Commits · b7b4e6393ed2a4d5a1e88c69c96aa7a86d13dccc · OpenDAS / vllm_cscc

"tests/compile/passes/__init__.py" did not exist on "2c8b9182b5ced00d83bed15ef8bc0ac6e079b6ee"

05 Oct, 2025 1 commit
- fix rope acc error · b7b4e639
  zhuwenwen authored Oct 05, 2025
  
  b7b4e639
29 Sep, 2025 1 commit
- update moe_sum interface · 00bbf0bb
  zhuwenwen authored Sep 29, 2025
  
  00bbf0bb
26 Sep, 2025 4 commits
- off rotary_embedding_deepseek_fuse · dc675592
  zhuwenwen authored Sep 26, 2025
  
  dc675592
- feat: moe_align_block_size 更新lightop 接口,加入对ep的支持 · 0dcc2e60
  jujl1 authored Sep 26, 2025
  
  0dcc2e60
- update op · 92c6171e
  zhuwenwen authored Sep 26, 2025
  
  92c6171e
- VLLM_USE_LIGHTOP and VLLM_USE_OPT_CAT · c1ece8c6
  zhuwenwen authored Sep 26, 2025
```
add shared_output and routed_scaling_factor of CompressedTensorsW8A8Int8MoEMethod
```
  c1ece8c6
25 Sep, 2025 3 commits
- add shared_output and routed_scaling_factor of w4a8 · b33ff2d6
  zhuwenwen authored Sep 25, 2025
  
  b33ff2d6
- DeepSeek-R1-Channel-INT8调用rmsquant融合。 · a5043e83
  wujl5 authored Sep 25, 2025
  
  a5043e83
- [kernels] add fused_rms_norm_contiguous and rotary_embedding_deepseek_fuse · 49810c37
  zhuwenwen authored Sep 25, 2025
```
[kernels] update moe_align_block_size and moe_sum interface
```
  49810c37
24 Sep, 2025 2 commits
- update VLLM_USE_OPT_CAT · 4d97c5fc
  zhuwenwen authored Sep 24, 2025
  
  4d97c5fc
- [kernel] add lightop's moe_sum(mul+add) fusion operator for deepseek · 8d2cac26
  zhuwenwen authored Sep 24, 2025
```
[FIX] 修复mtp和VLLM_USE_TRITON_CAT不能一起开的bug
```
  8d2cac26
23 Sep, 2025 1 commit
- 修复W4A8 PP8崩溃问题 · 4020670f
  zhuwenwen authored Sep 23, 2025
  
  4020670f
22 Sep, 2025 1 commit
- deepseek-r1-w4a8 mlp/moe调用silu-mul-quant融合 · 3964e019
  wujl5 authored Sep 22, 2025
  
  3964e019
15 Sep, 2025 1 commit
- update moe_align_block_size · d251d4f3
  zhuwenwen authored Sep 15, 2025
  
  d251d4f3
14 Sep, 2025 1 commit
- deepseek-r1-w4a8使用rmsquant融合算子及横向融合 · fc443d52
  wujl5 authored Sep 14, 2025
  
  fc443d52
11 Sep, 2025 2 commits
- 新增dpsk-v3.1-awq的支持 · e5f51b79
  yangql authored Sep 11, 2025
  
  e5f51b79
- fix: w4a8 marlin 中 weight重排接入lightop算子 · 082d41a1
  jujl1 authored Sep 10, 2025
  
  082d41a1
10 Sep, 2025 1 commit
- update flash-attn interface of apply_rotary_emb · 37707203
  zhuwenwen authored Sep 10, 2025
  
  37707203
09 Sep, 2025 3 commits
- 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
- 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
04 Sep, 2025 1 commit
- add glm-4.5-air-tp8 nn moe configs · 7a97637e
  zhuwenwen authored Sep 04, 2025
  
  7a97637e
03 Sep, 2025 1 commit
- [fix]fix tests of v1 and worker · a7668e46
  zhuwenwen authored Sep 03, 2025
  
  a7668e46
02 Sep, 2025 1 commit
- remove topk_ids · 4a62a3eb
  zhuwenwen authored Sep 02, 2025
  
  4a62a3eb
01 Sep, 2025 2 commits
- [fix]fix tests of fused_moe · 58bbb720
  zhuwenwen authored Sep 01, 2025
  
  58bbb720
- [BugFix]支持v1 engine pp deepseek · 4a946680
  zhuwenwen authored Sep 01, 2025
  
  4a946680
29 Aug, 2025 1 commit
- 修复residual FP16 overflow，解决mtp采样率和数据集精度的冲突 · 080ed180
  yangql authored Aug 29, 2025
  
  080ed180
27 Aug, 2025 1 commit
- feat: w4a8 marlin合入主分支, 通过-q slimquant_w4a8_marlin 可以启用 · 533d4c72
  jujl1 authored Aug 27, 2025
  
  533d4c72
26 Aug, 2025 2 commits
- feat: w4a8 marlin 初版合入 · d0181e5a
  jujl1 authored Aug 25, 2025
  
  d0181e5a
- [fix]fix tests of kernels · 31584b45
  zhuwenwen authored Aug 26, 2025
  
  31584b45
20 Aug, 2025 1 commit
- [fix]v0 SamplerOutput在非tree decoding时不传入logits · 01692be6
  王敏 authored Aug 20, 2025
  
  01692be6
15 Aug, 2025 2 commits
- [fix]修复mtp eager模式下显存占用增加问题 · 49559d79
  王敏 authored Aug 15, 2025
  
  49559d79
- 增加marlin对cache13的支持，以及新增flash mla的kvcache fp8的支持 · 8f25283a
  yangql authored Aug 15, 2025
  
  8f25283a
11 Aug, 2025 1 commit
- add qwen3-coder-480b moe configs · 60e5d686
  zhuwenwen authored Aug 11, 2025
  
  60e5d686
08 Aug, 2025 1 commit
- feat:新增VLLM_USE_GLOBAL_CACHE13 设置moe使用全局变量的cache13 · 333104ab
  jujl1 authored Jul 31, 2025
  
  333104ab
07 Aug, 2025 3 commits
- 修改增加SlimQuantW4A8Int8MoEMethod 获取intermediate_size_per_partition 支持 · e92bb9ea
  gaoqiong authored Aug 07, 2025
  
  e92bb9ea
- 修改增加lmslimquant_w4a8量化支持 · 8b1e4ef0
  gaoqiong authored Aug 07, 2025
  
  8b1e4ef0
- [feat]支持mtp模型full_cuda_graph · 89eecc55
  王敏 authored Aug 07, 2025
  
  89eecc55