Commits · 02bf31ef29a31e6dc05cc16f2c8411da6fdf1b8c · zhaoyu6 / sglang

21 Jun, 2025 1 commit
- [fix] PD disaggregation when enable mtp and tp!=dp (#7420) · 02bf31ef
  Atream authored Jun 22, 2025
  
  02bf31ef
20 Jun, 2025 2 commits
- [Feature] Comprehensive Hybrid Parallelism Support (#6389) · e879d8b7
  Cheng Wan authored Jun 20, 2025
  
  e879d8b7
- Fix judgment condition for enabling Deepseek V3/R1 shared expert fusion optimization (#7371) · a06912ad
  Li Hui authored Jun 20, 2025
  
  a06912ad
18 Jun, 2025 1 commit

Update python API of activation, topk, norm and rope and remove vllm dependency (#6614) · 094c116f

YanbingJiang authored Jun 18, 2025


Co-authored-by: Wu, Chunyuan <chunyuan.wu@intel.com>
Co-authored-by: jianan-gu <jianan.gu@intel.com>
Co-authored-by: sdp <sdp@gnr799219.jf.intel.com>

094c116f

17 Jun, 2025 4 commits
- fix: resolve b200 dsv3 mtp issue (#7286) · 4f204db5
  Yineng Zhang authored Jun 17, 2025
  
  4f204db5
- Fix AWQ Dequant and Weight Loading of deepseek v2 (#6842) · 3eb4a800
  AniZpZ authored Jun 18, 2025
  
  3eb4a800
- Fix Deepseek R1 0528 FP4 tensor name mismatch issue during weights loading. (#7164) · 8c16da33
  Charles Chen authored Jun 17, 2025
  
  8c16da33
- feat: mtp support dp-attention (#6081) · 10d60cd4
  u4lr451 authored Jun 17, 2025
```
Co-authored-by: austindeng <austindeng@tencent.com>
Co-authored-by: tianqilin.99 <tianqilin.99@bytedance.com>
Co-authored-by: Qiaolin Yu <liin1211@outlook.com>
Co-authored-by: ch-wan <cwan39@gatech.edu>
```
  10d60cd4
16 Jun, 2025 1 commit
- [fix] fix DeepGEMM blackwell input quant & ut & fix style and log (#7247) · 5ca07eed
  JieXin Liang authored Jun 17, 2025
  
  5ca07eed
15 Jun, 2025 2 commits
- Fix error when disabling new DeepGEMM (#7198) · 349bb2c9
  fzyzcjy authored Jun 15, 2025
  
  349bb2c9
- [fix] fix determine_num_fused_shared_experts (#7180) · 55561e25
  JieXin Liang authored Jun 15, 2025
  
  55561e25
14 Jun, 2025 4 commits
- [fix] fix dsv3 weight loader tqdm and simplify shared experts fusion (#7181) · ed54bf9d
  JieXin Liang authored Jun 15, 2025
  
  ed54bf9d
- Fix shared experts fusion + weight requant (#7177) · b57d87c2
  fzyzcjy authored Jun 14, 2025
  
  b57d87c2
- Support new DeepGEMM (#7172) · 93cec433
  fzyzcjy authored Jun 14, 2025
  
  93cec433
- Refactor DeepGEMM integration (#7150) · b4c41f72
  fzyzcjy authored Jun 14, 2025
  
  b4c41f72
13 Jun, 2025 3 commits
- Re-quantize DeepSeek model weights to support DeepGEMM new input format (#7156) · 5b1afa78
  fzyzcjy authored Jun 14, 2025
  
  5b1afa78
- Remove 200us slow concat kernel (part 2: srt) (#7020) · c49c1d92
  fzyzcjy authored Jun 14, 2025
  
  c49c1d92
- filter by num_hidden_layers (#7056) · 2f4ec752
  pansicheng authored Jun 13, 2025
  
