Commits · 026894204f84f182fc85f7ca3a14b46a5045d023 · OpenDAS / vllm_cscc

05 Feb, 2026 1 commit

feat: Support shared experts fusion. · bbe4df8b

wanglong3 authored Jan 29, 2026

feat: support moe sum when topk==9

bugfix: Fix mtp model load error when eable shared experts fusion.

bbe4df8b

23 Jan, 2026 2 commits
- fix interface error · fc55a25c
  zhuwenwen authored Jan 23, 2026
  
  fc55a25c
- support fa kvcache fp8, add VLLM_USE_QUERY_QUANT to not use q quant(todo) · b3062dab
  zhuwenwen authored Jan 23, 2026
  
  b3062dab
15 Jan, 2026 1 commit
- pcie 解决custom cudagraph模式需要拷贝的问题，这个需要配合dtk进行使用。 · d65c5085
  xiabo authored Jan 15, 2026
```
vllm：export VLLM_CUSTOM_CACHE=1
dtk：export HIP_KERNEL_EVENT_SYSTENFENCE=1
```
  d65c5085
18 Dec, 2025 1 commit
- 解决custom cudagraph模式需要拷贝的问题，这个需要配合dtk进行使用。 · 651925e8
  xiabo authored Dec 18, 2025
```
vllm：export VLLM_CUSTOM_CACHE=1
dtk：export HIP_KERNEL_EVENT_SYSTENFENCE=1
```
  651925e8
18 Nov, 2025 1 commit
- deepseekv2-w4a8支持custom-rms-quant融合 · 3e6729e0
  wujl5 authored Nov 18, 2025
  
  3e6729e0
13 Oct, 2025 1 commit
- 新增fp8—e5m2 · ed3cdc81
  zhuwenwen authored Oct 13, 2025
  
  ed3cdc81
29 Sep, 2025 2 commits
- Revert "新增fp8—e5m2" · 3903e15e
  zhuwenwen authored Sep 29, 2025
```
This reverts commit d6ba8ed5.
```
  3903e15e
- 新增fp8—e5m2 · d6ba8ed5
  zhuwenwen authored Sep 29, 2025
  
  d6ba8ed5
28 Sep, 2025 3 commits
- 修复kvcache-fp8—e5m2的不能开cp的bug · b2d14ba3
  yangql authored Sep 28, 2025
  
  b2d14ba3
- 新增fp8—e5m2_utils · bb13d854
  yangql authored Sep 28, 2025
  
  bb13d854
- 新增fp8—e5m2 · e29550e2
  yangql authored Sep 28, 2025
  
  e29550e2
23 Sep, 2025 1 commit
- add convert_vertical_slash_indexes and convert_vertical_slash_indexes_mergehead · 5086453d
  zhuwenwen authored Sep 23, 2025
  
  5086453d
10 Sep, 2025 1 commit
- add rocm merge_attn_states · f6324f60
  zhuwenwen authored Sep 10, 2025
  
  f6324f60
06 Sep, 2025 1 commit
- 增加reduce修改 · 863176e5
  SAC_fanth authored Sep 06, 2025
  
  863176e5
11 Aug, 2025 1 commit
- update act_and_mul opt · e04730ac
  zhuwenwen authored Aug 11, 2025
  
  e04730ac
18 Jul, 2025 1 commit
- remove unused sgl_moe_align_block_size · fd46df33
  zhuwenwen authored Jul 18, 2025
  
  fd46df33
16 Jul, 2025 1 commit
- support rms_norm_dynamic_per_token_quant · 91feb245
  zhuwenwen authored Jul 16, 2025
  
  91feb245
12 Jul, 2025 1 commit
- fix rmsnrom precision · ac61d69b
  wanghl6 authored Jul 12, 2025
  
