Commits

Breadcrumbs

History for

onmaster

User selector

All users

Datepicker

All time

Commit history

Commits on Sep 8, 2026

• ggml : update ggml_prec specification (#26675)

  Show description for 5a6caa0
  ggerganov

  authored
  5a6caa0View commit details

  Copy full SHA for 5a6caa0

  View code at this pointBrowse repository at this point

Commits on Sep 6, 2026

• convert : add `--fuse-qkv` flag to fuse Q/K/V into QKV during HF-to-GGUF conversion (#22780)

  JoursBleu

  authored
  465e49bView commit details

  Copy full SHA for 465e49b

  View code at this pointBrowse repository at this point

Commits on Sep 4, 2026

• model : add Tencent Hy 4 (hy_v4) preview architecture support (#28127)

  Show description for 49c0dc8
  

  4 peopleauthored
  49c0dc8View commit details

  Copy full SHA for 49c0dc8

  View code at this pointBrowse repository at this point

Commits on Sep 3, 2026

• model: add NVIDIA Nemotron-3-Puzzle-75B-A9B (NemotronHPuzzle) support (#25444)

  Show description for c61b98b
  YanissAmz

  authored
  c61b98bView commit details

  Copy full SHA for c61b98b

  View code at this pointBrowse repository at this point

Commits on Sep 2, 2026

• CUDA + ggml: add sparse-fa for DSV4/GLM (#27970)

  am17an

  authored
  8e93a97View commit details

  Copy full SHA for 8e93a97

  View code at this pointBrowse repository at this point

Commits on Aug 30, 2026

• ggml: add SWIGLU_CLAMP (#27930)

  Show description for 0190529
  am17an

  authored
  0190529View commit details

  Copy full SHA for 0190529

  View code at this pointBrowse repository at this point

Commits on Aug 21, 2026

• model: add dots3-note (#27060)

  Show description for 5a32f7b
  ngxson

  authored
  5a32f7bView commit details

  Copy full SHA for 5a32f7b

  View code at this pointBrowse repository at this point

Commits on Aug 20, 2026

• graph : create V as a view of K in the k_iswa build_attn (#27392)

  Show description for 929d47a
  ggerganov

  authored
  929d47aView commit details

  Copy full SHA for 929d47a

  View code at this pointBrowse repository at this point

Commits on Aug 15, 2026

• model: add Kimi-K3 text model (#26185)

  Show description for ad1de39
  

  7 peopleauthored
  ad1de39View commit details

  Copy full SHA for ad1de39

  View code at this pointBrowse repository at this point

Commits on Aug 10, 2026

• llama : support multi-output backend sampling (#25532)

  Show description for dd1ea52
  

  gaugarg-nv

  and

  ggerganov

  authored
  dd1ea52View commit details

  Copy full SHA for dd1ea52

  View code at this pointBrowse repository at this point

Commits on Aug 4, 2026

• llama : move n_vocab from llama_sampler_data to penalty_sampler (#26520)

  Show description for 935cad6
  ggerganov

  authored
  935cad6View commit details

  Copy full SHA for 935cad6

  View code at this pointBrowse repository at this point

Commits on Aug 3, 2026

• graph : fix unused input tensors in minimax m3 graph (#26519)

  ggerganov

  authored
  c50b34aView commit details

  Copy full SHA for c50b34a

  View code at this pointBrowse repository at this point

• model: M3: Move MSA into a new memory implementation (#26338)

  Show description for 67d5978
  

  timkhronos

  and

  ggerganov

  authored
  67d5978View commit details

  Copy full SHA for 67d5978

  View code at this pointBrowse repository at this point

• CUDA: Add backend sampler for penalties sampler (#25262)

