TeleChat-12B-V2-GPTQ-Int4

About Quantization

我们使用modelscope swift仓库进行GPTQ 4bit量化. 量化文档可以查看这里. 量化命令如下:

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

# Experimental Environment: A100
OMP_NUM_THREADS=14 \
    swift export \
    --quant_bits 4 \
    --model_type telechat-12b-v2 \
    --quant_method gptq \
    --dataset alpaca-zh alpaca-en sharegpt-gpt4-mini \
    --quant_seqlen 4096

Inference:

CUDA_VISIBLE_DEVICES=0 swift infer --model_type telechat-12b-v2-gptq-int4

SFT:

CUDA_VISIBLE_DEVICES=0 swift sft --model_type telechat-12b-v2-gptq-int4 --dataset leetcode-python-en

Original Model:

telechat-12b-v2