Commits · 9e27b5e4e4472d51cbac7778de86382e86f8dd48 · OpenDAS / vllm_cscc

03 Jul, 2025 1 commit
- 增加dpsk awq mtp推理的支持 · f9795c8c
  zhuwenwen authored Jul 03, 2025
  
  f9795c8c
01 Jul, 2025 1 commit
- [Model] Add Ernie4.5 and Ernie4.5MoE · d48d7497
  zhuwenwen authored Jul 01, 2025
  
  d48d7497
30 Jun, 2025 1 commit
- 修复channelwise w8a8 预热没有区分卡造成的triton kernel崩溃问题，另外增加block-int8的moe 拿到config步骤 · c0c0eb69
  gaoqiong authored Jun 30, 2025
  
  c0c0eb69
27 Jun, 2025 1 commit
- [Model] Add support for MiniMaxM1ForCausalLM (shares architecture with MiniMaxText01ForCausalLM) · 95d58c0d
  zhuwenwen authored Jun 27, 2025
  
  95d58c0d
21 Jun, 2025 2 commits
- support ndhwc of qwen2-vl and qwen2.5-vl · 5f71c3f8
  zhuwenwen authored Jun 21, 2025
  
  5f71c3f8
- skip is_rocm_aiter_moe_enabled and remove qwen unused int8 code · bbd14169
  zhuwenwen authored Jun 21, 2025
  
  bbd14169
17 Jun, 2025 1 commit

[fix]1.修复medusa、scorer等并行解码单测；2.修复moe... · 8db76782

王敏 authored Jun 17, 2025

[fix]1.修复medusa、scorer等并行解码单测；2.修复moe kernel单测问题，优化代码；3.修复rejection_sampler中test_compare_nonflashinfer_backend单测问题

8db76782

13 Jun, 2025 3 commits
- [fix]修复并行解码eagle和mlp相关单测问题 · 08cb5d8f
  王敏 authored Jun 13, 2025
  
  08cb5d8f
- 解耦w8a8 的gemm以及moe算子相关实现 · 942368c7
  gaoqiong authored Jun 13, 2025
  
  942368c7
- [New Model]: Support Qwen3 Embedding & Reranker · 9bcbaafc
  zhuwenwen authored Jun 13, 2025
  
  9bcbaafc
12 Jun, 2025 2 commits
- 修复qwen3-moe的awq配置导致fp16加载错误，修复dpsk-moe手写算子首字耗时增加问题 · 99ffef47
  zhuwenwen authored Jun 12, 2025
  
  99ffef47
- update qwen2&2.5-vl prefill interface and fix pixtral run error · 4fd5389b
  zhuwenwen authored Jun 12, 2025
  
  4fd5389b
10 Jun, 2025 1 commit
- Update qwen3_moe.py · 07c69390
  zhuwenwen authored Jun 10, 2025
  
  07c69390
06 Jun, 2025 3 commits
- [Model] Optimize nemotron_h implementation (#19249) · 7661e92e
  Jee Jee Li authored Jun 06, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  7661e92e
- [Model] Add VLLM_USE_NN to use nn layout · 6880bf15
  zhuwenwen authored Jun 06, 2025
  
  6880bf15
- [Bugfix] Fix EAGLE vocab embedding construction for Llama 70B (#19033) · 3465b87e
  Benjamin Chislett authored Jun 05, 2025
```
Signed-off-by: Benjamin Chislett <benjamin.chislett@centml.ai>
```
  3465b87e
05 Jun, 2025 3 commits
- [Model] NemotronH support (#18863) · cb6d572e
  Luis Vega authored Jun 05, 2025
```
Signed-off-by: Luis Vega <2478335+vegaluisjose@users.noreply.github.com>
Co-authored-by: Luis Vega <2478335+vegaluisjose@users.noreply.github.com>
```
  cb6d572e
- 新增qwen3 fusemoe优化支持 · 1cb37dab
  yangql authored Jun 05, 2025
  
  1cb37dab
- 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 3 commits
- 修复qwen3-moe-awq的bug问题 · 1a4e4cad
  yangql authored Jun 04, 2025
  
  1a4e4cad
- [CI/Build][Bugfix] Ensure compatibility with transformers 4.52 (#18678) · 01dc9a76
  Cyrus Leung authored Jun 04, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  01dc9a76
- Improve the output precision of embedding models (#19092) · 35cf32df
  wang.yuqi authored Jun 04, 2025
  