  2f4ec752
10 Jun, 2025 1 commit
- Support both approximate and exact expert distribution collection (#6964) · f6ebba53
  fzyzcjy authored Jun 10, 2025
  
  f6ebba53
09 Jun, 2025 1 commit
- Minor remove one kernel for DeepSeek (#6977) · de1350ea
  fzyzcjy authored Jun 09, 2025
  
  de1350ea
08 Jun, 2025 3 commits
- Fuse routed scaling factor in deepseek (#6970) · 3712abfa
  Xiaoyu Zhang authored Jun 09, 2025
  
  3712abfa
- Revert "Fuse routed scaling factor in topk_reduce kernel (#6220)" (#6968) · 1fb76ebb
  Yineng Zhang authored Jun 07, 2025
  
  1fb76ebb
- [DeepseekR1-FP4] Add Support for nvidia/DeepSeekR1-FP4 model (#6853) · c2c4f57f
  Pavani Majety authored Jun 07, 2025
```
Signed-off-by: Pavani Majety <pmajety@nvidia.com>
```
  c2c4f57f
07 Jun, 2025 3 commits
- Fuse routed scaling factor in topk_reduce kernel (#6220) · 515ef4fa
  Xiaoyu Zhang authored Jun 08, 2025
  
  515ef4fa
- [sgl-kernel] update deepgemm (#6942) · 22fe7878
  JieXin Liang authored Jun 07, 2025
  
  22fe7878
- [fix] logical_to_all_physical_map index 256 is out of bounds in EP parallel. (#6767) · f8eaaab8
  miter authored Jun 07, 2025
```
Signed-off-by: miter <miterv@outlook.com>
```
  f8eaaab8
06 Jun, 2025 1 commit
- AITER backend extension and workload optimizations (#6838) · b819381f
  HAI authored Jun 05, 2025
```
Co-authored-by: wunhuang <wunhuang@amd.com>
Co-authored-by: Hubert Lu <Hubert.Lu@amd.com>
```
  b819381f
04 Jun, 2025 2 commits
- Set `num_fused_shared_experts` as `num_shared_experts` when shared_experts... · 81964328
  Cheng Wan authored Jun 04, 2025
```
Set `num_fused_shared_experts` as `num_shared_experts` when shared_experts fusion is not disabled (#6736)
```
  81964328
- [Refactor] Rename `n_share_experts_fusion` as `num_fused_shared_experts` (#6735) · 8a548052
  Cheng Wan authored Jun 03, 2025
  
  8a548052
03 Jun, 2025 1 commit
- Fix wrong weight reference in dynamic EPLB (#6818) · 0ea330ca
  fzyzcjy authored Jun 03, 2025
  
  0ea330ca
30 May, 2025 1 commit
- Fix shared experts fusion error (#6289) · 69dd878b
  Li Hui authored May 30, 2025
  
  69dd878b
29 May, 2025 1 commit
- [fix][RL] Fix DeepSeekV3ForCausalLM.post_load_weights for multiple update weight (#6265) · 51cdd81f
  Zilin Zhu authored May 30, 2025
  
  51cdd81f
28 May, 2025 2 commits
- Speed up when having padding tokens two-batch overlap (#6668) · 31589e17
  fzyzcjy authored May 29, 2025
```
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
```
  31589e17
- Fuse routed_scaling_factor in DeepSeek (#6710) · 541a985f
  fzyzcjy authored May 29, 2025
  
  541a985f
27 May, 2025 2 commits
- DeepSeek: enable none block-quant FP8 quantizations (#6638) · 183d9f96
  HAI authored May 27, 2025
  
  183d9f96
- Support TP in attention for two batch overlap (#6634) · 32cd7070
  fzyzcjy authored May 27, 2025
  
  32cd7070
26 May, 2025 2 commits
- Tiny fix missing expert location dispatch info (#6620) · 0ca3e568
  fzyzcjy authored May 26, 2025
  
  0ca3e568
- refactor qwen moe code, use communicator to support tp+dp (#6581) · 65f09131
  Yi Zhang authored May 26, 2025
  
  65f09131
25 May, 2025 2 commits
- Support overlapping two batches (#4068) · 0d477880
  fzyzcjy authored May 25, 2025
  
  0d477880
- Add back DeepSeek non-TBO branches (#6578) · b2388433
  fzyzcjy authored May 25, 2025
  
  b2388433