Commits · debd6bbf0951222c9ad7a1a91b958027f2ed0782 · OpenDAS / vllm_cscc · GitLab

12 Mar, 2025 4 commits
- [Kernel] Add ModelOpt FP4 Checkpoint Support (#12520) · debd6bbf
  Pavani Majety authored Mar 11, 2025
```
Signed-off-by: Pavani Majety <pmajety@nvidia.com>
```
  debd6bbf
- [Kernel] GGUF MoE kernel (#14613) · e22ee1e7
  Szymon Ożóg authored Mar 12, 2025
```
Signed-off-by: SzymonOzog <szymon.ozog@aleph-alpha.com>
```
  e22ee1e7
- [Core] Refactor `QKVCrossParallelLinear` implementation to support BNB 4-bit quantization (#14545) · e392d858
  Isotr0py authored Mar 12, 2025
```
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  e392d858
- [Model] Extend Ultravox to accept audio longer than 30s (#13631) · 80e78d02
  Farzad Abdolhosseini authored Mar 11, 2025
```
Signed-off-by: Farzad Abdolhosseini <farzad@fixie.ai>
```
  80e78d02
11 Mar, 2025 5 commits
- dynamic distpatch of fp8 kernels (#14245) · a1c8f379
  Jeff Daily authored Mar 11, 2025
```
Signed-off-by: Jeff Daily <jeff.daily@amd.com>
```
  a1c8f379
- [VLM] Cleanup siglip legacy code and fix broken paligemma multimodal processor (#14602) · 1477ffc3
  Isotr0py authored Mar 11, 2025
```
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  1477ffc3
- [Perf]:Optimize qwen2-vl to reduce cudaMemcpyAsync (#14377) · 70b808fe
  yexin(叶鑫) authored Mar 11, 2025
```
Signed-off-by: cynthieye <987073381@qq.com>
```
  70b808fe
- [Bugfix] Fix FP16 overflow for DeepSeek V2 (#13232) · c982ac57
  Concurrensee authored Mar 10, 2025
```
Signed-off-by: Yida Wu <yida.wu@amd.com>
```
  c982ac57
- [Kernel] moe wna16 cuda kernel (#13321) · 90e88ab7
  Jinzhen Lin authored Mar 11, 2025
```
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
```
  90e88ab7
10 Mar, 2025 4 commits
- [Bugfix][v1] fixed llava-hf/llava-1.5-7b-hf is broken on V1 (#14554) · 92b0ce2a
  Chauncey authored Mar 11, 2025
```
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  92b0ce2a
- Correct capitalisation: `VLLM` -> `vLLM` (#14562) · 3b352a2f
  Harry Mellor authored Mar 10, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  3b352a2f
- [V1][Bugfix] Fix handing of `second_per_grid_ts` for Qwen2-VL & Qwen2.5-VL (#14548) · dea985ae
  Roger Wang authored Mar 10, 2025
```
Signed-off-by: Roger Wang <ywang@roblox.com>
```
  dea985ae
- [Kernel] Add more dtype support for GGUF kernels (#14043) · 89cdaa83
  Szymon Ożóg authored Mar 10, 2025
```
Signed-off-by: SzymonOzog <szymon.ozog@aleph-alpha.com>
Signed-off-by: SzymonOzog <szymon.ozog@gmail.com>
```
  89cdaa83
09 Mar, 2025 1 commit
- [Bugfix] Revert QKVCrossParallelLinear usage in Mllama to keep BNB quantization work (#14498) · fb16eea4
  Isotr0py authored Mar 09, 2025
```
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  fb16eea4
08 Mar, 2025 3 commits
- [Misc] add `use_tqdm_on_load` to reduce logs (#14407) · 0b7f06b4
  Aaron Pham authored Mar 08, 2025
```
Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
```
  0b7f06b4
- [VLM] Add TP support for Phi-4-MM (#14453) · 03fe18ae
  Isotr0py authored Mar 08, 2025
```
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  03fe18ae
- Revert "[Perf] Reduce MLA CPU overheads in V1 (#14384)" (#14471) · ca7a2d5f
  Tyler Michael Smith authored Mar 08, 2025
  
  ca7a2d5f
07 Mar, 2025 6 commits
- [Misc] Add Phi4-MM example (#14343) · 952a0749
  Jee Jee Li authored Mar 08, 2025
```
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
```
  952a0749
- [Kernel] optimize performance of gptq marlin kernel when n is small (#14138) · d0feea31
  Jinzhen Lin authored Mar 08, 2025
```
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
```
  d0feea31
- Fix missing `kv_caches` and `attn_metadata` in `OpenVINOCausalLM` (#14271) · f7a6bd0f
  Harry Mellor authored Mar 07, 2025
```
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
```
  f7a6bd0f
