Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 14, 2025

• convert : refactor rope scaling handling (#18013)

  Show description for 5c8a717
  CISC

  authored
  5c8a717

  Copy full SHA for 5c8a717

• mtmd: enhance image resizing in llava_uhd (#18014)

  bluebread

  authored
  37f5a10

  Copy full SHA for 37f5a10

• vulkan: fix mul_mat_vec_iq1_s formatting (#18026)

  0cc4m

  authored
  9e6649e

  Copy full SHA for 9e6649e

• graph: add f_attn_temp_offset (#18025)

  ngxson

  authored
  0759b09

  Copy full SHA for 0759b09

• common : refactor common_sampler + grammar logic changes (#17937)

  Show description for 254098a
  ggerganov

  authored
  254098a

  Copy full SHA for 254098a

• vulkan: Fix data race/hang in scalar/cm1 flash attention (#17887)

  jeffbolznv

  authored
  3238b14

  Copy full SHA for 3238b14

• vulkan: improve mul_mat_vec_iq1_s speed (#17874)

  lovedheart

  authored
  4722671

  Copy full SHA for 4722671

• vulkan: faster q6_k matmul (#17813)

  Show description for d15d177
  netrunnereve

  authored
  d15d177

  Copy full SHA for d15d177

• model-conversion : cast logits to float32 (#18009)

  ggerganov

  authored
  77ad854

  Copy full SHA for 77ad854

• models : fix YaRN regression + consolidate logic (#18006)

  Show description for 609a2d0
  ggerganov

  authored
  609a2d0

  Copy full SHA for 609a2d0

• ggml : arm repack fix build

  ggerganov

  committed
  a63cbaf

  Copy full SHA for a63cbaf

• sync : ggml

  ggerganov

  committed
  0e59224

  Copy full SHA for 0e59224

• ggml : arm repack fix build (whisper/0)

  ggerganov

  committed
  71fdcf0

  Copy full SHA for 71fdcf0

• cmake : set `CMAKE_RUNTIME_OUTPUT_DIRECTORY` for non standalone build (ggml/1394)

  Show description for 615655a
  

  HerrCai0907

  authored and

  ggerganov

  committed
  615655a

  Copy full SHA for 615655a

Commits on Dec 13, 2025

• scripts: add script to compare logprobs of llama.cpp against other frameworks (#17947)

  Show description for c00ff92
  ngxson

  authored
  c00ff92

  Copy full SHA for c00ff92

• server-models.cpp: add missing <filesystem> (#18000)

  Show description for 4ed2bae
  barracuda156

  authored
  4ed2bae

  Copy full SHA for 4ed2bae

• llama_context: synchronize before reallocating output buffer (#17974)

  jeffbolznv

  authored
  5266379

  Copy full SHA for 5266379

• arg: fix common_params_parse not accepting negated arg (#17991)

  ngxson

  authored
  4d5ae24

  Copy full SHA for 4d5ae24

• cmake: correct scope - link ws2_32 for MinGW/w64devkit builds in cpp-httplib (#17972)

  Show description for 66ba512
  gustrd

  authored
  66ba512

  Copy full SHA for 66ba512

• vulkan: support get_rows for i32 (#17941)

  jeffbolznv

  authored
  36255a2

  Copy full SHA for 36255a2

• vulkan: support GGML_OP_DIAG (#17893)

  jeffbolznv

  authored
  3229a23

  Copy full SHA for 3229a23

• vulkan: Multi-pass softmax for large number of cols (#17892)

  Show description for 303f861
  jeffbolznv

  authored
  303f861

  Copy full SHA for 303f861

• speculative-simple : free batch on exit (#17985)

  ggerganov

  authored
  3c6391e

  Copy full SHA for 3c6391e

• common : skip model validation when --completion-bash is requested (#17975)

  CISC

  authored
  8e4d678

  Copy full SHA for 8e4d678

• vulkan: Allow non-pow2 n_experts in topk_moe (#17872)

  jeffbolznv

  authored
  07a10c1

  Copy full SHA for 07a10c1

• add llama-completion to completion-bash executables (#17976)

  CISC

  authored
  2bc94e7

  Copy full SHA for 2bc94e7

• model-conversion : use CONVERTED_MODEL value for converted model [no ci] (#17984)

  Show description for fd1085f
  danbev

  authored
  fd1085f

  Copy full SHA for fd1085f

Commits on Dec 12, 2025

• common: support negated args (#17919)

  Show description for 380b4c9
  

  ngxson

  and

  CISC

  authored
  380b4c9

  Copy full SHA for 380b4c9

• clip: move model cgraphs into their own files (#17965)

  Show description for e39a2ce
  ngxson

  authored
  e39a2ce

  Copy full SHA for e39a2ce

• ci : change the cann version and the container pull method (#17953)

  Show description for a8c7f33
  xuedinge233

  authored
  a8c7f33

  Copy full SHA for a8c7f33

• docker : include legacy llama-completion binary (#17964)

  CISC

  authored
  b7f5f46

  Copy full SHA for b7f5f46

• CUDA: fix overflow in MMA kernel without stream-k (#17939)

  JohannesGaessler

  authored
  4822114

  Copy full SHA for 4822114

• models : fix the attn_factor for mistral3 graphs + improve consistency (#17945)

  Show description for 7bed317
  ggerganov

  authored
  7bed317

  Copy full SHA for 7bed317

• cann : fix ops broken by circular padding guard (#17825)

  CISC

  authored
  dcb7d17

  Copy full SHA for dcb7d17

• ggml-cpu : fix RISC-V Q4_0 repack select and RVV feature reporting (#17951)

  Show description for 5160443
  

  ixgbe

  and

  ggerganov

  authored
  5160443

  Copy full SHA for 5160443

Pagination

Previous

Next