Commits · aa9f3dc38572c6e3f25ea5d3c3373c78ec7e9c8b · OpenDAS / vllm_cscc · GitLab

16 May, 2025 2 commits
- [Bugfix] fix `an illegal memory access was encountered` of marlin kernel + act_order (#18245) · e73b7dfd
  Jinzhen Lin authored May 17, 2025
  
  e73b7dfd
- use ceil_div in cutlass block scaling shape check (#17918) · e23564cb
  Lain authored May 16, 2025
  
  e23564cb
15 May, 2025 1 commit

fix: typos (#18151) · a9944aab

omahs authored May 15, 2025


Signed-off-by: omahs <73983677+omahs@users.noreply.github.com>

a9944aab

14 May, 2025 4 commits
- [Kernel] Have rotary embeddings support tensors (#18046) · d93c976a
  Lucas Wilkinson authored May 14, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
```
  d93c976a
- Modularize fused experts and integrate PPLX kernels (#15956) · f9c069c8
  bnellnm authored May 14, 2025
  
  f9c069c8
- [Bugfix] fix moe marlin `topk_weight` loading (#18080) · d4154c35
  Jinzhen Lin authored May 14, 2025
```
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  d4154c35
- [AMD][torch.compile] Enable silu+fp8_quant fusion for rocm (#18082) · 7b2f28de
  Charlie Fu authored May 14, 2025
```
Signed-off-by: charlifu <charlifu@amd.com>
```
  7b2f28de
13 May, 2025 3 commits
- Fix Broken macro for cutlass moe (#18049) · e57e4d6e
  Driss Guessous authored May 12, 2025
```
Signed-off-by: drisspg <drisspguessous@gmail.com>
```
  e57e4d6e
- [ROCm]: Fix build from source failure with gcc14 and ROCm 6.3 (#13779) · d8487ef5
  Arjun Kathuria authored May 13, 2025
```
Signed-off-by: Arjun Kathuria <arjun.kathuria8@gmail.com>
```
  d8487ef5
- Implements dual-chunk-flash-attn backend for dual chunk attention with sparse... · 60f76243
  Tao He authored May 13, 2025
```
Implements dual-chunk-flash-attn backend for dual chunk attention with sparse attention support (#11844)
```
  60f76243
11 May, 2025 1 commit
- [Kernel] fp4 marlin kernel (#17687) · d74e5f37
  Jinzhen Lin authored May 11, 2025
```
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
```
  d74e5f37
09 May, 2025 1 commit
- [Hardware/NVIDIA/Kernel] Enable nvidia/DeepSeek-R1-FP4 Model (#16362) · 0c0fdae8
  Pavani Majety authored May 09, 2025
  
  0c0fdae8
08 May, 2025 4 commits
- Add cutlass support for blackwell fp8 blockwise gemm (#14383) · 376786fa
  Shu Wang authored May 08, 2025
```
Signed-off-by: Shu Wang <shuw@nvidia.com>
```
  376786fa
- [Easy] Eliminate c10::optional usage in vllm/csrc (#17819) · f50dcb7c
  Lu Fang authored May 08, 2025
  