  ac61d69b
06 Jul, 2025 2 commits
- [BugFix] Fix: ImportError when building on hopper systems (#20513) · 40b86aa0
  Lucas Wilkinson authored Jul 06, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
  40b86aa0
- [BUG] Fix #20484. Support empty sequence in cuda penalty kernel (#20491) · f73d02aa
  Vadim Gimpelson authored Jul 06, 2025
```
Signed-off-by: Vadim Gimpelson <vadim.gimpelson@centml.ai>
```
  f73d02aa
04 Jul, 2025 2 commits
- [feat]: CUTLASS block scaled group gemm for SM100 (#19757) · 3d184b95
  Duncan Moss authored Jul 04, 2025
```
Signed-off-by: Duncan Moss <djm.moss@gmail.com>
Co-authored-by: Duncan Moss <dmoss@nvidia.com>
```
  3d184b95
- [Perf] Optimize Vectorization Utils for Int 8 Quantization Kernels (#20331) · 783921d8
  Wentao Ye authored Jul 04, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
```
  783921d8
02 Jul, 2025 3 commits
- [NVIDIA] Support Cutlass w8a8 FP8 for Blackwell Geforce GPUs (sm120) (#17280) · 9e5552aa
  Joonchen Liau authored Jul 02, 2025
```
Signed-off-by: kaln27 <liaojuncheng123@foxmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  9e5552aa
- [Kernel][Bugfix] Fixup some warnings in nvfp4_blockwise_moe when CUDA < 12.8 (#20324) · 3be8d312
  Tyler Michael Smith authored Jul 01, 2025
```
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
```
  3be8d312
- remove unused variables in marlin_template.h (#20236) · 9290de56
  周周周 authored Jul 02, 2025
  
  9290de56
01 Jul, 2025 3 commits
- [CPU] Update custom ops for the CPU backend (#20255) · 6cc1e7d9
  Li, Jiang authored Jul 01, 2025
```
Signed-off-by: jiang1.li <jiang1.li@intel.com>
```
  6cc1e7d9
- Fix `numel()` downcast in vllm/csrc/moe/moe_align_sum_kernels.cu +2 (#17082) · 86debab5
  Richard Barnes authored Jul 01, 2025
```
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  86debab5
- 解决PA崩溃 · 5d3fb1d4
  zhuwenwen authored Jul 01, 2025
  
  5d3fb1d4
30 Jun, 2025 1 commit
- 解决PA 大batch，32k，崩溃问题 · 9d5187eb
  zhuwenwen authored Jun 30, 2025
  
  9d5187eb
27 Jun, 2025 6 commits
- [Bugfix] Fix some narrowing conversion warnings (#20141) · e8c3bd2c
  Tyler Michael Smith authored Jun 27, 2025
```
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
```
  e8c3bd2c
- [Fix][ROCm] Remove unused variables to fix build error on GFX11/12 (#19891) · 94a55c76
  Hosang authored Jun 27, 2025
```
Signed-off-by: Hosang Yoon <hosang.yoon@amd.com>
```
  94a55c76
- update cutlass fa and pa · 4ddb5447
  zhuwenwen authored Jun 27, 2025
  
  4ddb5447
- [Feature] add quick all reduce (#19744) · 0740e29b
  li haoyang authored Jun 27, 2025
```
Signed-off-by: ilmarkov <imarkov@redhat.com>
Signed-off-by: Haoyang Li <Haoyang.Li@amd.com>
Co-authored-by: ilmarkov <imarkov@redhat.com>
```
  0740e29b
- [Bugfix] Build moe_data for both sm100 and sm90 (#20086) · 44d2e6af
  Michael Goin authored Jun 27, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  44d2e6af
- [Perf] SM100 FP8 GEMM Optimizations after cutlass_profiler (#20071) · 2d7779f8
  Ilya Markov authored Jun 27, 2025
```
Signed-off-by: ilmarkov <imarkov@redhat.com>
Co-authored-by: ilmarkov <imarkov@redhat.com>
```
  2d7779f8
26 Jun, 2025 1 commit
- [CPU] Fix torch version in x86 CPU backend (#19258) · 0567c824
  Li, Jiang authored Jun 26, 2025
```
Signed-off-by: jiang1.li <jiang1.li@intel.com>
```
  0567c824
21 Jun, 2025 2 commits
- skip silu_and_mul_quant · d6a856f5
  zhuwenwen authored Jun 21, 2025
  
  d6a856f5
- fix build error · e87de76c
  zhuwenwen authored Jun 21, 2025
  
  e87de76c
17 Jun, 2025 1 commit
- [Perf] Optimize `moe_align_block_size` CUDA kernel (#19572) · ffb2cd6b
  Wentao Ye authored Jun 17, 2025
```
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  ffb2cd6b