Commits

Branch selector

 master

User selector

All users

All time

Commit History

Commits on Jan 18, 2024

• ggml : sync ggml-metal.m

  ggerganov

  committedJan 18, 2024
  fb466b3

Commits on Jan 17, 2024

• sync : llama.cpp

  ggerganov

  committedJan 17, 2024
  1f50a7d

• sync : ggml

  ggerganov

  committedJan 17, 2024
  1de21b9

• ggml : add IQ2 to test-backend-ops + refactoring (llama/4990)

  ggerganov

  committedJan 17, 2024
  4aea058

• imatrix : offload to GPU support (llama/4957)

  ggerganov

  committedJan 17, 2024
  fd10234

• backend : add eval callback (llama/4935)

  ggerganov

  committedJan 17, 2024
  8fb5c6a

• metal : create autorelease pool during library build (llama/4970)

  ggerganov

  committedJan 17, 2024
  2fe5fbf

• ggml : importance matrix support for legacy quants (llama/4969)

  

  2 peopleauthored and

  ggerganov

  committedJan 17, 2024
  01637e1

• metal : log recommendedMaxWorkingSetSize on iOS 16+ (llama/4936)

  

  azarovalex

  and

  ggerganov

  committedJan 17, 2024
  1b349eb

• ggml : introduce GGML_CALL function annotation (llama/4850)

  

  jart

  authored and

  ggerganov

  committedJan 17, 2024
  138eaeb

• cuda : fix dequantize kernel names (llama/4938)

  ggerganov

  committedJan 17, 2024
  61b9192

• CUDA: faster dequantize kernels for Q4_0 and Q4_1 (llama/4938)

  

  2 peopleauthored and

  ggerganov

  committedJan 17, 2024
  161b51d

• Add ability to use importance matrix for all k-quants (llama/4930)

  

  2 peopleauthored and

  ggerganov

  committedJan 17, 2024
  f904b31

Commits on Jan 16, 2024

• talk-llama : optional wake-up command and audio confirmation (ggerganov#1765)

  Rakksor

  committedJan 16, 2024
  f661415

Commits on Jan 15, 2024

• server : fix building and simplify lib deps on Windows (ggerganov#1772)

  przemoc

  committedJan 15, 2024
  f5f159c

Commits on Jan 14, 2024

• talk-llama : sync llama.cpp

  ggerganov

  committedJan 14, 2024
  6ebba52

• talk-llama : llama.cpp

  ggerganov

  committedJan 14, 2024
  2a58744

• sync : ggml

  ggerganov

  committedJan 14, 2024
  d08445c

• metal : correctly set SIMD support flags on iOS (llama/4923)

  

  azarovalex

  authored and

  ggerganov

  committedJan 14, 2024
  4a94569

• 2-bit quantizations (llama/4897)

  

  2 peopleauthored and

  ggerganov

  committedJan 14, 2024
  dabc964

• scripts : sync-ggml-am.sh add option to skip commits

  ggerganov

  committedJan 14, 2024
  654baf6

Commits on Jan 13, 2024

• talk-llama : sync llama.cpp

  ggerganov

  committedJan 13, 2024
  f001a3b

• sync : ggml

  ggerganov

  committedJan 13, 2024
  c615f2c

• examples : adapt to metal API

  ggerganov

  committedJan 13, 2024
  d839dd0

• ggml: cache sin/cos for RoPE (llama/4908)

  

  JohannesGaessler

  authored and

  ggerganov

  committedJan 13, 2024
  4358478

• metal : remove old API (llama/4919)

  ggerganov

  committedJan 13, 2024
  182f290

• metal : disable log for loaded kernels (llama/4794)

  ggerganov

  committedJan 13, 2024
  447dfc1

• gguf : fix potential infinite for-loop (llama/4600)

  

  2 peopleauthored and

  ggerganov

  committedJan 13, 2024
  9aa9f3b

• metal : refactor kernel loading code (llama/4794)

  ggerganov

  committedJan 13, 2024
  396ebd1

• CUDA: faster q8_0 -> f16 dequantization (llama/4895)

  

  JohannesGaessler

  authored and

  ggerganov

  committedJan 13, 2024
  12490f4

• talk-llama : add optional CLI arg to set the bot name (ggerganov#1764)

  RhinoDevel

  committedJan 13, 2024
  db078a9

• examples : add python example for transcription (ggerganov#1744)

  contractorwolf

  committedJan 13, 2024
  a13a7da

• whisper : load the model into multiple buffers of max size 1GB (ggerganov#1763)

  ggerganov

  committedJan 13, 2024
  519f8e8

Commits on Jan 12, 2024

• talk-llama : sync llama.cpp

  ggerganov

  committedJan 12, 2024
  40ae096

• sync : ggml

  ggerganov

  committedJan 12, 2024
  1560288

Pagination

PreviousNext