Discofork
Repository brief

ray-project/ray

Read the upstream summary on the left, browse the cached forks below it, and load each fork comparison into the right-hand panel.

Cached analysis
cached 2026-03-30T15:59:28.397Z
1mo ago

ray-project/ray

Ray is a large, active open source AI compute engine for scaling Python and ML workloads from a laptop to a cluster. It combines a distributed runtime with higher-level AI libraries for data, training, tuning, reinforcement learning, and serving, and it is very widely adopted by fork and star counts.

GitHub
Loading tags...
Stars41,877
Forks7,395
Default branchmaster
Last pushed2026-03-30T15:45:57Z
Recommended shortcuts

Jump straight into Discofork's strongest cached fork picks, or open a compare view in one click.

Forks

Choose a fork to inspect

10 of 10 fork briefs
Selected

Choose this fork if you need Ray with internal production workflow changes and are comfortable owning a significant downstream branch. Choose upstream if you want the newest community fixes and the lowest maintenance burden.

Prefer this fork only if you need its existing customizations and are willing to own long-term divergence. For most adopters, upstream Ray is the better choice because this fork is stale and likely missing many recent fixes and operational improvements.

Choose this fork if ROCm/AMD compatibility is the goal and you need downstream packaging and deployment support. Choose upstream Ray if you want the newest general-purpose Ray work and the broadest compatibility with minimal fork-specific maintenance.

Prefer this fork only if you need its LMCO-specific Ray/RLlib and packaging customizations and can accept a stale upstream base. If you want current Ray Core, Serve, Data, or platform support, upstream is the better choice.

Choose this fork if you need Tencent-specific hardware or deployment support and can absorb upstream lag. Stick with upstream Ray if you want the broadest feature coverage, fastest access to fixes, and the least merge risk.

Choose this fork only if you need its specific downstream testing, release, or Serve-related customizations and are willing to own a large rebase burden. If you want current Ray functionality and active upstream support, upstream is the better default.

Prefer upstream unless you specifically need the fork's pinned environment or custom Serve/export behavior. This fork is not a good default adoption choice for new work because it is stale, heavily diverged, and missing a large amount of later Ray functionality and maintenance.

Choose this fork only if you need its older custom training/serving workflow changes and can accept major upstream lag. For most adopters, current upstream Ray is the safer and more complete choice.

Prefer this fork only if you need its legacy customizations and are prepared to own the maintenance burden. For new adoption, upstream Ray is the safer choice because this fork is materially behind and highly divergent.

Prefer upstream unless you specifically need this fork's older behavior or one of its narrow patches. For new work, this fork is too stale and too far behind current Ray to be a safe default.