Repository navigation

[Perf][Model] Enable CoVo audio encoder CUDA graphs - #8572

Open

jeffaa729 wants to merge 11 commits into

vllm-project:mainfrom

jeffaa729:perf/covo-audio-encoder-cudagraph

Open

[Perf][Model] Enable CoVo audio encoder CUDA graphs#8572
jeffaa729 wants to merge 11 commits into
vllm-project:mainfrom
jeffaa729:perf/covo-audio-encoder-cudagraph

Commits on Oct 6, 2026
- [Perf][Model] Add CoVo audio encoder CUDA graph support
  jeffaa729
  committed
  1a1772e
Commits on Oct 7, 2026
- [Test][Model] Cover CoVo encoder CUDA graph execution
  jeffaa729
  committed
  52e5666
Commits on Oct 13, 2026
- Share CoVo encoder computation and token constant
  jeffaa729
  committed
  8ac6b09
Commits on Oct 17, 2026
- Merge branch 'main' into perf/covo-audio-encoder-cudagraph
  jeffaa729
  authored
  d09b06d
- Covo eagar fallback and test runner dispatch
  jeffaa729
  committed
  3352cae
Commits on Oct 30, 2026
- Address CoVo encoder CUDA graph constants
  jeffaa729
  committed
  88d2ac0
- [Bugfix] Validate CoVo audio lengths in eager and graph paths
  jeffaa729
  committed
  b8ca825
- validate CoVo audio lengths before encoder batching
  jeffaa729
  committed
  4b44b0d
- check CoVo audio duration validation and test boundaries, better error log
  jeffaa729
  committed
  44006b8
- add extra 1 sample,use AUDIO_SAMPLE_RATE
  jeffaa729
  committed
  5cdf6c9
- Guard CoVo processor sample rate
  jeffaa729
  committed
  c7c08a7