Commits · 6880bf15fcbb29f16121df987200ffa7db15d771 · OpenDAS / vllm_cscc

06 Jun, 2025 1 commit
- [Model] Add VLLM_USE_NN to use nn layout · 6880bf15
  zhuwenwen authored Jun 06, 2025
  
  6880bf15
05 Jun, 2025 3 commits
- 新增qwen3 fusemoe优化支持 · 1cb37dab
  yangql authored Jun 05, 2025
  
  1cb37dab
- Update w8a8_int8.py · 53250530
  gaoqiong authored Jun 05, 2025
  
  53250530
- 修改deepseelk block-int8 权重处理流程，增加per-channel bestconfig配置以及首次triton warmup代码 · 40b94473
  gaoqiong authored Jun 05, 2025
  
  40b94473
04 Jun, 2025 1 commit
- [feat]add 120cu nn fused_moe configs of qwen3, mixtral8*7b and mixtral8*22b · d36deb1a
  zhuwenwen authored Jun 04, 2025
  
  d36deb1a
29 May, 2025 1 commit
- 增加w8a8 线性gemm config优化 · f52afe3f
  gaoqiong authored May 29, 2025
  
  f52afe3f
28 May, 2025 1 commit
- [feat]量化模型添加支持 moe_fused_gate kernel，并使用VLLM_ENABLE_MOE_FUSED_GATE环境变量控制开关，默认打开 · 1016e339
  王敏 authored May 28, 2025
  
  1016e339
27 May, 2025 2 commits
- [feat]适配sgl moe_fused_gate kernel · 0b229519
  王敏 authored May 27, 2025
  
  0b229519
- 修复0.8.5 config找不到的bug · 823b3332
  gaoqiong authored May 27, 2025
  
  823b3332
23 May, 2025 1 commit
- tbo增加ds int8支持，修改tb zerooverhead环境变量到envs.py中统一管理 · dbbb148b
  lizhigong authored May 23, 2025
  
  dbbb148b
22 May, 2025 2 commits
- 优化awq权重转换的方式，修复可能会导致蒸馏模型加载时报错的bug · a94ed3ea
  yangql authored May 22, 2025
  
  a94ed3ea
- fix tbo to support deepseek · 0cc7c880
  lizhigong authored May 22, 2025
  
  0cc7c880
16 May, 2025 1 commit
- 修复zero-overhead首字正确性问题，zero-overhead不使用默认流调整，增加two-batch-overlap功能 · 2a935929
  lizhigong authored May 16, 2025
  
  2a935929
13 May, 2025 1 commit
- add qwen3 moe configs · 09e372e7
  zhuwenwen authored May 13, 2025
```
support telechat2 and glm4 nn layout
remove log of request_id
```
  09e372e7
09 May, 2025 6 commits
- roll back unused change · 29e922ac
  lizhigong authored May 09, 2025
  
  29e922ac
- rm debug log · 7d224eb2
  lizhigong authored May 08, 2025
  
  7d224eb2
- debug spec decode zero overhead · 0ecda6d1
  lizhigong authored May 08, 2025
  
  0ecda6d1
- add spec decode zero overhead · 01c30741
  lizhigong authored Apr 27, 2025
  
  01c30741
- debug v0 zero overhead schedule · 4ff58b66
  lizhigong authored Apr 14, 2025
  
  4ff58b66
- add v0 zero overhead · 54294854
  lizhigong authored Apr 11, 2025
  
  54294854
08 May, 2025 1 commit
- set MOE_NN=1 · e5497373
  zhuwenwen authored May 08, 2025
  
  e5497373
30 Apr, 2025 2 commits
- skip apply_rotary_emb from vllm_fa · ffd8e40d
  zhuwenwen authored Apr 30, 2025
  
  ffd8e40d
- skip is_rocm_aiter_moe_enabled and add mla pad · daefd764
  zhuwenwen authored Apr 30, 2025
  
