Commits

Branch selector

 avoid-gnu-source

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 25, 2023

• ggml : do not use _GNU_SOURCE gratuitously

  ggerganov

  committed
  78fafca

  Copy full SHA for 78fafca

• readme : add new roadmap + manifesto

  ggerganov

  authored
  447ccbe

  Copy full SHA for 447ccbe

• ggml : sync latest ggml (custom operators)

  ggerganov

  committed
  bd34cdd

  Copy full SHA for bd34cdd

• fix server sampling: top k sampler first (#1977)

  Show description for c2a08f8
  

  anon998

  and

  anon

  authored
  c2a08f8

  Copy full SHA for c2a08f8

• readme : add Azure CI discussion link

  ggerganov

  authored
  66a2555

  Copy full SHA for 66a2555

• zig : upgrade build system support (#1981)

  Show description for e65ca7e
  

  coderonion

  and

  ggerganov

  authored
  e65ca7e

  Copy full SHA for e65ca7e

Commits on Jun 24, 2023

• #1869 Fix null reference errors when training from scratch with CUDA (#1907)

  Show description for 5ec8dd5
  

  robyngraf

  and

  ggerganov

  authored
  5ec8dd5

  Copy full SHA for 5ec8dd5

• tests : sync test-grad0 from ggml

  ggerganov

  committed
  65bdd52

  Copy full SHA for 65bdd52

• flake : fix ggml-metal.metal path and run nixfmt (#1974)

  novafacing

  authored
  fdd1860

  Copy full SHA for fdd1860

• convert : fix invalid params in write_vocab_only (#1975)

  aisk

  authored
  c943d82

  Copy full SHA for c943d82

• ggml : improve ggml_graph_dump_dot, add ggml_format_name (#1978)

  Show description for f2c754e
  slaren

  authored
  f2c754e

  Copy full SHA for f2c754e

• readme : fix whitespaces

  ggerganov

  committed
  11da1a8

  Copy full SHA for 11da1a8

• readme : fixed termux instructions (#1973)

  albbus-stack

  authored
  235b610

  Copy full SHA for 235b610

• llama : fix top-p sampling to match the canonical definition (#1953)

  Show description for b061ba9
  alexrenda

  authored
  b061ba9

  Copy full SHA for b061ba9

• llama : make model stateless and context stateful (llama_state) (#1797)

  Show description for 527b6fb
  

  didzis

  and

  ggerganov

  authored
  527b6fb

  Copy full SHA for 527b6fb

Commits on Jun 23, 2023

• Add OpenLLaMA instructions to the README (#1954)

  Show description for d7b7484
  

  eiery

  authored
  d7b7484

  Copy full SHA for d7b7484

Commits on Jun 22, 2023

• rework convert.py to read hyper-parameters from config.json (#1958)

  Show description for 7487137
  Green-Sky

  authored
  7487137

  Copy full SHA for 7487137

Commits on Jun 21, 2023

• cmake: revert CUDA arch default to 52, 61 if f16 (#1959)

  JohannesGaessler

  authored
  bbca06e

  Copy full SHA for bbca06e

• Fix typo in README.md (#1961)

  RahulVivekNair

  authored
  fb98254

  Copy full SHA for fb98254

Commits on Jun 20, 2023

• readme : add link to p1

  ggerganov

  authored
  049aa16

  Copy full SHA for 049aa16

• Fix typo (#1949)

  sammysun0711

  authored
  2322ec2

  Copy full SHA for 2322ec2

• llama : fix params struct slignment (#1936)

  Show description for aacdbd4
  mudler

  authored
  aacdbd4

  Copy full SHA for aacdbd4

Commits on Jun 19, 2023

• [Fix] Reenable server embedding endpoint (#1937)

  Show description for 20568fe
  SlyEcho

  authored
  20568fe

  Copy full SHA for 20568fe

• ggml : fix bug in LBFGS optimizer (found by ggml tests)

  ggerganov

  committed
  18b3562

  Copy full SHA for 18b3562

• llama : use aligned memory during ggml_init call from loading saved sessions (#1934)

  Show description for ba4e85a
  l3utterfly

  authored
  ba4e85a

  Copy full SHA for ba4e85a

• cmake : fix trailing whitespaces

  ggerganov

  committed
  23fc5c2

  Copy full SHA for 23fc5c2

• llama : only use Q6_K for output weights if tensor size is multiple of 256 (#1932)

  Show description for cb40dfc
  

  ikawrakow

  and

  Kawrakow

  authored
  cb40dfc

  Copy full SHA for cb40dfc

• cuda : faster k-quants on older GPUs (#1930)

  Show description for ca7c3f4
  

  ikawrakow

  and

  Kawrakow

  authored
  ca7c3f4

  Copy full SHA for ca7c3f4

• ggml : sync latest ggml repo (#1924)

  Show description for b97ca43
  ggerganov

  authored
  b97ca43

  Copy full SHA for b97ca43

• cmake : fix build shared ggml when CUDA is enabled (#1929)

  Show description for 1e3abfc
  

  howard0su

  and

  ggerganov

  authored
  1e3abfc

  Copy full SHA for 1e3abfc

• Convert vector to f16 for dequantize mul mat vec (#1913)

  Show description for 16b9cd1
  JohannesGaessler

  authored
  16b9cd1

  Copy full SHA for 16b9cd1

Commits on Jun 18, 2023

• Added tokens per second to info prints (#1928)

  JohannesGaessler

  authored
  b24c304

  Copy full SHA for b24c304

• Fixed incorrectly applying RMS norm twice (#1925)

  JohannesGaessler

  authored
  0ede372

  Copy full SHA for 0ede372

• ggml : fix bug in ggml_compute_forward_add_q_f32 (#1918)

  l3utterfly

  authored
  8596af4

  Copy full SHA for 8596af4

• readme : update Android build instructions (#1922)

  Show description for e1886cf
  mikeyang01

  authored
  e1886cf

  Copy full SHA for e1886cf

Pagination

Previous

Next