Commits · 34bf6014a574e468b6c73142b37d0ebacdc9ec76 · OpenDAS / vllm_cscc

27 Oct, 2025 3 commits
- Merge branch 'v0.9.2-dev' into v0.9.2-step3v · 34bf6014
  zhuwenwen authored Oct 27, 2025
  
  34bf6014
- set VLLM_USE_LIGHTOP_MOE_SUM_MUL_ADD=0 · cd738b68
  zhuwenwen authored Oct 27, 2025
  
  cd738b68
- Merge branch 'v0.9.2-dev' into v0.9.2-step3v · 9b35bbfa
  zhuwenwen authored Oct 27, 2025
  
  9b35bbfa
24 Oct, 2025 5 commits
- update apply_rotary_emb for z100l&k100 (keye) · 4ba4b755
  zhuwenwen authored Oct 24, 2025
  
  4ba4b755
- use apply_rotary_emb_torch on z100l&k100 (qwen2-vl) · 62fe9a48
  zhuwenwen authored Oct 24, 2025
  
  62fe9a48
- use _apply_rotary_emb_torch for z100l&k100 · 48d8707e
  zhuwenwen authored Oct 24, 2025
  
  48d8707e
- Merge branch 'v0.9.2-dev' into v0.9.2-step3v · 7f54652e
  zhuwenwen authored Oct 24, 2025
  
  7f54652e
- set USE_FUSED_RMS_QUANT=0 and USE_FUSED_SILU_MUL_QUANT=0 · 46dd30e7
  zhuwenwen authored Oct 24, 2025
```
add VLLM_USE_LIGHTOP_MOE_SUM_MUL_ADD
```
  46dd30e7
23 Oct, 2025 3 commits
- Merge branch 'v0.9.2-dev' into v0.9.2-step3v · 10a836fb
  zhuwenwen authored Oct 23, 2025
  
  10a836fb
- update test_moe.py · 86d92eb9
  zhuwenwen authored Oct 23, 2025
```
set USE_FUSED_RMS_QUANT=1 and USE_FUSED_SILU_MUL_QUANT=1
```
  86d92eb9
- support prefix cache on kme · 6f1db287
  zhuwenwen authored Oct 23, 2025
```
fix the error in test_moe caused by moe align not supporting 511
multi-modal switching to torch implementation on z100l&k100
```
  6f1db287
21 Oct, 2025 2 commits
- Merge branch 'v0.9.2-dev' into v0.9.2-step3v · d9934804
  zhuwenwen authored Oct 21, 2025
  
  d9934804
- update linear of RowParallelLinear and UnquantizedLinearMethod apply · 56ebbba3
  zhuwenwen authored Oct 21, 2025
  
  56ebbba3
20 Oct, 2025 4 commits
- 修复pd分离开启异步发送时decode侧判断条件卡住 · 384b6bd9
  zhuwenwen authored Oct 20, 2025
  
  384b6bd9
- Merge branch 'v0.9.2-dev' into v0.9.2-step3v · ee346e93
  zhuwenwen authored Oct 20, 2025
  
  ee346e93
- [models] support step3v · 583034f1
  zhuwenwen authored Oct 20, 2025
  
  583034f1
- update lightop == 0.6.0 · 72ada8dc
  zhuwenwen authored Oct 20, 2025
  
  72ada8dc
17 Oct, 2025 2 commits
- 支持pd分离p2p_async & 解决oom问题 · 4824a3fa
  zhuwenwen authored Oct 17, 2025
  
  4824a3fa
- Set the default value of routed_scaling_factor to 1 · 9a505ffb
  zhuwenwen authored Oct 17, 2025
  
  9a505ffb
16 Oct, 2025 3 commits
- fix v0 eager fa-pa acc error · 0adf9cda
  zhuwenwen authored Oct 16, 2025
  
  0adf9cda
- update triton3.1 · aed7c9f8
  zhuwenwen authored Oct 16, 2025
  
  aed7c9f8
- Merge branch 'v0.9.2-dev-yql' into 'v0.9.2-dev' · 7dceb45f
  zhuwenwen authored Oct 16, 2025
```
修复awq 的mtp中的blockint8的问题

See merge request dcutoolkit/deeplearing/vllm!229
```
  7dceb45f
15 Oct, 2025 6 commits
- 修复awq 的mtp中的blockint8的问题 · 0cc9bc74
  yangql authored Oct 15, 2025
  
  0cc9bc74
- Merge branch 'v0.9.2-dev-yql' into 'v0.9.2-dev' · 1b98d0bb
  zhuwenwen authored Oct 15, 2025
```
删除DPSK_FP16_QUICK，以及增加awq和blockwiseint8的shared_output接口

See merge request dcutoolkit/deeplearing/vllm!228
```
  1b98d0bb
- 删除DPSK_FP16_QUICK，以及增加awq和blockwiseint8的shared_output接口 · 50cb9270
  yangql authored Oct 15, 2025
  
  50cb9270
- release v0.9.2 of das1.7 · 0004c57a
  zhuwenwen authored Oct 15, 2025
  
  0004c57a
- set VLLM_USE_OPT_MOE_SUM=1 and VLLM_USE_LIGHTOP_MOE_SUM=1 · 15ef12c1
  zhuwenwen authored Oct 15, 2025
  
  15ef12c1
- support --no-enable-chunked-prefill of v1 · 2c4b2c80
  zhuwenwen authored Oct 15, 2025
  
  2c4b2c80
13 Oct, 2025 11 commits
- Merge branch 'v0.9.2-dev-wm-1' into 'v0.9.2-dev' · f7e9c329
  zhuwenwen authored Oct 13, 2025
```
去掉all2all ep相关代码

See merge request dcutoolkit/deeplearing/vllm!226
```
  f7e9c329
- 去掉all2all ep相关代码 · 0b467604
  王敏 authored Oct 13, 2025
  
  0b467604
- update the envs for moe_sum and moe_align · 766663e6
  zhuwenwen authored Oct 13, 2025
  
  766663e6
- update moe_sum and moe_align · 1277ff09
  zhuwenwen authored Oct 13, 2025
  
  1277ff09
- optimize the implementation of moe_sum (lightop) · 7e68a7fe
  zhuwenwen authored Oct 13, 2025
  
  7e68a7fe
- remove --no-enable-chunked-prefill of v1 · bdaaf39d
  zhuwenwen authored Oct 13, 2025
  
  bdaaf39d
- optimize the implementation of moe_sum · 06a1bee2
  zhuwenwen authored Oct 13, 2025
  
  06a1bee2
- disable cascade_attn · b7989b07
  zhuwenwen authored Oct 13, 2025
  
  b7989b07
- support dsv32 · 633f8199
  zhuwenwen authored Oct 13, 2025
  
  633f8199
- 新增fp8—e5m2 · ed3cdc81
  zhuwenwen authored Oct 13, 2025
  
  ed3cdc81
- Merge branch 'v0.9.2-dev' of ssh://10.16.6.30:10022/dcutoolkit/deeplearing/vllm into v0.9.2-dev · cf13152f
  zhuwenwen authored Oct 13, 2025
  
  cf13152f
11 Oct, 2025 1 commit
- Merge branch 'v0.9.2-dev-maxiao2' into 'v0.9.2-dev' · 99863602
  zhuwenwen authored Oct 11, 2025
```
fix pd send async perfomance

See merge request dcutoolkit/deeplearing/vllm!224
```
  99863602