Commits

Breadcrumbs

History for

onmaster

User selector

All users

Datepicker

All time

Commit history

Commits on Oct 6, 2026

• ggml: refactor selective expert copying to user code (#29943)

  Show description for 6753a03
  

  am17an

  and

  ggerganov

  authored
  6753a03View commit details

  Copy full SHA for 6753a03

  View code at this pointBrowse repository at this point

• llama: fix k-pool scatter data race on shared sequences (#29994)

  Show description for 43fe9c6
  ServeurpersoCom

  authored
  43fe9c6View commit details

  Copy full SHA for 43fe9c6

  View code at this pointBrowse repository at this point

Commits on Oct 5, 2026

• server: support vision input for Clef (#29969)

  Show description for 9871df5
  ngxson

  authored
  9871df5View commit details

  Copy full SHA for 9871df5

  View code at this pointBrowse repository at this point

• llama : fix unexpected graph reallocation in the k-pool models (#29958)

  Show description for c173a53
  

  ggerganov

  and

  ServeurpersoCom

  authored
  c173a53View commit details

  Copy full SHA for c173a53

  View code at this pointBrowse repository at this point

• kv-cache: fix restoring mismatched KV cache rotation by saving exact rotation metadata (#28498)

  Show description for 2107910
  

  eapache

  and

  ggerganov

  authored
  2107910View commit details

  Copy full SHA for 2107910

  View code at this pointBrowse repository at this point

Commits on Oct 4, 2026

• llama: support both embd + raw tokens in batch (#29622)

  Show description for 0bb496d
  ngxson

  authored
  0bb496dView commit details

  Copy full SHA for 0bb496d

  View code at this pointBrowse repository at this point

Commits on Oct 3, 2026

• graph: gather the recurrent states once so the reserve covers every split (#29856)

  Show description for 436f6f8
  ServeurpersoCom

  authored
  436f6f8View commit details

  Copy full SHA for 436f6f8

  View code at this pointBrowse repository at this point

• qwen4exp : halve the indexer score memory (#29825)

  Show description for 889edf4
  

  ServeurpersoCom

  and

  0cc4m

  authored
  889edf4View commit details

  Copy full SHA for 889edf4

  View code at this pointBrowse repository at this point

• model: add support for clef decision model (text-only) (#29831)

  Show description for 99b9548
  

  ngxson

  and

  CISC

  authored
  99b9548View commit details

  Copy full SHA for 99b9548

  View code at this pointBrowse repository at this point

Commits on Oct 2, 2026

• llama, server: add /v1/systemone API (models: laya, julia-1, lev, openjev, kev) (#29818)

  Show description for a4cb4c6
  ngxson

  authored
  a4cb4c6View commit details

  Copy full SHA for a4cb4c6

  View code at this pointBrowse repository at this point

• llama : silence unused-result warnings (#29839)

  Show description for 8d81559
  angt

  authored
  8d81559View commit details

  Copy full SHA for 8d81559

  View code at this pointBrowse repository at this point

• llama : use GGML_ABORT instead of throw (#29840)

  Show description for 6805ae3
  angt

  authored
  6805ae3View commit details

  Copy full SHA for 6805ae3

  View code at this pointBrowse repository at this point

• qwen4exp : optimize mask constructions (#29824)

  Show description for 4e2713c
  ggerganov

  authored
  4e2713cView commit details

  Copy full SHA for 4e2713c

  View code at this pointBrowse repository at this point

• qwen4exp: fix tests (#29819)

  am17an

  authored
  159c651View commit details

  Copy full SHA for 159c651

  View code at this pointBrowse repository at this point

Commits on Oct 1, 2026

• llama: refer to segment documentation [no ci] (#29074)

  JohannesGaessler

  authored
  78e2964View commit details

  Copy full SHA for 78e2964

  View code at this pointBrowse repository at this point

• llama : clamp kpool re-pool bound to existing pools (#29805)

  Show description for 81e39ad
  ggerganov

  authored
  81e39adView commit details

  Copy full SHA for 81e39ad

  View code at this pointBrowse repository at this point

• llama : fix invalid assert in recurrent memory (#29799)

  ggerganov

  authored
  869034bView commit details

  Copy full SHA for 869034b

  View code at this pointBrowse repository at this point

• CUDA: Handle compute type for NVFP4 on cublass path (#29173)

