arishiki/Llama-3.2-1B-Instruct-quantized-gptq-8g01

This model is a quantized version of Llama-3.2-1B-Instruct. The quantization was done using AutoGPTQ (https://github.com/AutoGPTQ/AutoGPTQ). Code used for generation is as follows:

from transformers import AutoModelForCausalLM, AutoTokenizer, GPTQConfig
import torch

model_id = "meta-llama/Llama-3.2-1B-Instruct"

quantization_config = GPTQConfig(
     bits=8,
     group_size=128,
     dataset="c4",
     desc_act=False,
)

tokenizer = AutoTokenizer.from_pretrained(model_id)
quant_model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=quantization_config, device_map='auto')