub dependency graph: 22 direct and 12 indirect (transitive) packages. The transitive closure is complete when the repository commits a lockfile.

RegistryPackageVersionRelation
crates.ioanyhow—direct
crates.ioaskama—direct
crates.ioatty—direct
crates.ioaxum—direct
crates.iochrono—direct
crates.iofern—direct
crates.iogethostname—direct
crates.iolog—direct
crates.ioprost—direct
crates.ioprost-types—direct
crates.iopyo3—direct
crates.iorand—direct
crates.ioslog—direct
crates.ioslog-stdlog—direct
crates.iostderrlog—direct
crates.iostructopt—direct
crates.iotokio—direct
crates.iotonic—direct
PyPIopentelemetry-api—direct
PyPIopentelemetry-exporter-otlp-proto-http—direct
PyPIopentelemetry-sdk—direct
PyPItorch—direct
crates.iotonic-build—indirect
PyPIdocutils0.16indirect
PyPImaturin—indirect
PyPIpytest8.3.4indirect
PyPIpytorch-sphinx-theme20.2.0indirect
PyPIsphinx5.3.0indirect
PyPIsphinx-autobuild—indirect
PyPIsphinx-design0.4.0indirect
PyPIsphinx-sitemap2.7.1indirect
PyPIsphinxcontrib-katex0.8.6indirect
PyPIsphinxcontrib-mermaid1.0.0indirect
PyPIsphinxext-opengraph0.9.1indirect
Raw JSON report machine-readable

Feedback

Spotted something off in this report, or have thoughts to share? Wrong measurements, missed tooling, ideas, questions — anything is welcome. Every message is read and gets a response.

The message is kept through sign-in.

Related records

Inspected repositories sharing catalogue tags or ecosystems with meta-pytorch/torchft. Read them side by side →

93pytorch/TensorRTPyTorch/TorchScript/FX compiler for NVIDIA GPUs using TensorRTExceptional94databendlabs/openraftrust raft with improvementsExceptional100onnx/onnxOpen standard for machine learning interoperabilityExceptional99huggingface/transformers🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training. Exceptional98NVIDIA/Megatron-LMOngoing research training transformer models at scaleExceptional98aws/sagemaker-python-sdkA library for training and deploying machine learning models on Amazon SageMakerExceptional

Scores are signals, not warranties. They reflect publicly visible practices on GitHub — not a code audit, and not a security guarantee.

Missing data is excluded and weights renormalized, never scored as zero. Methodology is versioned and open: metrics v2.10.0, schema v0.15.0 — full methodology · metrics wiki.

How one result sits in the wider record: aggregate statistics — PyPI.