Skip to content

v0.0.19

Choose a tag to compare

@xiaguan xiaguan released this 13 Apr 08:07
· 108 commits to master since this release
5f716db

What's Changed

  • feat(core): add flush_saves and expand integration tests by @xiaguan in #194
  • [codex] refresh AGENTS.md guidance by @xiaguan in #200
  • feat(core): add CUDA 12/13 feature flags with cuMemcpyBatchAsync_v2 by @xiaguan in #201
  • fix(connector): align decode save and load block indexes by @xiaguan in #204
  • fix(connector): fix MLA skip-save lifecycle and remove blocking wait_for_save by @xiaguan in #205
  • fix(connector): isolate cross-layer KV cache across PP ranks by @xiaguan in #206
  • chore(release): bump pegaflow version to 0.19.0 by @xiaguan in #207

Full Changelog: v0.0.18...v0.0.19