Commits · e45bb0fe22a105cd10dfab5d2a2174189559cd12 · OpenDAS / dynamo · GitLab

08 Apr, 2026 2 commits
- feat: vllm omni disagg (#7409) · 1c199f88
  Ayush Agarwal authored Apr 08, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
  1c199f88
- feat(vllm): compute Qwen VL grid_thw from PIL images for P/D disagg (#7885) · d517fb80
  Qi Wang authored Apr 08, 2026
```
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  d517fb80
07 Apr, 2026 4 commits
- refactor: split build_gpu_mem_args into engine-specific functions (#7916) · 4cdc49c2
  Keiven C authored Apr 07, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  4cdc49c2
- feat(sglang): add video input support for aggregated serving (#7941) · b1c18bb1
  Zhuangcheng Gu authored Apr 07, 2026
```
Signed-off-by: Chokoyo <40918450+Chokoyo@users.noreply.github.com>
```
  b1c18bb1
- fix: broken vLLM nightly gpu_2 test and vLLM API compatibility errors (#7865) · 3ad7b7c8
  Keiven C authored Apr 07, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  3ad7b7c8
- feat: add a few TRT-LLM example support for GPU-parallel test execution (#7880) · ad2205eb
  Keiven C authored Apr 06, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  ad2205eb
06 Apr, 2026 1 commit
- perf(multimodal): strip inline base64 from TCP transport in frontend decoding (#7895) · 6b8545fc
  Qi Wang authored Apr 06, 2026
```
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  6b8545fc
03 Apr, 2026 1 commit
- chore(multimodal): Cleanup multimodal docs and consolidate launch scripts (#7845) · dacb2980
  Ryan McCormick authored Apr 03, 2026
  
  dacb2980
02 Apr, 2026 3 commits
- refactor(vLLM): Move video support from example to backend (#7663) · 4791aaaa
  Ryan McCormick authored Apr 02, 2026
  
  4791aaaa
- feat: GPU-parallel test runner with VRAM-aware scheduling (#7560) · 6dc85fbc
  Keiven C authored Apr 02, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  6dc85fbc
- feat: add examples and unit tests for vLLM aggregated serving on XPU (#7146) · dd7ceb4a
  Yi Yao authored Apr 02, 2026
```
Signed-off-by: Yi Yao <yi.a.yao@intel.com>
Co-authored-by: Ryan McCormick <rmccormick@nvidia.com>
```
  dd7ceb4a
01 Apr, 2026 1 commit
- test: Add ci coverage for trtllm multimodal raw embeddings (#7540) · c84c0934
  Indrajit Bhosale authored Apr 01, 2026
```
Signed-off-by: Indrajit Bhosale <iamindrajitb@gmail.com>
```
  c84c0934
31 Mar, 2026 2 commits
- feat: add initial audio/TTS pipeline support for vLLM-omni backend (#7495) · cbbde3d0
  Zhongdongming Dai authored Mar 31, 2026
```
Signed-off-by: Zhongdongming Dai <zhongdongmin@nvidia.com>
```
  cbbde3d0
- fix: align video diffusion pipeline with TRT-LLM 1.3.0rc9 API changes (#7529) · 10fb23d2
  Indrajit Bhosale authored Mar 31, 2026
```
Signed-off-by: Indrajit Bhosale <iamindrajitb@gmail.com>
```
  10fb23d2
20 Mar, 2026 2 commits
- feat: GPU VRAM profiler + profiled test markers for 1 GPU SGLang (part 3) (#7508) · 6ed8ba0a
  Keiven C authored Mar 20, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  6ed8ba0a
- feat: [vLLM] add remote embedding loader to prefill/decode worker. Port E/P/D usage (#7507) · 4d302ab6
  GuanLuo authored Mar 19, 2026
```
Signed-off-by: Guan Luo <41310872+GuanLuo@users.noreply.github.com>
Signed-off-by: GuanLuo <41310872+GuanLuo@users.noreply.github.com>
```
  4d302ab6
19 Mar, 2026 4 commits
- fix: unskip multimodal_disagg_qwen EPD test (#7523) · fa92958a
  KrishnanPrash authored Mar 19, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  fa92958a
- fix: increase `/dev/shm` for CI pytest containers to fix NIXL UCX segfault (#7515) · 4fc2b5e8
  KrishnanPrash authored Mar 19, 2026
  
  4fc2b5e8
- fix: restore serve test timeouts for slower CI GPUs (#7510) · 934b0669
  Keiven C authored Mar 19, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
Co-authored-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  934b0669
- feat: add NixlEmbeddingSender / NixlEmbeddingReceiver in sglang backend (#7153) · 70892fc1
  Wang, Yi authored Mar 19, 2026
```
Signed-off-by: Wang, Yi <yi.a.wang@intel.com>
Co-authored-by: KrishnanPrash <140860868+KrishnanPrash@users.noreply.github.com>
```
  70892fc1
18 Mar, 2026 2 commits

feat: GPU VRAM profiler via memory fraction injection + profiled test markers... · 0b20745e