  Show description for 96278e3
  kmorennv

  authored
  96278e3View commit details

  Copy full SHA for 96278e3

  View code at this pointBrowse repository at this point

Commits on Aug 2, 2026

• DeepseekV4 MTP + DSpark (#25784)

  am17an

  authored
  596a579View commit details

  Copy full SHA for 596a579

  View code at this pointBrowse repository at this point

Commits on Jul 26, 2026

• model: Add MiniMax-M3 (MSA: MiniMax Sparse Attention) support (#24908)

  Show description for b1d4c65
  

  4 peopleauthored
  b1d4c65View commit details

  Copy full SHA for b1d4c65

  View code at this pointBrowse repository at this point

Commits on Jul 11, 2026

• 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 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: refactor fused ops (#24646)

  am17an

  authored
  90e0f5cView commit details

  Copy full SHA for 90e0f5c

  View code at this pointBrowse repository at this point

Commits on Jul 7, 2026

• 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 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 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

Commits on Jun 16, 2026

• Fix and restrict NVFP4 edge-cases in llama-graph (#24331)

  Show description for 02810c7
  ORippler

  authored
  02810c7View commit details

  Copy full SHA for 02810c7

  View code at this pointBrowse repository at this point

Commits on Jun 12, 2026

• spec: add EAGLE3 speculative decoding support (#18039)

  Show description for 88a3927
  

  5 peopleauthored
  88a3927View commit details

  Copy full SHA for 88a3927

  View code at this pointBrowse repository at this point

Commits on Jun 9, 2026

• graph: Fix granite speech model inference by applying embedding scale when deepstack is not used (#24357)

  Show description for d73cd07
  

  arnu515

  and

  ngxson

  authored
  d73cd07View commit details

  Copy full SHA for d73cd07

  View code at this pointBrowse repository at this point

Commits on Jun 8, 2026

• graph: guard iswa kq_mask on its own buffer (#24294)

  Show description for a66d505
  

  ServeurpersoCom

  and

  ggerganov

  authored
  a66d505View commit details

  Copy full SHA for a66d505

  View code at this pointBrowse repository at this point

Commits on Jun 7, 2026

• llama : add Gemma4 MTP (#23398)

  am17an

  authored
  04eb4c4View commit details

  Copy full SHA for 04eb4c4

  View code at this pointBrowse repository at this point

Commits on Jun 5, 2026

• model, mtmd: Granite4 Vision (#23545)

  Show description for 64086f2
  gabe-l-hart

  authored
  64086f2View commit details

  Copy full SHA for 64086f2

  View code at this pointBrowse repository at this point

• hparams : refactor `hparams.n_layer` (#24060)

  Show description for 7acb4e8
  

  ggerganov

  and

  CISC

  authored
  7acb4e8View commit details

  Copy full SHA for 7acb4e8

  View code at this pointBrowse repository at this point

Commits on Jun 3, 2026

• qwen35: use post-norm hidden state for MTP (#24025)

  Show description for 166fe29
  am17an

  authored
  166fe29View commit details

  Copy full SHA for 166fe29

  View code at this pointBrowse repository at this point

Commits on May 29, 2026

• graph : ensure DS32 kq_mask_lid is F32 (#23864)

  CISC

  authored
  764f1e6View commit details

  Copy full SHA for 764f1e6

  View code at this pointBrowse repository at this point

• model : support for DeepseekV32ForCausalLM with generic DeepSeek Sparse Attention (DSA) implementation (#23346)

  Show description for 1f0aa2a
  

  4 peopleauthored
  1f0aa2aView commit details

  Copy full SHA for 1f0aa2a

  View code at this pointBrowse repository at this point

• llama: use f16 mask for FA to save VRAM (#23764)

  Show description for 031ddb2
  am17an

  authored
  031ddb2View commit details

  Copy full SHA for 031ddb2

  View code at this pointBrowse repository at this point

• llama: add llm_graph_input_mtp (#23643)

  Show description for eef59a7
  

  am17an

  and

  ggerganov

  authored
  eef59a7View commit details

  Copy full SHA for eef59a7

  View code at this pointBrowse repository at this point