  35cf32df
03 Jun, 2025 3 commits
- [Misc] Add SPDX-FileCopyrightText (#19100) · 02f0c7b2
  Simon Mo authored Jun 03, 2025
```
Signed-off-by: simon-mo <simon.mo@hey.com>
```
  02f0c7b2
- [Misc] Update `WeightsMapper` for qwen2-vl/qwen2.5-vl (#19054) · ec2dcd80
  Isotr0py authored Jun 03, 2025
```
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  ec2dcd80
- Add tarsier model support (#18985) · 1282bd81
  汪志鹏 authored Jun 03, 2025
```
Signed-off-by: 汪志鹏 <wangzhipeng628@gmail.com>
```
  1282bd81
02 Jun, 2025 1 commit

[Model] enable data parallel for Llama4 vision encoder (#18368) · ebb1ec93

jennyyyyzhen authored Jun 02, 2025


Signed-off-by: yzhen <yzhen@devgpu093.cco2.facebook.com>
Co-authored-by: yZhen <yZhen@fb.com>
Co-authored-by: yzhen <yzhen@devgpu093.cco2.facebook.com>

ebb1ec93

01 Jun, 2025 2 commits

[LoRA] Support dynamically initialize `packed_modules_mapping` for VLM with... · a35ca765

Isotr0py authored Jun 01, 2025


[LoRA] Support dynamically initialize `packed_modules_mapping` for VLM with arbitrary components (#18987)
Signed-off-by: isotr0py <2037008807@qq.com>
Signed-off-by: Isotr0py <2037008807@qq.com>

a35ca765

[Bugfix] Fix EAGLE3 broken logits (#18909) · 1bc86a3d
Benjamin Chislett authored May 31, 2025
```
Signed-off-by: Benjamin Chislett <benjamin.chislett@centml.ai>
```
1bc86a3d

30 May, 2025 6 commits
- [VLM] Add PP support and fix GPTQ inference for Ovis models (#18958) · 5a864163
  Isotr0py authored May 31, 2025
```
Signed-off-by: isotr0py <2037008807@qq.com>
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  5a864163
- [Feature] minicpm eagle support (#18943) · e1fadf11
  Shawn Huang authored May 30, 2025
```
Signed-off-by: huangyuxiang03 <huangyx0321@gmail.com>
Co-authored-by: huangyuxiang03 <huangyx0321@gmail.com>
```
  e1fadf11
- [Model] Use in-place adds in SigLIP (#18922) · c3bb9f23
  Lukas Geiger authored May 30, 2025
```
Signed-off-by: Lukas Geiger <lukas.geiger94@gmail.com>
```
  c3bb9f23
- [Deprecation] Remove mean pooling default for `Qwen2EmbeddingModel` (#18913) · 4f4a6b84
  Cyrus Leung authored May 30, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  4f4a6b84
- [Model] Use AutoWeightsLoader for mamba2 (#18918) · 3987e2ae
  iLeGend authored May 30, 2025
```
Signed-off-by: iLeGend <824040212@qq.com>
```
  3987e2ae
- [Misc] Update type annotation for rotary embedding `base` (#18914) · 1aa2f81b
  Cyrus Leung authored May 30, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  1aa2f81b
29 May, 2025 2 commits
- [LoRA] Add LoRA support for InternVL (#18842) · 34d6c447
  Jee Jee Li authored May 29, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  34d6c447
- Prevent the cross-encoder logic from being applied to classification tasks (#18838) · 515b413e
  Maximilien de Bayser authored May 28, 2025
```
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
```
  515b413e
28 May, 2025 2 commits
- [Misc] fix olmoe model layer can't laod in tp gt 1 (#18828) · c68b5c63
  rongfu.leng authored May 29, 2025
```
Signed-off-by: rongfu.leng <rongfu.leng@daocloud.io>
```
  c68b5c63
- [Bugfix] Fix nomic max_model_len (#18755) · 3e9ce609
  wang.yuqi authored May 28, 2025
  
  3e9ce609
27 May, 2025 2 commits
- [Core] Automatically cast multi-modal input dtype (#18756) · 696259ca
  Cyrus Leung authored May 27, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  696259ca
- [feat]适配sgl moe_fused_gate kernel · 0b229519
  王敏 authored May 27, 2025
  
  0b229519