  Show description for b56f34a
  

  ynankani

  and

  JohannesGaessler

  authored
  b56f34aView commit details

  Copy full SHA for b56f34a

  View code at this pointBrowse repository at this point

• Qwen4Exp: add MTP (#29761)

  Show description for c061df1
  

  am17an

  and

  ggerganov

  authored
  c061df1View commit details

  Copy full SHA for c061df1

  View code at this pointBrowse repository at this point

• llama: fix qwen4exp (#29751)

  Show description for 66e0c17
  am17an

  authored
  66e0c17View commit details

  Copy full SHA for 66e0c17

  View code at this pointBrowse repository at this point

• llama-mmap : avoid a second full-size copy of each tensor with direct-io (#29749)

  Show description for 32dd62e
  

  praneshgo

  and

  Pranesh Gonegandla

  authored
  32dd62eView commit details

  Copy full SHA for 32dd62e

  View code at this pointBrowse repository at this point

• vocab : honor BOS/EOS settings for PLaMo-2 and PLaMo-3 (#29734)

  Show description for db33d3c
  

  tokinasin

  and

  CISC

  authored
  db33d3cView commit details

  Copy full SHA for db33d3c

  View code at this pointBrowse repository at this point

• model : re-enable -sm tensor for qwen4exp (#28569)

  Show description for 10f340d
  kh0pper

  authored
  10f340dView commit details

  Copy full SHA for 10f340d

  View code at this pointBrowse repository at this point

Commits on Sep 30, 2026

• llama : preserve original batch order for speculative decoding layer inputs (#29019)

  Show description for 4453b53
  hthadicherla

  authored
  4453b53View commit details

  Copy full SHA for 4453b53

  View code at this pointBrowse repository at this point

• llama: properly handle KV on training (#28520)

  Show description for 81ff93e
  ngxson

  authored
  81ff93eView commit details

  Copy full SHA for 81ff93e

  View code at this pointBrowse repository at this point

• glm5-next: give dead indexer slots unique scatter rows (#29745)

  Show description for 05af0d2
  ServeurpersoCom

  authored
  05af0d2View commit details

  Copy full SHA for 05af0d2

  View code at this pointBrowse repository at this point

• mimo : support dflash (convert + feature extraction) (#29650)

  Show description for 22bdcc4
  

  ggerganov

  and

  CISC

  authored
  22bdcc4View commit details

  Copy full SHA for 22bdcc4

  View code at this pointBrowse repository at this point

• llama: llama_prefetch_rows (#29599)

  Show description for 185103d
  

  3 peopleauthored
  185103dView commit details

  Copy full SHA for 185103d

  View code at this pointBrowse repository at this point

• add GLM-5.3-Flash (GLM5-Next) support (#27773)

  Show description for 649dcb1
  

  4 peopleauthored
  649dcb1View commit details

  Copy full SHA for 649dcb1

  View code at this pointBrowse repository at this point

Commits on Sep 29, 2026

• model : support classifier_pooling for rerankers (#29627)

  Show description for 48de2a1
  boshjerns

  authored
  48de2a1View commit details

  Copy full SHA for 48de2a1

  View code at this pointBrowse repository at this point

• vocab : keep </s> NORMAL in PLaMo-2 and PLaMo-3 (#29580)

  Show description for a3f84fa
  tokinasin

  authored
  a3f84faView commit details

  Copy full SHA for a3f84fa

  View code at this pointBrowse repository at this point

• ggml : require input tensors to be GGML_OP_NONE (#29647)

  ggerganov

  authored
  b5cf8ceView commit details

  Copy full SHA for b5cf8ce

  View code at this pointBrowse repository at this point

Commits on Sep 28, 2026

• models: pad on the left with ggml_pad_ext (#29567)

  Show description for 57b557c
  ServeurpersoCom

  authored
  57b557cView commit details

  Copy full SHA for 57b557c

  View code at this pointBrowse repository at this point

• context : do not re-reserve the scheduler when toggling causal_attn (#28751)

  Show description for ed7ac35
  

  sihanyu03

  and

  ggerganov

  authored
  ed7ac35View commit details

  Copy full SHA for ed7ac35

  View code at this pointBrowse repository at this point

• tests : fix ggml init (#29554)

  Show description for 81ef10e
  ggerganov

  authored
  81ef10eView commit details

  Copy full SHA for 81ef10e

  View code at this pointBrowse repository at this point