Code

Use full range for q4_0 quantization#729

Code

Merged

ggerganov merged 6 commits intoggml-org:masterggml-org/llama.cpp:masterfrom

unbounded:q4_0_range_fixunbounded/llama.cpp:q4_0_range_fixCopy head branch name to clipboard

Apr 25, 2023

Commits

Commits on Apr 24, 2023

• Use full range for q4_0 quantization

  Show description for f57433c
  unbounded

  committed
  f57433c

  Copy full SHA for f57433c

• Update quantize_row_q4_0 for AVX/AVX2

  unbounded

  committed
  84aa7d8

  Copy full SHA for 84aa7d8

• Update quantize_row_q4_0 for WASM

  Show description for 73a92d2
  unbounded

  committed
  73a92d2

  Copy full SHA for 73a92d2

• Update quantize_row_q4_0 for Arm NEON

  unbounded

  committed
  fea8d10

  Copy full SHA for fea8d10

• Update quantize_row_q4_0 for PowerPC

  Show description for d09f97e
  unbounded

  committed
  d09f97e

  Copy full SHA for d09f97e

• Use full range for q4_2 quantization

  unbounded

  committed
  5808fcf

  Copy full SHA for 5808fcf