Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 24, 2024

• fix: refactor adapter weight loading and mapping (huggingface#2193)

  Show description for 5d85a95
  drbh

  authored
  5d85a95

• Split up layers.marlin into several files (huggingface#2292)

  Show description for 93d2b9f
  danieldk

  authored
  93d2b9f

• fix of use of unquantized weights in cohere GQA loading, also enable … (huggingface#2291)

  Show description for 8642250
  sywangyi

  authored
  8642250

• fix crash in multi-modal (huggingface#2245)

  Show description for 5ad39dd
  sywangyi

  authored
  5ad39dd

Commits on Jul 23, 2024

• hotfix: update nccl

  OlivierDehaene

  committed
  a895029

• chore: update to torch 2.4 (huggingface#2259)

  Show description for e7e3aa6
  OlivierDehaene

  authored
  e7e3aa6

• hotfix: pin numpy (huggingface#2289)

  danieldk

  authored
  bc9593a

• Add support for Llama 3 rotary embeddings (huggingface#2286)

  Show description for 4ab4173
  danieldk

  authored
  4ab4173

• Preparing for release. (huggingface#2285)

  Show description for 5d121a9
  Narsil

  authored
  5d121a9

• [WIP] Add support for Mistral-Nemo by supporting head_dim through config (huggingface#2254)

  Show description for 3961e32
  

  shaltielshmid

  and

  Narsil

  authored
  3961e32

• Add support for repacking AWQ weights for GPTQ-Marlin (huggingface#2278)

  Show description for 9935720
  danieldk

  authored
  9935720

• fix(l4): fix fp8 logic on l4 (huggingface#2277)

  Show description for 5fca30e
  OlivierDehaene

  authored
  5fca30e

• Fixing mistral nemo. (huggingface#2276)

  Narsil

  authored
  abc3253

Commits on Jul 22, 2024

• use proper name for ci (huggingface#2274)

  XciD

  authored
  4700465

• Softcapping for gemma2. (huggingface#2273)

  Show description for 6aeb669
  Narsil

  authored
  6aeb669

• fix(server): fix fp8 weight loading (huggingface#2268)

  Show description for 4844ff7
  OlivierDehaene

  authored
  4844ff7

• fix(ci): test new instances (huggingface#2272)

  Show description for 6aebf44
  XciD

  authored
  6aebf44

• legacy warning on text_generation client (huggingface#2271)

  Show description for 07441f5
  ErikKaum

  authored
  07441f5

• Hotfix: fix of use of unquantized weights in Mixtral GQA loading (huggingface#2269)

  Show description for 4e42072
  icyxp

  authored
  4e42072

Commits on Jul 21, 2024

• fix(server): fix deepseekv2 loading (huggingface#2266)

  OlivierDehaene

  authored
  f3435ba

Commits on Jul 20, 2024

• feat(fp8): use fbgemm kernels and load fp8 weights directly (huggingface#2248)

  Show description for 53ec0b7
  OlivierDehaene

  authored
  53ec0b7

• Add FP8 release test (huggingface#2261)

  danieldk

  authored
  e5c1d6d

• re-push to internal registry (huggingface#2242)

  Show description for 11123a8
  XciD

  authored
  11123a8

Commits on Jul 19, 2024

• Add support for Deepseek V2 (huggingface#2224)

  Show description for e52be9b
  danieldk

  authored
  e52be9b

• fix: adjust default tool choice (huggingface#2244)

  Show description for 68a9685
  drbh

  authored
  68a9685

• add usage stats to toctree (huggingface#2260)

  Show description for 40f5dc3
  ErikKaum

  authored
  40f5dc3

• usage stats and crash reports (huggingface#2220)

  Show description for 4c19593
  

  3 peopleauthored
  4c19593

• Hotfix: pass through model revision in VlmCausalLM (huggingface#2258)

  danieldk

  authored
  3f37a66

• Hotfix: fix MPT after recent refactor (huggingface#2257)

  danieldk

  authored
  3b41e93

• Hotfix: various GPT-based model fixes (huggingface#2256)

  danieldk

  authored
  18db78f

• Hotfix: fix of use of unquantized weights in Gemma GQA loading (huggingface#2255)

  danieldk

  authored
  80adb5b

• Improve the handling of quantized weights (huggingface#2250)

  Show description for ba291da
  danieldk

  authored
  ba291da

Commits on Jul 18, 2024

• fix(server): fix cohere (huggingface#2249)

  OlivierDehaene

  authored
  1d1b1ef

Commits on Jul 16, 2024

• Remove stray quantize argument in get_weights_col_packed_qkv (huggingface#2237)

  Show description for da82c63
  danieldk

  authored
  da82c63

• server quantize: expose groupsize option (huggingface#2225)

  danieldk

  authored
  2cb1842

Pagination

PreviousNext