Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on Nov 15, 2024

• [Bugfix] Qwen-vl output is inconsistent in speculative decoding (#10350)

  skylee-01

  authored
  2ec8827

  Copy full SHA for 2ec8827

• [Model] Support Qwen2 embeddings and use tags to select model tests (#10184)

  DarkLight1337

  authored
  b40cf64

  Copy full SHA for b40cf64

• [Misc] Change RedundantReshapesPass and FusionPass logging from info to debug (#10308)

  Show description for 2885ba0
  tlrmchlsmth

  authored
  2885ba0

  Copy full SHA for 2885ba0

• [bugfix] Fix static asymmetric quantization case (#10334)

  Show description for bf2ddc6
  

  ProExpertProg

  and

  danieldk

  authored
  bf2ddc6

  Copy full SHA for bf2ddc6

• [Bugfix] Fix unable to load some models (#10312)

  Show description for 972112d
  DarkLight1337

  authored
  972112d

  Copy full SHA for 972112d

• [Tool parsing] Improve / correct mistral tool parsing (#10333)

  patrickvonplaten

  authored
  11cd1ae

  Copy full SHA for 11cd1ae

• [Bugfix] use AF_INET6 for OpenAI Compatible Server with ipv6 (#9583)

  Show description for 554af92
  jxpxxzj

  authored
  554af92

  Copy full SHA for 554af92

• [Perf] Reduce peak memory usage of llama (#10339)

  Show description for b2e0ad3
  andoorve

  authored
  b2e0ad3

  Copy full SHA for b2e0ad3

Commits on Nov 14, 2024

• Support Roberta embedding models (#9387)

  Show description for 4a18fd1
  

  maxdebayser

  and

  flaviabeo

  authored
  4a18fd1

  Copy full SHA for 4a18fd1

• [Model] Add BNB quantization support for Idefics3 (#10310)

  Show description for 294bf46
  

  linitra24

  and

  jeejeelee

  authored
  294bf46

  Copy full SHA for 294bf46

• [BugFix]: properly deserialize `tool_calls` iterator before processing by mistral-common when MistralTokenizer is used (#9951)

  Show description for 52b48c1
  gcalmettes

  authored
  52b48c1

  Copy full SHA for 52b48c1

• [Frontend] Pythonic tool parser (#9859)

  Show description for f67ce05
  mdepinet

  authored
  f67ce05

  Copy full SHA for f67ce05

• [misc] error early for old-style class (#10304)

  Show description for 504ac53
  youkaichao

  authored
  504ac53

  Copy full SHA for 504ac53

• [Bugfix] Fix tensor parallel for qwen2 classification model (#10297)

  Show description for 15bb833
  Isotr0py

  authored
  15bb833

  Copy full SHA for 15bb833

Commits on Nov 13, 2024

• [Bugfix] bitsandbytes models fail to run pipeline parallel (#10200)

  Show description for ac49b59
  HoangCongDuc

  authored
  ac49b59

  Copy full SHA for ac49b59

• [1/N] Initial prototype for multi-modal processor (#10044)

  Show description for 0b8bb86
  DarkLight1337

  authored
  0b8bb86

  Copy full SHA for 0b8bb86

• [V1] Add missing tokenizer options for `Detokenizer` (#10288)

  Show description for bb7991a
  ywang96

  authored
  bb7991a

  Copy full SHA for bb7991a

• [Model][LoRA]LoRA support added for idefics3 (#10281)

  Show description for d909acf
  linitra24

  authored
  d909acf

  Copy full SHA for d909acf

• [Model] Adding Support for Qwen2VL as an Embedding Model. Using MrLight/dse-qwen2-2b-mrl-v1 (#9944)

  Show description for 1b886aa
  FurtherAI

  authored
  1b886aa

  Copy full SHA for 1b886aa

• [Model] Add support for Qwen2-VL video embeddings input & multiple image embeddings input with varied resolutions (#10221)

  Show description for 3945c82
  imkero

  authored
  3945c82

  Copy full SHA for 3945c82

• [Doc] Fix typo in arg_utils.py (#10264)

  Show description for 032fcf1
  xyang16

  authored
  032fcf1

  Copy full SHA for 032fcf1

• [V1] Support VLMs with fine-grained scheduling (#9871)

  Show description for bbd3e86
  

  WoosukKwon

  and

  ywang96

  authored
  bbd3e86

  Copy full SHA for bbd3e86

• [core][distributed] use tcp store directly (#10275)

  Show description for 0d4ea3f
  youkaichao

  authored
  0d4ea3f

  Copy full SHA for 0d4ea3f

• [V1] Fix CI tests on V1 engine (#10272)

  Show description for 112fa0b
  WoosukKwon

  authored
  112fa0b

  Copy full SHA for 112fa0b

Commits on Nov 12, 2024

• [V1] Enable Inductor when using piecewise CUDA graphs (#10268)

  Show description for 1f55e05
  WoosukKwon

  authored
  1f55e05

  Copy full SHA for 1f55e05

• [LoRA] Adds support for bias in LoRA (#5733)

  Show description for 8a06428
  

  followumesh

  and

  Umesh Deshpande

  authored
  8a06428

  Copy full SHA for 8a06428

• [Encoder Decoder] Update Mllama to run with both FlashAttention and XFormers (#9982)

  Show description for b41fb9d
  sroy745

  authored
  b41fb9d

  Copy full SHA for b41fb9d

• [V1] Use pickle for serializing EngineCoreRequest & Add multimodal inputs to EngineCoreRequest (#10245)

  Show description for 7c65527
  WoosukKwon

  authored
  7c65527

  Copy full SHA for 7c65527

• [Frontend] Add per-request number of cached token stats (#10174)

  zifeitong

  authored
  47db6ec

  Copy full SHA for 47db6ec

• [Bugfix] Fix QwenModel argument (#10262)

  Show description for 176fcb1
  DamonFool

  authored
  176fcb1

  Copy full SHA for 176fcb1

• [Misc]Fix Idefics3Model argument (#10255)

  Show description for a838ba7
  jeejeelee

  authored
  a838ba7

  Copy full SHA for a838ba7

• [BugFix] Do not raise a `ValueError` when `tool_choice` is set to the supported `none` option and `tools` are not defined. (#10000)

  Show description for 36c513a
  gcalmettes

  authored
  36c513a

  Copy full SHA for 36c513a

• [Misc][LoRA] Replace hardcoded cuda device with configurable argument (#10223)

  Show description for 7f5edb5
  jeejeelee

  authored
  7f5edb5

  Copy full SHA for 7f5edb5

• [1/N] torch.compile user interface design (#10237)

  Show description for eea55cc
  youkaichao

  authored
  eea55cc

  Copy full SHA for eea55cc

• [Doc] Update help text for `--distributed-executor-backend` (#10231)

  Show description for 9cdba96
  russellb

  authored
  9cdba96

  Copy full SHA for 9cdba96