Labels

• bug

  Something isn't working
• component: alignment

  Tasks involving RL loss functions such as DPO and GRPO, and mathematical alignment logic
• component: distributed

  Tasks involving Ray actor management, cross-node scheduling, and communication synchronization.
• component: executors

  Tasks involving the interaction of vLLM inference and DeepSpeed ​​training endpoints.
• component: kernels

  Tasks involving the development of CUDA and Triton underlying operators
• component: observability
• component: sync-bridge

  It is a tag specifically for weight synchronization and zero-copy transfer.
• component: testing

  Add test cases and benchmark-related tasks
• do-not-close
• documentation

  Improvements or additions to documentation
• duplicate

  This issue or pull request already exists
• enhancement

  New feature or request
• feature
• good first issue

  Good for newcomers
• help wanted

  Extra attention is needed
• invalid

  This doesn't seem right
• needs-gpu-ci
• next-phase
• pinned

  Representing the core long-term mission suspended
• platform: cuda

  Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
• platform: rocm

  Specific tasks specific to AMD graphics cards (such as CK, bpreshuffle/FA)
• platform: triton

  Cross-platform Triton kernel related tasks
• priority: high

  Severe congestion issues require the highest priority for resolution.
• priority: medium
• question

  Further information is requested
• sprint-0615
• status: blocked

  paused due to lack of prerequisites, such as the upstream vLLM has not yet released a new version.
• status: needs-review

  The code is complete and awaits review by the core maintainers.
• type: ci-cd

  Modify GitHub Actions, automated tests, and packaging/deployment tasks.
• type: design

  Issues requiring in-depth discussion of architecture design