Labels

related to API changes

related to the auto-parallel feature

Something isn't working

ChatGPT Application

related to compatibility

stable diffusion related

Improvements or additions to documentation

This issue or pull request already exists

New feature or request

example-related issuer or pull request

related to the gemini feature

Good for newcomers

Extra attention is needed

(pilot) Someone is working on it or a PR is open; exempt from stale

This doesn't seem right

known issue to be fixed

Lazy initialization

(pilot) MoE models and expert parallelism: expert sharding, routing, Mixtral/DeepSeek/Qwen MoE

(pilot) Sequence parallelism: all_to_all (Ulysses), ring attention, split_gather

(pilot) ZeRO (LowLevelZero) and Gemini: sharding, offload, chunk management

related to the pipeline-parallel feature

(pilot) Affects mainstream usage or numerical correctness; handle first

Further information is requested

a release PR for a version bump

related to sequence parallel