  f50dcb7c
- [Hardware][Power] Enable compressed tensor W8A8 INT8 quantization for POWER (#17153) · e515668e
  Akash kaothalkar authored May 08, 2025
```
Signed-off-by: Akash Kaothalkar <akash.kaothalkar@ibm.com>
Co-authored-by: Akash Kaothalkar <akash.kaothalkar@ibm.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  e515668e
- [Kernel][Hardware][AMD] Bf16 mfma opt for ROCm skinny GEMMs (#17071) · 5a499e70
  Hashem Hashemi authored May 07, 2025
```
Signed-off-by: Hashem Hashemi <hashem.hashemi@amd.com>
Signed-off-by: charlifu <charlifu@amd.com>
Co-authored-by: charlifu <charlifu@amd.com>
```
  5a499e70
07 May, 2025 5 commits
- [ROCm][FP8][Kernel] FP8 quantization fused into Custom Paged Attention (#17139) · 32aa74c0
  Gregory Shtrasberg authored May 07, 2025
```
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
```
  32aa74c0
- Make key optional for rotary embedding (#17566) · 98c89e16
  Yong Hoon Shin authored May 07, 2025
```
Signed-off-by: Yong Hoon Shin <yhshin@meta.com>
```
  98c89e16
- [Kernel] Use fused rmsnorm for some models like qwen3 series (#17735) · f80ae5bd
  Wanrui Dai authored May 07, 2025
```
Signed-off-by: evian <eviantai@u.nus.edu>
Co-authored-by: evian <eviantai@u.nus.edu>
```
  f80ae5bd
- [Kernel] GGUF MoeVec kernel (#16780) · 1a45a613
  Szymon Ożóg authored May 07, 2025
```
Signed-off-by: SzymonOzog <szymon.ozog@aleph-alpha.com>
Signed-off-by: SzymonOzog <szymon.ozog@gmail.com>
Signed-off-by: Isotr0py <2037008807@qq.com>
Co-authored-by: Isotr0py <2037008807@qq.com>
```
  1a45a613
- Removed unused marlin cuda code (#17684) · a17cef70
  Michael Goin authored May 06, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  a17cef70
05 May, 2025 1 commit
- [Kernel] some optimizations for dense marlin and moe marlin (#16850) · 1d0c9d6b
  Jinzhen Lin authored May 06, 2025
```
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
```
  1d0c9d6b
02 May, 2025 1 commit
- permute/unpermute kernel for moe optimization (#14568) · 3e887d2e
  Caleb_Du authored May 03, 2025
```
Signed-off-by: Caleb_Du <Caleb_Du@zju.edu.cn>
```
  3e887d2e
01 May, 2025 2 commits

[ROCM] Add gfx950 to the custom attention archs (#16034) · 811a6c09

Juan Villamizar authored May 01, 2025


Signed-off-by: jpvillam <Juan.Villamizar@amd.com>
Signed-off-by: seungrokjung <seungrok.jung@amd.com>
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Co-authored-by: seungrokjung <seungrok.jung@amd.com>
Co-authored-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>

811a6c09

[torch.compile] Add torch inductor pass for fusing silu_and_mul with... · 460a2b11

Sage Moore authored May 01, 2025


[torch.compile] Add torch inductor pass for fusing silu_and_mul with subsequent scaled_fp8_quant operations (#10867)
Signed-off-by: Sage Moore <sage@neuralmagic.com>

460a2b11

29 Apr, 2025 4 commits
- [Misc][ROCm] Exclude `cutlass_mla_decode` for ROCm build (#17289) · 06ffc7e1
  TY-AMD authored Apr 30, 2025
```
Signed-off-by: Tianyuan Wu <Tianyuan.Wu@amd.com>
```
  06ffc7e1
- `pre-commit autoupdate` (#17380) · 40896bdf
  Harry Mellor authored Apr 29, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  40896bdf
- skip cutlass_mla_decode · 2d465ec7
  zhuwenwen authored Apr 29, 2025
  
  2d465ec7
- [Bugfix] Fix `numel()` downcast in fused_layernorm_dynamic_per_token_quant.cu (#17316) · d6da8a8f
  Richard Barnes authored Apr 28, 2025
  
  d6da8a8f
28 Apr, 2025 1 commit
- [Bugfix] Fix cutlass dispatch for fp8/int8 to properly invoke M<=16 c… (#16751) · c12df53b
  TherLF authored Apr 28, 2025
```
Signed-off-by: Ther-LF <2639852836@qq.com>
```
  c12df53b
27 Apr, 2025 1 commit
- [NVIDIA] Support Cutlass MLA for Blackwell GPUs (#16032) · ed7a29d9
  Kaixi Hou authored Apr 27, 2025
```
Signed-off-by: kaixih <kaixih@nvidia.com>
```
  ed7a29d9
26 Apr, 2025 2 commits
- Allocate kv_cache with stride order (#16605) · 9e96f56e
  Shu Wang authored Apr 26, 2025
```
Signed-off-by: shuw <shuw@nvidia.com>
```
  9e96f56e
- [MISC][AMD] Add unused annotation to rocm kernel file (#17097) · c8e5be35
  Lu Fang authored Apr 25, 2025
```
Signed-off-by: Lu Fang <lufang@fb.com>
```
  c8e5be35
22 Apr, 2025 2 commits
- [Performance][ROCm] Add skinny gemms for unquantized linear on ROCm (#15830) · 188b7f9b
  Charlie Fu authored Apr 21, 2025
```
Signed-off-by: charlifu <charlifu@amd.com>
Co-authored-by: Tyler Michael Smith <tysmith@redhat.com>
```
  188b7f9b
- [Kernel] Add expert_map support to Cutlass FP8 MOE (#16861) · 7b8a2ab7
  Varun Sundar Rabindranath authored Apr 21, 2025
```
Signed-off-by: varun sundar rabindranath <vsundarr@redhat.com>
Co-authored-by: varun sundar rabindranath <vsundarr@redhat.com>
```
  7b8a2ab7
18 Apr, 2025 1 commit
- [BugFix] Accuracy fix for llama4 int4 - improperly casted scales (#16801) · 7eb42556
  Lucas Wilkinson authored Apr 18, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
```
  7eb42556
16 Apr, 2025 1 commit
- [Bugfix][Kernel] fix potential cuda graph broken for merge_attn_states kernel (#16693) · e82ee40d
  DefTruth authored Apr 16, 2025
```
Signed-off-by: DefTruth <qiustudent_r@163.com>
```
  e82ee40d
15 Apr, 2025 1 commit

[Kernel] moe wna16 marlin kernel (#14447) · d06ba4ed

Jinzhen Lin authored Apr 15, 2025


Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
Co-authored-by: Michael Goin <michael@neuralmagic.com>
Co-authored-by: mgoin <mgoin64@gmail.com>

d06ba4ed

12 Apr, 2025 1 commit
- fix: spelling (#16466) · 4a3a5187
  Tianer Zhou authored Apr 12, 2025
```
Signed-off-by: Tianer Zhou <ezhoureal@gmail.com>
```
  4a3a5187
11 Apr, 2025 1 commit
- [Kernel] support merge_attn_states CUDA kernel, 3x speedup (#16173) · e9528f6d
  DefTruth authored Apr 11, 2025
```
Signed-off-by: DefTruth <qiustudent_r@163.com>
```
  e9528f6d