Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit history

Commits on Jul 18, 2026

• model: rotate injected K/V cache for DFlash (#25823)

  Show description for 571d0d5
  

  ruixiang63

  and

  ggerganov

  authored
  571d0d5View commit details

  Copy full SHA for 571d0d5

  View code at this pointBrowse repository at this point

• llama-quant : exclude i32 ffn_gate_tid2eid routing table from quantization (#25787)

  Show description for 4937ca8
  devYRPauli

  authored
  4937ca8View commit details

  Copy full SHA for 4937ca8

  View code at this pointBrowse repository at this point

Commits on Jul 16, 2026

• DeepseekV4: Add fused hyper-connection ops (#25585)

  Show description for 0dc74e3
  am17an

  authored
  0dc74e3View commit details

  Copy full SHA for 0dc74e3

  View code at this pointBrowse repository at this point

• TP: fix Phi3, Bert, Plamo2/3, ChatGLM (#25536)

  JohannesGaessler

  authored
  2e1fd76View commit details

  Copy full SHA for 2e1fd76

  View code at this pointBrowse repository at this point

• quant : allow using manual tensor types with --pure (#25716)

  ggerganov

  authored
  56d6e9dView commit details

  Copy full SHA for 56d6e9d

  View code at this pointBrowse repository at this point

Commits on Jul 15, 2026

• DeepseekV4: reduce graph splits (#25702)

  am17an

  authored
  33a75f4View commit details

  Copy full SHA for 33a75f4

  View code at this pointBrowse repository at this point

Commits on Jul 14, 2026

• DeepseekV4: fix seq_rm (#25588)

  Show description for 7f575c3
  am17an

  authored
  7f575c3View commit details

  Copy full SHA for 7f575c3

  View code at this pointBrowse repository at this point

Commits on Jul 13, 2026

• model: add Hy3 (hy_v3) support with MTP speculative decoding (#25395)

  Show description for 2969d6d
  

  3 peopleauthored
  2969d6dView commit details

  Copy full SHA for 2969d6d

  View code at this pointBrowse repository at this point

• spec: add Minimax2 eagle3 support

  Show description for 259ae1d
  

  adrianisk

  and

  ggerganov

  authored
  259ae1dView commit details

  Copy full SHA for 259ae1d

  View code at this pointBrowse repository at this point

Commits on Jul 11, 2026

• Fix conditional to display 'LLAMA_SPLIT_MODE_TENSOR not implemented for architecture' message (#24926)

  kdkd

  authored
  e3546c7View commit details

  Copy full SHA for e3546c7

  View code at this pointBrowse repository at this point

• DeepseekV4: clear cache only for seq rather than full (#25521)

  am17an

  authored
  13f2b28View commit details

  Copy full SHA for 13f2b28

  View code at this pointBrowse repository at this point

• ggml : add GGML_OP_LIGHTNING_INDEXER that implements DeepSeek V3.2/V4 lightning indexer (#24231)

  Show description for 00f5442
  

  fairydreaming

  and

  sszymczy

  authored
  00f5442View commit details

  Copy full SHA for 00f5442

  View code at this pointBrowse repository at this point

Commits on Jul 10, 2026

• llama : make tensor-split regex patterns static (#24710)

  Show description for c749cb0
  eduardopessin

  authored
  c749cb0View commit details

  Copy full SHA for c749cb0

  View code at this pointBrowse repository at this point

• llama : make all KQ masks f16 if FA is used, remove zero attention bias, remove raw_k repeats in DeepSeek V4 (#25370)

  Show description for 2ed3c1a
  

  fairydreaming

  and

  sszymczy

  authored
  2ed3c1aView commit details

  Copy full SHA for 2ed3c1a

  View code at this pointBrowse repository at this point

Commits on Jul 8, 2026

• llama-batch: fix allowed decreasing pos in a seq (#25449)

  am17an

  authored
  1ee0939View commit details

  Copy full SHA for 1ee0939

  View code at this pointBrowse repository at this point

• llama: refactor fused ops (#24646)

  am17an

  authored
  90e0f5cView commit details

  Copy full SHA for 90e0f5c

  View code at this pointBrowse repository at this point

• llama-batch: add n_keep_tail in split_equal for recurrent models (#25278)

  am17an

  authored
  230ea9dView commit details

  Copy full SHA for 230ea9d

  View code at this pointBrowse repository at this point

• fix: OOB reads in UGM tokenizer (precompiled_charsmap handling) (#18750)

