whx 5998704c08 [BugFix] Fix ascend scheduler bugs. (#822) ...

This PR fixes two bugs in AscendScheduler:
1. When running with high concurrency, the length of running queue may
exceed the limit of max_num_seqs
2. When some requests are prempted and recomputing is activated, the
logic of computing new tokens is wrong.

Signed-off-by: whx-sjtu <2952154980@qq.com>

2025-05-12 21:15:17 +08:00

..

attention

[Enhancement] Add padding for ACL Graph (#803)

2025-05-12 20:26:22 +08:00

core

[BugFix] Fix ascend scheduler bugs. (#822)

2025-05-12 21:15:17 +08:00

device_allocator

[MISC] Clean up torch_npu (#688)

2025-04-29 18:03:38 +08:00

distributed

[Disaggregated Prefill] P2P Disaggregated Prefill based on llm_datadist (#694)

2025-05-01 22:31:36 +08:00

lora

[Bugfix] fix import error (#600)

2025-04-22 08:57:25 +08:00

models

feat: support torchair graph mode in v1 engine (#789)

2025-05-12 19:14:07 +08:00

ops

[Bugfix] Correct method call for _set_cos_sin_cache (#774)

2025-05-09 12:55:57 +08:00

patch

[Bugfix] Tweak distributed process group initialization and add dummy… (#816)

2025-05-12 17:31:29 +08:00

quantization

[Perf] Optimize fused_experts quantization code to save npu memory (#784)

2025-05-09 15:09:37 +08:00

worker

[Enhancement] Add padding for ACL Graph (#803)

2025-05-12 20:26:22 +08:00

__init__.py

[Core] Cleanup triton patch which has been fixed in vllm (#764)

2025-05-06 18:52:15 +08:00

envs.py

[MISC] Make vllm version configurable (#651)

2025-04-28 14:19:06 +08:00

platform.py

[Enhancement] Add padding for ACL Graph (#803)

2025-05-12 20:26:22 +08:00

utils.py

[Enhancement] Add padding for ACL Graph (#803)

2025-05-12 20:26:22 +08:00