We use the modelscope swift repository to perform GPTQ 4bit quantization. Quantization documentation can be found here. The quantization command is as follows:

bash

# Experimental Environment: A100
OMP_NUM_THREADS=14 \
	swift export \
	--quant_bits 4 \
	--model_type yi-1_5-34b-chat \
	--quant_method gptq \
	--dataset alpaca-zh alpaca-en sharegpt-gpt4-mini \
	--quant_seqlen 4096

Inference:

CUDA_VISIBLE_DEVICES=0 swift infer --model_type yi-1_5-34b-chat-gptq-int4

SFT:

CUDA_VISIBLE_DEVICES=0 swift sft --model_type yi-1_5-34b-chat-gptq-int4 --dataset leetcode-python-en

Original Model:

YI1.5-34B-Chat

modelscope/Yi-1.5-34B-Chat-GPTQ

About Quantization

modelscope/
Yi-1.5-34B-Chat-GPTQ