Code

Integer quantisation support#540

Code

Merged

ggerganov merged 15 commits intomasterggerganov/whisper.cpp:masterfrom 4-bitggerganov/whisper.cpp:4-bitCopy head branch name to clipboardApr 30, 2023

+3,180-1,007

Commits

Commits on Apr 30, 2023

• whisper : add integer quantization support

  ggerganov

  committed
  e9d78e6

  Copy full SHA for e9d78e6

• examples : add common-ggml + prepare to add "quantize" tool

  ggerganov

  committed
  d814e95

  Copy full SHA for d814e95

• whisper : quantization tool ready

  ggerganov

  committed
  9d6bcad

  Copy full SHA for 9d6bcad

• whisper : fix F32 support

  ggerganov

  committed
  8b91958

  Copy full SHA for 8b91958

• whisper : try to fix shared lib linkage

  ggerganov

  committed
  23f63e0

  Copy full SHA for 23f63e0

• wasm : update quantized models to Q5

  ggerganov

  committed
  f641252

  Copy full SHA for f641252

• bench.wasm : remove "medium" button

  ggerganov

  committed
  ec3e882

  Copy full SHA for ec3e882

• bench.wasm : fix custom model button

  ggerganov

  committed
  9364cfd

  Copy full SHA for 9364cfd

• ggml : add Q5_0 and Q5_1 WASM SIMD

  ggerganov

  committed
  524470d

  Copy full SHA for 524470d

• wasm : add quantized models to all WASM examples

  ggerganov

  committed
  a3e96e7

  Copy full SHA for a3e96e7

• wasm : bump DB version number to 2

  ggerganov

  committed
  7c43711

  Copy full SHA for 7c43711

• talk-llama : update example to latest llama.cpp

  ggerganov

  committed
  05137ce

  Copy full SHA for 05137ce

• node : increase test timeout to 10s

  ggerganov

  committed
  ced04d0

  Copy full SHA for ced04d0

• readme : add information for model quantization

  ggerganov

  committed
  d378f5d

  Copy full SHA for d378f5d

• wasm : add links to other examples

  ggerganov

  committed
  7aeb79d

  Copy full SHA for 7aeb79d