Junhong Liu 2ee4a07adb [Refactor][Perf] Qwen3-omni: code predictor with re-prefill + SDPA and eliminate decode hot-path CPU round-trips (#1758) ...

Signed-off-by: Junhong Liu <98734602+LJH-LBJ@users.noreply.github.com>

2026-03-11 07:52:21 +08:00

..

__init__.py

Initial vLLM-omni repository structure

2025-09-11 14:32:51 +08:00

base.py

[Debug] Add trigger to concurrent stage init (#1274)

2026-02-11 14:18:18 +08:00

gpu_ar_model_runner.py

[Bagel]: Support multistage img2img (#1669)

2026-03-09 22:23:50 +08:00

gpu_ar_worker.py

[Rebase] Rebase to vllm v0.16.0 (#1357)

2026-02-13 18:05:07 +08:00

gpu_generation_model_runner.py

Dev/rebase v0170 (#1639)

2026-03-09 17:29:38 +08:00

gpu_generation_worker.py

[Rebase] Rebase to vllm v0.16.0 (#1357)

2026-02-13 18:05:07 +08:00

gpu_memory_utils.py

Process-Scoped GPU Memory Accounting (#1204)

2026-02-08 22:43:02 +08:00

gpu_model_runner.py

[Refactor][Perf] Qwen3-omni: code predictor with re-prefill + SDPA and eliminate decode hot-path CPU round-trips (#1758)

2026-03-11 07:52:21 +08:00

mixins.py

[Hardware] Support platforms and plugin system (#774)

2026-01-28 22:12:52 +08:00