  daefd764
28 Apr, 2025 5 commits
- [Model] Add tuned triton fused_moe configs for Qwen3Moe (#17328) · ba41cc90
  Michael Goin authored Apr 28, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  ba41cc90
- [Model] Qwen3 Dense FP8 Compat Fixes (#17318) · dcbac4cb
  Simon Mo authored Apr 28, 2025
```
Signed-off-by: simon-mo <xmo@berkeley.edu>
```
  dcbac4cb
- [Bugfix] Fix moe weight losing all extra attrs after `process_weights_after_loading`. (#16854) · ed246203
  Charlie Fu authored Apr 28, 2025
```
Signed-off-by: charlifu <charlifu@amd.com>
```
  ed246203
- Explicitly explain quant method override ordering and ensure all overrides are ordered (#17256) · c7941cca
  Harry Mellor authored Apr 28, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  c7941cca
- Make name of `compressed-tensors` quant method consistent across vLLM (#17255) · b6dd32aa
  Harry Mellor authored Apr 28, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  b6dd32aa
26 Apr, 2025 5 commits
- [Minor] Fix lint error in main branch (#17233) · 43eea295
  Woosuk Kwon authored Apr 26, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
  43eea295
- [Bugfix] Fix Qwen2.5-Omni M-RoPE position ids generation (#16878) · de7eb10c
  Kero Liang authored Apr 27, 2025
```
Signed-off-by: imkero <kerorek@outlook.com>
```
  de7eb10c
- [Hardware][Intel-Gaudi] Update hpu-extension and update bucketing system for HPU device (#17186) · c48334d4
  Agata Dobrzyniewicz authored Apr 26, 2025
```
Signed-off-by: Agata Dobrzyniewicz <adobrzyniewicz@habana.ai>
```
  c48334d4
- [ROCm][Misc] Follow-ups for Skinny Gemms on ROCm. (#17011) · 54271bb7
  Charlie Fu authored Apr 26, 2025
```
Signed-off-by: charlifu <charlifu@amd.com>
```
  54271bb7
- [AMD][FP8][BugFix] Remove V1 check in arg_utils.py for FP8 since it is not necessary (#17215) · 68af5f6c
  rasmith authored Apr 25, 2025
```
Signed-off-by: Randall Smith <Randall.Smith@amd.com>
```
  68af5f6c
25 Apr, 2025 5 commits

[Quantization][FP8] Add support for FP8 models with input_scale for output... · a41351f3

rasmith authored Apr 25, 2025


[Quantization][FP8] Add support for FP8 models with input_scale for output projection and QK quantization (#15734)
Signed-off-by: Randall Smith <Randall.Smith@amd.com>
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Luka Govedič <lgovedic@redhat.com>

a41351f3

增加per-channel 量化策略支持 · 313d629e
gaoqiong authored Apr 25, 2025

313d629e

[Perf]Optimize rotary_emb implementation to use Triton operator for improved... · b22980a1

yexin(叶鑫) authored Apr 25, 2025


[Perf]Optimize rotary_emb implementation to use Triton operator for improved inference performance (#16457)
Signed-off-by: cynthieye <yexin93@qq.com>
Co-authored-by: MagnetoWang <magnetowang@outlook.com>

b22980a1

[Bugfix][Misc] Use TritonPlaceholderModule to defensively import triton (#15099) · 2f540455
Mengqing Cao authored Apr 25, 2025
```
Signed-off-by: Mengqing Cao <cmq0113@163.com>
```
2f540455

[FEAT] [ROCm]: AITER Fused MOE V1 Support (#16752) · eef36472

vllmellm authored Apr 25, 2025


Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
Co-authored-by: tjtanaa <tunjian.tan@embeddedllm.com>

eef36472

23 Apr, 2025 1 commit
- 更新curr_topk_ids · f9a784a7
  yangql authored Apr 23, 2025
  
  f9a784a7
22 Apr, 2025 1 commit
- 修复ds3 int8调用的功能 · cd87548a
  gaoqiong authored Apr 22, 2025
  
  cd87548a