From a6f39ce0d3d17827cad0ce43bb00d40d6b19b168 Mon Sep 17 00:00:00 2001 From: Lance Wang Date: Wed, 22 May 2024 16:16:43 -0700 Subject: [PATCH] Update README.md, the quantize flag is no longer available, quantize_type assumes the role of the original flag. --- README.md | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 42c27765..1c72eca3 100644 --- a/README.md +++ b/README.md @@ -69,10 +69,9 @@ Need to manually modify the `config.json` in the checkpoint folder to make it a ```bash export input_ckpt_dir=Original llama weights directory export output_ckpt_dir=The output directory -export quantize=True #whether to quantize export model_name="llama-3" # or "llama-2", "gemma" -export quantize_type="int8_per_channel" # Availabe quantize type: {"int8", "int4"} x {"per_channel", "blockwise"} -python -m convert_checkpoints --model_name=$model_name --input_checkpoint_dir=$input_ckpt_dir --output_checkpoint_dir=$output_ckpt_dir --quantize=$quantize --quantize_type=$quantize_type +export quantize_type="int8_per_channel" # Availabe quantize type: {"int8", "int4"} x {"per_channel", "blockwise"}, setting this will quantize the weights +python -m convert_checkpoints --model_name=$model_name --input_checkpoint_dir=$input_ckpt_dir --output_checkpoint_dir=$output_ckpt_dir --quantize_type=$quantize_type ```