Commits · b4610c06af6ec1c93aae68aa413974c29fac9176 · OpenDAS / vllm_cscc

14 Sep, 2025 1 commit
- [fix]修复eagle 创建cu_num_tokens类型错误问题 · b4610c06
  王敏 authored Sep 14, 2025
  
  b4610c06
13 Sep, 2025 3 commits
- add VLLM_USE_MERGE_ATTN_STATES_OPT to control merge_attn_states support · 2071c380
  zhuwenwen authored Sep 13, 2025
  
  2071c380
- set VLLM_HAS_CONTEXT_DEFAULT=1 · 48b4c41d
  zhuwenwen authored Sep 13, 2025
  
  48b4c41d
- update triton cat scheduling · b1329ff2
  zhuwenwen authored Sep 13, 2025
```
update the default values of VLLM_USE_TRITON_CAT and VLLM_USE_LIGHT_OP to True
```
  b1329ff2
12 Sep, 2025 1 commit
- update the cat implementation of triton's non contiguous memory for the decode phase · 2c169409
  zhuwenwen authored Sep 12, 2025
  
  2c169409
11 Sep, 2025 5 commits
- 新增dpsk-v3.1-awq的支持 · e5f51b79
  yangql authored Sep 11, 2025
  
  e5f51b79
- Merge branch 'v0.9.2-dev' of http://10.16.6.30/dcutoolkit/deeplearing/vllm into v0.9.2-dev · 5c288d91
  zhuwenwen authored Sep 11, 2025
  
  5c288d91
- update fa interface · bf3d75f4
  zhuwenwen authored Sep 11, 2025
  
  bf3d75f4
- Merge branch 'v0.9.2-dev-w4a8' into 'v0.9.2-dev' · e128ad19
  zhuwenwen authored Sep 11, 2025
```
fix: w4a8 marlin 中 weight重排接入lightop算子

See merge request dcutoolkit/deeplearing/vllm!202
```
  e128ad19
- fix: w4a8 marlin 中 weight重排接入lightop算子 · 082d41a1
  jujl1 authored Sep 10, 2025
  
  082d41a1
10 Sep, 2025 8 commits
- add rocm merge_attn_states · f6324f60
  zhuwenwen authored Sep 10, 2025
  
  f6324f60
- update triton kernel to optimize torch cat for ds prefill · 90e40f49
  zhuwenwen authored Sep 10, 2025
  
  90e40f49
- support cascade_attention · ead74dfa
  zhuwenwen authored Sep 10, 2025
  
  ead74dfa
- update the conditions for pad_v on v0 · d6dc122f
  zhuwenwen authored Sep 10, 2025
  
  d6dc122f
- update the conditions for pad_v · aaa89e82
  zhuwenwen authored Sep 10, 2025
  
  aaa89e82
- update flash-attn interface of apply_rotary_emb · 37707203
  zhuwenwen authored Sep 10, 2025
  
  37707203
- update VLLM_USE_TRITON_CAT during the prefill phase · 7d959770
  zhuwenwen authored Sep 10, 2025
  
  7d959770
- use VLLM_USE_TRITON_CAT during the prefill phase · 072d4638
  zhuwenwen authored Sep 10, 2025
  
  072d4638
09 Sep, 2025 7 commits
- update flashmla.py · ff090f36
  zhuwenwen authored Sep 09, 2025
  
  ff090f36
- import concatv3Tritonfinal · 5973c805
  zhuwenwen authored Sep 09, 2025
  
  5973c805
- add VLLM_USE_TRITON_CAT to opt torch cat · 379315a0
  zhuwenwen authored Sep 09, 2025
  
  379315a0
- update qwen2_vl and qwen2_5_vl conv layout · 71cb74ff
  zhuwenwen authored Sep 09, 2025
  
  71cb74ff
- update flash-attn interface to support keye · 4d53d14c
  zhuwenwen authored Sep 09, 2025
  
  4d53d14c
- Merge branch 'v0.9.2-dev-marlin' into 'v0.9.2-dev' · a5f106eb
  zhuwenwen authored Sep 09, 2025
```
fix: 优化w4a8 marlin 中 weight重排耗时

See merge request dcutoolkit/deeplearing/vllm!200
```
  a5f106eb
- fix: 优化w4a8 marlin 中 weight重排耗时 · ffab74dd
  jujl1 authored Sep 09, 2025
  
  ffab74dd
08 Sep, 2025 1 commit
- update op.moe_fused_gate · bc9aee38
  zhuwenwen authored Sep 08, 2025
  
  bc9aee38
07 Sep, 2025 1 commit
- add VLLM_USE_LIGHT_OP to optimize moe_align_block_size and moe_fused_gate · a54ab95d
  zhuwenwen authored Sep 07, 2025
  
  a54ab95d
06 Sep, 2025 9 commits
- Merge branch 'v0.9.2-dev-fth' into 'v0.9.2-dev' · ca034fd5
  zhuwenwen authored Sep 06, 2025
```
V0.9.2 dev fth

See merge request dcutoolkit/deeplearing/vllm!199
```
  ca034fd5
- Update version.py · 59a345ac
  gaoqiong authored Sep 06, 2025
  
  59a345ac
- 回退修改的version.py · 4e471b8a
  gaoqiong authored Sep 06, 2025
  
  4e471b8a
- [P/D][Feature]显存调度优化，及时释放tensor · ed392378
  zhuwenwen authored Sep 06, 2025
  
  ed392378
- Merge branch 'v0.9.2-dev' of http://10.16.6.30/dcutoolkit/deeplearing/vllm into v0.9.2-dev · 596c18f6
  zhuwenwen authored Sep 06, 2025
  
  596c18f6
- [fix]fix tests of neuron, quantization etc · dc2aff4c
  zhuwenwen authored Sep 06, 2025
  
  dc2aff4c
- Merge branch 'v0.9.2-dev-lzg' into 'v0.9.2-dev' · 5e1ca86d
  zhuwenwen authored Sep 06, 2025
```
fix precision issue in mtp

See merge request dcutoolkit/deeplearing/vllm!198
```
  5e1ca86d
- fix precision issue in mtp · 2a4a2877
  lizhigong authored Sep 05, 2025
  
  2a4a2877
- 增加reduce修改 · 863176e5
  SAC_fanth authored Sep 06, 2025
  
  863176e5
05 Sep, 2025 3 commits
- Merge branch 'v0.9.2-dev' of http://10.16.6.30/dcutoolkit/deeplearing/vllm into v0.9.2-dev · a5d54d38
  zhuwenwen authored Sep 05, 2025
  
  a5d54d38
- [fix]fix tests of core, samplers and tokenization etc. · d3473ba4
  zhuwenwen authored Sep 05, 2025
  
  d3473ba4
- Merge branch 'v0.9.2-dev-lzg' into 'v0.9.2-dev' · 9a7d05c2
  zhuwenwen authored Sep 05, 2025
```
fix bugs in zero overhead and tbo

See merge request dcutoolkit/deeplearing/vllm!197
```
  9a7d05c2
04 Sep, 2025 1 commit
- fix performance issues caused by enabling TBO · 424cccfe
  lizhigong authored Sep 04, 2025
  
  424cccfe