Commits · 3f674a49b5033a6ed778ab960e86e03cfa64aa1f · OpenDAS / vllm_cscc · GitLab

14 Aug, 2024 5 commits
- [VLM][Core] Support profiling with multiple multi-modal inputs per prompt (#7126) · 3f674a49
  Cyrus Leung authored Aug 15, 2024
  
  3f674a49
- [Misc] Deprecation Warning when setting --engine-use-ray (#7424) · 70b746ef
  Wallas Henrique authored Aug 14, 2024
```
Signed-off-by: Wallas Santos <wallashss@ibm.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: Nick Hill <nickhill@us.ibm.com>
Co-authored-by: youkaichao <youkaichao@126.com>
```
  70b746ef
- [Bugfix][Frontend] Disable embedding API for chat models (#7504) · 67d115db
  jack authored Aug 15, 2024
```
Co-authored-by: jack <jack@alex>
```
  67d115db
- Fix empty output when temp is too low (#2937) · c134a464
  Chang Su authored Aug 13, 2024
```
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
```
  c134a464
- [Bugfix][CI] Import ray under guard (#7486) · 59edd0f1
  Woosuk Kwon authored Aug 13, 2024
  
  59edd0f1
13 Aug, 2024 14 commits
- [TPU] Support multi-host inference (#7457) · a08df832
  Woosuk Kwon authored Aug 13, 2024
  
  a08df832
- [misc][plugin] add plugin system implementation (#7426) · 16422ea7
  youkaichao authored Aug 13, 2024
  
  16422ea7
- [Misc] `compressed-tensors` code reuse (#7277) · 373538f9
  Kyle Sayers authored Aug 13, 2024
  
  373538f9
- [frontend] spawn engine process from api server process (#7484) · 33e5d7e6
  youkaichao authored Aug 13, 2024
  
  33e5d7e6
- [Misc] Update `awq` and `awq_marlin` to use `vLLMParameters` (#7422) · b1e5afc3
  Dipika Sikka authored Aug 13, 2024
  
  b1e5afc3
- [Misc] Update Fused MoE weight loading (#7334) · d3bdfd3a
  Dipika Sikka authored Aug 13, 2024
  
  d3bdfd3a
- [Misc] Update `gptq_marlin` to use new vLLMParameters (#7281) · fb377d7e
  Dipika Sikka authored Aug 13, 2024
  
  fb377d7e
- [Frontend][Core] Add plumbing to support audio language models (#7446) · 00c3d68e
  Peter Salas authored Aug 13, 2024
  
  00c3d68e
- [TPU] Suppress import custom_ops warning (#7458) · d6e634f3
  Woosuk Kwon authored Aug 13, 2024
  
  d6e634f3
- [hardware] unify usage of is_tpu to current_platform.is_tpu() (#7102) · 4d2dc507
  youkaichao authored Aug 13, 2024
  
  4d2dc507
- [Bugfix] Fix weight loading for Chameleon when TP>1 (#7410) · 7025b11d
  Cyrus Leung authored Aug 13, 2024
  
  7025b11d
- [Misc] improve logits processors logging message (#7435) · 97a6be95
  Andrew Wang authored Aug 12, 2024
  
  97a6be95
- [mypy] Misc. typing improvements (#7417) · 9ba85bc1
  Cyrus Leung authored Aug 13, 2024
  
  9ba85bc1
- [Core] Shut down aDAG workers with clean async llm engine exit (#7224) · 198d6a28
  Rui Qiao authored Aug 12, 2024
```
Signed-off-by: Rui Qiao <ruisearch42@gmail.com>
```
  198d6a28
12 Aug, 2024 8 commits
- [Bugfix] Handle PackageNotFoundError when checking for xpu version (#7398) · 91294d56
  sasha0552 authored Aug 12, 2024
  
  91294d56
- [Core/Bugfix] Add FP8 K/V Scale and dtype conversion for prefix/prefill Triton Kernel (#7208) · a046f863
  jon-chuang authored Aug 12, 2024
```
Co-authored-by: Cody Yu <hao.yu.cody@gmail.com>
```
  a046f863
- [Core] Consolidate `GB` constant and enable float GB arguments (#7416) · 4ddc4743
  Cyrus Leung authored Aug 13, 2024
  
  4ddc4743
- [Bugfix] Fix logit soft cap in flash-attn backend (#7425) · cfba4def
  Woosuk Kwon authored Aug 12, 2024
  
  cfba4def
- [Frontend] Disallow passing `model` as both argument and option (#7347) · 24154f86
  Cyrus Leung authored Aug 12, 2024
  
  24154f86
- [Core][VLM] Support image embeddings as input (#6613) · e6e42e4b
  Roger Wang authored Aug 12, 2024
  
  e6e42e4b
- [Kernel] Flashinfer correctness fix for v0.1.3 (#7319) · ec2affa8
  Lily Liu authored Aug 12, 2024
  
  ec2affa8
- [CI/Build] Minor refactoring for vLLM assets (#7407) · 86ab567b
  Roger Wang authored Aug 11, 2024
  
  86ab567b
11 Aug, 2024 2 commits
- [core] [2/N] refactor worker_base input preparation for multi-step (#7387) · c08e2b30
  William Lin authored Aug 11, 2024
  
  c08e2b30
- [TPU] Use mark_dynamic to reduce compilation time (#7340) · 90bab18f
  Woosuk Kwon authored Aug 10, 2024
  
  90bab18f
10 Aug, 2024 1 commit
- [Bugfix] Fix phi3v batch inference when images have different aspect ratio (#7392) · 4c5d8e8e
  Isotr0py authored Aug 11, 2024
  
  4c5d8e8e
09 Aug, 2024 10 commits
- [Core] Fix edge case in chunked prefill + block manager v2 (#7380) · baa24025
  Cade Daniel authored Aug 09, 2024
  
  baa24025
- [Misc] Add numpy implementation of `compute_slot_mapping` (#7377) · 999ef0b9
  Antoni Baum authored Aug 09, 2024
  
  999ef0b9
- [Bugfix] Fix `PerTensorScaleParameter` weight loading for fused models (#7376) · 5c6c54d6
  Dipika Sikka authored Aug 09, 2024
  
  5c6c54d6
- [Core] Add span metrics for model_forward, scheduler and sampler time (#7089) · 933790c2
  Mahesh Keralapura authored Aug 09, 2024
  
  933790c2
- [Frontend] Support embeddings in the run_batch API (#7132) · 249b8822
  Pooya Davoodi authored Aug 09, 2024
```
Co-authored-by: Simon Mo <simon.mo@hey.com>
```
  249b8822
- [Bugfix] Fix reinit procedure in ModelInputForGPUBuilder (#7360) · 74af2bbd
  Alexander Matveev authored Aug 09, 2024
  
  74af2bbd
- [Performance] e2e overheads reduction: Small followup diff (#7364) · fc7b8d1e
  Alexander Matveev authored Aug 09, 2024
  
  fc7b8d1e
- [Model][Jamba] Mamba cache single buffer (#6739) · 07ab1607
  Mor Zusman authored Aug 09, 2024
```
Co-authored-by: Mor Zusman <morz@ai21.com>
```
  07ab1607
- [Core] Streamline stream termination in `AsyncLLMEngine` (#7336) · b4e9528f
  Nick Hill authored Aug 09, 2024
  
  b4e9528f
- [Speculative decoding] [Multi-Step] decouple should_modify_greedy_probs_inplace (#6971) · 57b7be0e
  William Lin authored Aug 08, 2024
  
  57b7be0e