Repository navigation

[model, feature] qwen3-omni: add packed sequence support and shared sequence utilities - #4304

Open

hbhflw2000 wants to merge 11 commits into

NVIDIA-NeMo:mainfrom

hbhflw2000:pr4_omni3_packseq_sequence_utils

Open

[model, feature] qwen3-omni: add packed sequence support and shared sequence utilities#4304
hbhflw2000 wants to merge 11 commits into
NVIDIA-NeMo:mainfrom
hbhflw2000:pr4_omni3_packseq_sequence_utils

Commits on Sep 2, 2026
- [model] Add Qwen3-Omni packed sequence utilities
  hbhflw2000
  committed
  be82426
- [model] Cover Qwen3-Omni packseq padding helper
  hbhflw2000
  committed
  c9a3e33
- [model] Address Qwen3-Omni packseq review comments
  hbhflw2000
  committed
  75d0d8b
- [test] Cover Qwen3-Omni packed CP forward step
  hbhflw2000
  committed
  af982f2
- [model] Use MCore process group size utility
  hbhflw2000
  committed
  79ca807
- [test] Use Qwen3-Omni in-batch packing flag
  hbhflw2000
  committed
  1e9b750
- [style] Fix packed sequence utility import spacing
  hbhflw2000
  committed
  c795383
- [test] Update Qwen3-Omni CP mock for MCore helper
  hbhflw2000
  committed
  52b5a84
- [model] Align Qwen3-Omni packing with collator metadata
  hbhflw2000
  committed
  5f8636a
- [test] Adapt Qwen3-Omni audio fake to shared length helper
  hbhflw2000
  committed
  6030b5f
Commits on Sep 23, 2026
- Merge branch 'main' into pr4_omni3_packseq_sequence_utils
  hbhflw2000
  authored
  64b6731