Keiven C authored Mar 18, 2026


feat: GPU VRAM profiler via memory fraction injection + profiled test markers (part 2 - vLLM only) (#6719)
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>

0b20745e

chore: add vllm omni e2e tests (#7454) · f890199f
Ayush Agarwal authored Mar 18, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
f890199f

17 Mar, 2026 3 commits
- ci: add E/P/D multimodal CI coverage for SGLang (#7444) · d92d5c77
  KrishnanPrash authored Mar 17, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  d92d5c77
- ci(trtllm): fix aggregated_multimodal_router test (#7460) · 22fb3398
  KrishnanPrash authored Mar 17, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  22fb3398
- chore: unskip several multi-gpu tests and enable sglang multi-gpu tests (#7443) · 9681225a
  Dmitry Tokarev authored Mar 17, 2026
```
Signed-off-by: Dmitry Tokarev <dtokarev@nvidia.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  9681225a
15 Mar, 2026 1 commit
- ci: enable multi-GPU tests in PR pipeline (#7355) · 4751f769
  Anant Sharma authored Mar 15, 2026
```
Signed-off-by: Anant Sharma <anants@nvidia.com>
```
  4751f769
13 Mar, 2026 1 commit
- ci: fit sglang multimodal EPD test on 1 GPU (#7046) · be2f1dc1
  KrishnanPrash authored Mar 13, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  be2f1dc1
12 Mar, 2026 1 commit
- feat(kv-router): add worker_type to router selection log (#7258) · dcbccbcd
  Yan Ru Pei authored Mar 12, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  dcbccbcd
11 Mar, 2026 3 commits
- test: pytests able to run locally now (#7219) · 4bd6299b
  Graham King authored Mar 11, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  4bd6299b
- fix: serialize disagg first_gen_log_probs int keys for Rust transport (#7145) · e930526b
  Yuewei Na authored Mar 10, 2026
```
Signed-off-by: Yuewei Na <nv-yna@users.noreply.github.com>
Co-authored-by: Yuewei Na <nv-yna@users.noreply.github.com>
```
  e930526b
- chore: skip lmcache tests on cuda13 env (#7196) · 9eec94b2
  Dmitry Tokarev authored Mar 10, 2026
```
Signed-off-by: Dmitry Tokarev <dtokarev@nvidia.com>
```
  9eec94b2
07 Mar, 2026 1 commit
- chore: bump trtllm to 1.3.0rc5.post1 (#7040) · 25393fb3
  Harrison Saturley-Hall authored Mar 06, 2026
```
Signed-off-by: Harrison King Saturley-Hall <hsaturleyhal@nvidia.com>
```
  25393fb3
03 Mar, 2026 3 commits
- ci: skip completions_only test (#6811) · 51e42e39
  KrishnanPrash authored Mar 03, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  51e42e39
- fix: pass through extra args in trtllm agg.sh launch script (#6787) · 146eb3b4
  KrishnanPrash authored Mar 02, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  146eb3b4
- fix: Fix E + PD Multimodal Flow in trtllm (#6726) · 47ed1227
  Indrajit Bhosale authored Mar 02, 2026
```
Signed-off-by: Indrajit Bhosale <iamindrajitb@gmail.com>
```
  47ed1227
01 Mar, 2026 1 commit
- chore: fix post merge (#6704) · 1412e44b
  Dmitry Tokarev authored Mar 01, 2026
```
Signed-off-by: Dmitry Tokarev <dtokarev@nvidia.com>
```
  1412e44b
28 Feb, 2026 1 commit
- feat(kv-router): add ActiveSequences benchmark and extract common bench utils (#6633) · e5850e23
  Yan Ru Pei authored Feb 27, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  e5850e23
27 Feb, 2026 2 commits
- fix: Phase out llava and make EPD single GPU (#6674) · 4ffa1082
  Indrajit Bhosale authored Feb 27, 2026
```
Signed-off-by: Indrajit Bhosale <iamindrajitb@gmail.com>
```
  4ffa1082
- chore: migrate vllm e/p/d test from gpu_2 -> gpu_1 (#6638) · b1930a61
  KrishnanPrash authored Feb 27, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  b1930a61
24 Feb, 2026 1 commit

fix: restore E/P/D multimodal disagg serving and add Qwen3-VL-30B-A3B support (#6533) · 8e236375

KrishnanPrash authored Feb 24, 2026


Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
Signed-off-by: KrishnanPrash <140860868+KrishnanPrash@users.noreply.github.com>

8e236375