Commits

Branch selector

 gg/kv-cache-use-set-rows

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 23, 2025

• kv-cache : utilize ggml_set_rows broadcast

  Show description for 36f8e20
  ggerganov

  committed
  36f8e20

  Copy full SHA for 36f8e20

• cont : support non-continuous slots

  Show description for 332f073
  ggerganov

  committed
  332f073

  Copy full SHA for 332f073

• cont : kv-cells cp/set for non-cont slots

  Show description for 39d0b1e
  ggerganov

  committed
  39d0b1e

  Copy full SHA for 39d0b1e

• cont : migrate to using set of indices instead of slot head

  Show description for f875d6c
  ggerganov

  committed
  f875d6c

  Copy full SHA for f875d6c

• cont : gate the ggml_set_rows usage with env var

  Show description for db2bb37
  ggerganov

  committed
  db2bb37

  Copy full SHA for db2bb37

• kv-cache : use ggml_set_rows

  Show description for 79dac3c
  ggerganov

  committed
  79dac3c

  Copy full SHA for 79dac3c

• ggml : fix supports_op

  

  rgerganov

  authored and

  ggerganov

  committed
  1f647b5

  Copy full SHA for 1f647b5

• ggml : simplify forward_dup_f32

  

  rgerganov

  authored and

  ggerganov

  committed
  eba9757

  Copy full SHA for eba9757

• metal : add ggml_set_rows implementation

  Show description for c0cfc2f
  ggerganov

  committed
  c0cfc2f

  Copy full SHA for c0cfc2f

• tests : add ggml_set_rows

  ggerganov

  committed
  828e5d2

  Copy full SHA for 828e5d2

• ggml : ggml_set_rows update comment + better index name

  ggerganov

  committed
  e73690a

  Copy full SHA for e73690a

• ggml : support GGML_TYPE_F32 ".from_float" trait

  ggerganov

  committed
  e897097

  Copy full SHA for e897097

• ggml : ggml_set_rows support quantized dst

  Show description for 630c84a
  ggerganov

  committed
  630c84a

  Copy full SHA for 630c84a

• ggml : ggml_set_rows support broadcast

  ggerganov

  committed
  df71c80

  Copy full SHA for df71c80

• ggml : add ggml_is_contiguous_rows

  ggerganov

  committed
  313a444

  Copy full SHA for 313a444

• ggml : add repeat impl for i64

  ggerganov

  committed
  695b6b7

  Copy full SHA for 695b6b7

• use I64 for indices

  

  rgerganov

  authored and

  ggerganov

  committed
  f2cd962

  Copy full SHA for f2cd962

• ggml : add ggml_set_rows

  Show description for c1a581a
  

  rgerganov

  authored and

  ggerganov

  committed
  c1a581a

  Copy full SHA for c1a581a

• kv-cells : fix tracking of seq_pos (#14339)

  Show description for 7b50d58
  ggerganov

  authored
  7b50d58

  Copy full SHA for 7b50d58

• vulkan: update windows SDK in CI (#14334)

  jeffbolznv

  authored
  3a9457d

  Copy full SHA for 3a9457d

Commits on Jun 22, 2025

• quantize : handle user-defined pruning of whole layers (blocks) (#13037)

  EAddario

  authored
  fa4a9f2

  Copy full SHA for fa4a9f2

• gguf-py : fix SpecialVocab parsing when post_processor is null (#14330)

  CISC

  authored
  238005c

  Copy full SHA for 238005c

• run : avoid double tokenization (#14327)

  Show description for 66aba7a
  retr0reg

  authored
  66aba7a

  Copy full SHA for 66aba7a

• examples : fix is_first logic for tokenization (#14329)

  Show description for f1f5e82
  ggerganov

  authored
  f1f5e82

  Copy full SHA for f1f5e82

• HIP: enable vec fattn on RDNA4 (#14323)

  IMbackK

  authored
  af3373f

  Copy full SHA for af3373f

• mtmd : fix Pixtral OOM with large images by capping image_size to 1024 (#14326)

  Show description for 5d5c066
  yuiseki

  authored
  5d5c066

  Copy full SHA for 5d5c066

• common : use std::string_view now that we target c++17 (#14319)

  CISC

  authored
  40bfa04

  Copy full SHA for 40bfa04

• CUDA: add mean operation (#14313)

  Show description for aa064b2
  am17an

  authored
  aa064b2

  Copy full SHA for aa064b2

Commits on Jun 21, 2025

• gguf-py : fix Qwen3-Embedding eos token (#14314)

  CISC

  authored
  aa0ef5c

  Copy full SHA for aa0ef5c

• Add support for VK_EXT_debug_utils to add labels to Vulkan objects. (#13792)

  Show description for bb16041
  mtavenrath

  authored
  bb16041

  Copy full SHA for bb16041

• gguf-py : fix TemplateProcessing pair when bos/eos is missing (#14312)

  CISC

  authored
  58cba76

  Copy full SHA for 58cba76

• metal : fix thread-safety (#14300)

  Show description for 67ae531
  ggerganov

  authored
  67ae531

  Copy full SHA for 67ae531

• memory : rename interface to llama_memory_context_i (#14296)

  Show description for 692e3cd
  ggerganov

  authored
  692e3cd

  Copy full SHA for 692e3cd

• convert : fix Llama 4 conversion (#14311)

  danielhanchen

  authored
  b23fa0b

  Copy full SHA for b23fa0b

Commits on Jun 20, 2025

• sync : ggml

  Show description for 06cbedf
  ggerganov

  committed
  06cbedf

  Copy full SHA for 06cbedf

Pagination

Previous

Next