  Show description for 4a7ee31
  

  3 peopleauthored
  4a7ee31View commit details

  Copy full SHA for 4a7ee31

  View code at this pointBrowse repository at this point

Commits on Jul 7, 2026

• Add Q2_0 quantization: type definition and CPU backend (#24448)

  khosravipasha

  authored
  bec4772View commit details

  Copy full SHA for bec4772

  View code at this pointBrowse repository at this point

• llama: fix quantized kv-cache for dsv4 (#25202)

  am17an

  authored
  024c46aView commit details

  Copy full SHA for 024c46a

  View code at this pointBrowse repository at this point

Commits on Jul 6, 2026

• CUDA: remove -sm row, refactor cuBLAS (#24216)

  Show description for 74976e1
  JohannesGaessler

  authored
  74976e1View commit details

  Copy full SHA for 74976e1

  View code at this pointBrowse repository at this point

Commits on Jul 5, 2026

• Fix stale tensor-split params for draft models (#24814)

  Show description for 2da6686
  RedToasty

  authored
  2da6686View commit details

  Copy full SHA for 2da6686

  View code at this pointBrowse repository at this point

Commits on Jul 4, 2026

• llama : add guard for K/V rotation input when buffer is unallocated (#25215)

  Show description for a410713
  liminfei-amd

  authored
  a410713View commit details

  Copy full SHA for a410713

  View code at this pointBrowse repository at this point

Commits on Jul 2, 2026

• llama : add llama_model_ftype_name() (#25134)

  Show description for fdb1db8
  angt

  authored
  fdb1db8View commit details

  Copy full SHA for fdb1db8

  View code at this pointBrowse repository at this point

Commits on Jun 30, 2026

• model : register t_layer_inp for qwen3next (#25141)

  Show description for 4f31eed
  jschmied

  authored
  4f31eedView commit details

  Copy full SHA for 4f31eed

  View code at this pointBrowse repository at this point

Commits on Jun 29, 2026

• DeepSeek V4 (#24162)

  Show description for 8c146a8
  

  6 peopleauthored
  8c146a8View commit details

  Copy full SHA for 8c146a8

  View code at this pointBrowse repository at this point

Commits on Jun 28, 2026

• spec : add DFlash support (#22105)

  Show description for d1b3425
  

  ruixiang63

  and

  kashif

  authored
  d1b3425View commit details

  Copy full SHA for d1b3425

  View code at this pointBrowse repository at this point

• logs : reduce v2 (#25078)

  Show description for 27c8bb4
  ggerganov

  authored
  27c8bb4View commit details

  Copy full SHA for 27c8bb4

  View code at this pointBrowse repository at this point

Commits on Jun 26, 2026

• mamba2: remove hardcoded 2x expansion factor and invalid d_inner % d_state check (#23082)

  Show description for 960d628
  

  limloop

  and

  CISC

  authored
  960d628View commit details

  Copy full SHA for 960d628

  View code at this pointBrowse repository at this point

Commits on Jun 25, 2026

• model : Add label for LFM2.5-230M (#25008)

  tdakhran

  authored
  9d5d882View commit details

  Copy full SHA for 9d5d882

  View code at this pointBrowse repository at this point

• quant : fix quantizing moe with mtp (#24986)

  CISC

  authored
  b3ce5ceView commit details

  Copy full SHA for b3ce5ce

  View code at this pointBrowse repository at this point

Commits on Jun 24, 2026

• model : Add LFM2.5-ColBERT-350M and LFM2.5-Embedding-350M (#24913)

  Show description for 88636e1
  tdakhran

  authored
  88636e1View commit details

  Copy full SHA for 88636e1

  View code at this pointBrowse repository at this point

Commits on Jun 22, 2026

• sampling : remove unconditional softmax+sort in top-n-sigma sampler (#22645)

  TimNN

  authored
  37957e8View commit details

  Copy full SHA for 37957e8

  View code at this pointBrowse repository at this point

Commits on Jun 21, 2026

• spec : Support Step3.5/3.7 flash mtp3 (#24340)

  Show description for d789527
  

  forforever73

  and

  am17an

  authored
  d789527View commit details

  Copy full SHA for d789527

  View code at this pointBrowse repository at this point

Commits on Jun 20, 2026

• llama : use LLM_KV for quantization_version & file_type (#24802)

  Show description for 84de01a
  angt

  authored
  84de01aView commit details

  Copy full SHA for 84de01a

  View code at this pointBrowse repository at this point