Commits

Branch selector

 master

User selector

All users

All time

Commit history

Commits on Oct 9, 2025

• webui: updated the chat service to only include max_tokens in the req… (#16489)

  Show description for 1faa13a
  ServeurpersoCom

  authored
  1faa13a

  Copy full SHA for 1faa13a

• cpu : optimize the ggml NORM operation (#15953)

  Show description for 1deee0f
  

  3 peopleauthored
  1deee0f

  Copy full SHA for 1deee0f

• server : host-memory prompt caching (#16391)

  Show description for d00cbea
  

  ggerganov

  and

  ngxson

  authored
  d00cbea

  Copy full SHA for d00cbea

• No markdown in cot (#16483)

  Show description for 8328fd4
  ServeurpersoCom

  authored
  8328fd4

  Copy full SHA for 8328fd4

• model-conversion : add support for SentenceTransformers (#16387)

  Show description for 56b4795
  danbev

  authored
  56b4795

  Copy full SHA for 56b4795

• ci: add ARM64 Kleidiai build and test support (#16462)

  sudhiarm

  authored
  2c0d875

  Copy full SHA for 2c0d875

• CANN: Improve ACL graph matching (#16166)

  Show description for aa4711d
  noemotiovon

  authored
  aa4711d

  Copy full SHA for aa4711d

• kleidiai: kernel interface refactoring (#16460)

  chaxu01

  authored
  d80d6d2

  Copy full SHA for d80d6d2

• [SYCL] refactor soft_max, add soft_max_back (#16472)

  Show description for b260213
  

  NeoZhangJianyu

  and

  arthw

  authored
  b260213

  Copy full SHA for b260213

• model: EmbeddingGemma Adding Support for SentenceTransformers Dense Modules (#16367)

  Show description for e08db42
  

  3 peopleauthored
  e08db42

  Copy full SHA for e08db42

Commits on Oct 8, 2025

• refactor: centralize CoT parsing in backend for streaming mode (#16394)

  Show description for 12bbc3f
  

  3 peopleauthored
  12bbc3f

  Copy full SHA for 12bbc3f

• Disable CUDA host buffers on integrated GPUs (#16308)

  ai-fonsi

  authored
  9d08828

  Copy full SHA for 9d08828

• server : fix cancel pending task (#16467)

  Show description for d2ee056
  

  issixx

  and

  DevAI

  authored
  d2ee056

  Copy full SHA for d2ee056

• metal : mark FA blocks (#16372)

  Show description for b2c08c9
  ggerganov

  authored
  b2c08c9

  Copy full SHA for b2c08c9

• server : improve context checkpoint logic (#16440)

  ggerganov

  authored
  7fdd16b

  Copy full SHA for 7fdd16b

Commits on Oct 7, 2025

• ggml webgpu: profiling, CI updates, reworking of command submission (#16452)

  Show description for 74b8fc1
  reeselevine

  authored
  74b8fc1

  Copy full SHA for 74b8fc1

• llama : support LiquidAI LFM2-MoE hybrid model (#16464)

  Show description for aeaf8a3
  tdakhran

  authored
  aeaf8a3

  Copy full SHA for aeaf8a3

• server : add `/v1/health` endpoint (#16461)

  Show description for df1b612
  ggerganov

  authored
  df1b612

  Copy full SHA for df1b612

• webui : added download action (#13552) (#16282)

  Show description for 4e0388a
  

  srogmann

  and

  allozaur

  authored
  4e0388a

  Copy full SHA for 4e0388a

• presets : fix pooling param for embedding models (#16455)

  ggerganov

  authored
  ef4c5b8

  Copy full SHA for ef4c5b8

• rpc : update documentation (#16441)

  Show description for c61ae20
  

  rgerganov

  and

  slaren

  authored
  c61ae20

  Copy full SHA for c61ae20

• memory : use sequential equal splits for recurrent modules (#16442)

  ggerganov

  authored
  0123ff3

  Copy full SHA for 0123ff3

• metal : add support for non-padded FA KV (#16148)

  Show description for 0a319bb
  ggerganov

  authored
  0a319bb

  Copy full SHA for 0a319bb

• tests : add -INF blocks to the KQ mask in the FA tests (#16380)

  Show description for 1d6092f
  

  ggerganov

  and

  jeffbolznv

  authored
  1d6092f

  Copy full SHA for 1d6092f

• metal : various optimizations + refactoring (#16446)

  Show description for 8ae32dc
  ggerganov

  authored
  8ae32dc

  Copy full SHA for 8ae32dc

Commits on Oct 6, 2025

• llama : add --no-host to disable host buffers (#16310)

  Show description for 3df2244
  

  Gadflyii

  and

  slaren

  authored
  3df2244

  Copy full SHA for 3df2244

• chat : Granite Docling stopping (#16438)

  Show description for c08002a
  gabe-l-hart

  authored
  c08002a

  Copy full SHA for c08002a

• ci : refactor sdk caching to minimize storage (#16414)

  Show description for 3a002af
  CISC

  authored
  3a002af

  Copy full SHA for 3a002af

• ggml : fix unaligned access in AMX code (#16315)

  ggerganov

  authored
  a23b9bd

  Copy full SHA for a23b9bd

• ci : remove missing reranker model files (#16444)

  Show description for 04e632a
  danbev

  authored
  04e632a

  Copy full SHA for 04e632a

• ggml-cpu : fix leftover handling in ggml_vec_scale_f32 for SVE (#16443)

  Show description for a80ff18
  danbev

  authored
  a80ff18

  Copy full SHA for a80ff18

• nix : removed metal for nix (#16118)

  yuannan

  authored
  1d49ca3

  Copy full SHA for 1d49ca3

• server: update readme to mention n_past_max metric (#16436)

  Show description for c5fef0f
  okuvshynov

  authored
  c5fef0f

  Copy full SHA for c5fef0f

Commits on Oct 5, 2025

• model : Granite docling + Idefics3 preprocessing (SmolVLM) (#16206)

  Show description for ca71fb9
  

  gabe-l-hart

  and

  ngxson

  authored
  ca71fb9

  Copy full SHA for ca71fb9

• ggml webgpu: actually add softmax, fix rms_norm offset (#16400)

  Show description for 3526657
  reeselevine

  authored
  3526657

  Copy full SHA for 3526657