- [Bugfix] Clean up multi-modal processors (#14417) · 05fb6718
  Cyrus Leung authored Mar 07, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  05fb6718
- [FP8] Refactor apply_fp8_linear and apply_fp8_linear_generic into an object (#14390) · e1744502
  Luka Govedič authored Mar 07, 2025
```
Signed-off-by: luka <luka@neuralmagic.com>
```
  e1744502
- [Perf] Reduce MLA CPU overheads in V1 (#14384) · dae68969
  Lucas Wilkinson authored Mar 06, 2025
```
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
```
  dae68969
06 Mar, 2025 6 commits
- [Bugfix] Fix use_direct_call condition in FusedMoE layer for (#14382) · 958adce4
  Tyler Michael Smith authored Mar 06, 2025
```
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
```
  958adce4
- [Distributed] Add enable_expert_parallel arg (#14305) · cc2f9b32
  Tyler Michael Smith authored Mar 06, 2025
```
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
```
  cc2f9b32
- [Core] Optimizing cross-attention `QKVParallelLinear` computation (#12325) · 69ff99fd
  Nicolò Lucchesi authored Mar 06, 2025
```
Signed-off-by: NickLucche <nlucches@redhat.com>
Signed-off-by: NickLucche <nick@nlucches-4xa100.c.openshift-330514.internal>
Co-authored-by: NickLucche <nick@nlucches-4xa100.c.openshift-330514.internal>
```
  69ff99fd
- [V1][VLM][Pixtral-HF] Support Pixtral-HF on V1 (#14275) · 5d802522
  lkchen authored Mar 06, 2025
```
Signed-off-by: Linkun Chen <github@lkchen.net>
```
  5d802522
- [Model] Update Paligemma multimodal processing with PromptUpdate (#14015) · 17699280
  kYLe authored Mar 06, 2025
```
Signed-off-by: Kyle Huang <kylhuang@nvidia.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
```
  17699280
- [Bugfix][Structured Output] Support outlines engine with reasoning outputs for DeepSeek R1 (#14114) · f5f7f00c
  Ce Gao authored Mar 06, 2025
  
  f5f7f00c
05 Mar, 2025 5 commits
- [LoRA] Remove linear hack outside transformers backend (#14177) · e17e4488
  Isotr0py authored Mar 05, 2025
```
Signed-off-by: Isotr0py <2037008807@qq.com>
```
  e17e4488
- [Misc][V1] Avoid using `envs.VLLM_USE_V1` in mm processing (#14256) · ec79b67c
  Roger Wang authored Mar 04, 2025
```
Signed-off-by: Roger Wang <ywang@roblox.com>
```
  ec79b67c
- [V1] EP/TP MoE + DP Attention (#13931) · 72c62eae
  Tyler Michael Smith authored Mar 05, 2025
  
  72c62eae
- [Model] New model support for Phi-4-multimodal-instruct (#14119) · 0a995d54
  Congcong Chen authored Mar 04, 2025
  
  0a995d54
- [Bugfix] Fix gptq_marlin for deepseek-v3 (#13750) · 0df25101
  rainkert authored Mar 05, 2025
```
Signed-off-by: dangshunya <dangshunya@baichuan-inc.com>
Co-authored-by: dangshunya <dangshunya@baichuan-inc.com>
```
  0df25101
04 Mar, 2025 4 commits
- Clean up unused padding_idx variables across many model definitions (#13240) · 4f5b059f
  Tyler Michael Smith authored Mar 04, 2025
```
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
```
  4f5b059f
- [V1][Molmo] Fix get_multimodal_embeddings() in molmo.py (#14161) · b3cf368d
  lkchen authored Mar 04, 2025
  
  b3cf368d
- [core] moe fp8 block quant tuning support (#14068) · bb5b6403
  Divakar Verma authored Mar 03, 2025
```
Signed-off-by: Divakar Verma <divakar.verma@amd.com>
```
  bb5b6403
- [Model] Add support for GraniteMoeShared models (#13313) · c060b714
  Travis Johnson authored Mar 03, 2025
```
Signed-off-by: Travis Johnson <tsjohnso@us.ibm.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
```
  c060b714
03 Mar, 2025 2 commits
- [Kernel] Optimize moe intermediate_cache usage (#13625) · 19d98e0c
  Michael Goin authored Mar 03, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  19d98e0c
- [Bugfix] Allow shared_experts skip quantization for DeepSeekV2/V3 (#14100) · 2b04c209
  Michael Goin authored Mar 03, 2025
```
Signed-off-by: mgoin <mgoin64@gmail.com>
```
  2b04c209