metadata
base_model: facebook/opt-350m
language:
- en
license: other
model_name: opt-350m
pipeline_tag: text-generation
quantized_by: iproskurina
tags:
- gptq
- 4-bit
base_model_relation: quantized
inference: false
model_creator: facebook
model_type: opt
OPT-350M - GPTQ
The model published in this repo was quantized to 4bit using AutoGPTQ.
Quantization details
All quantization parameters were taken from GPTQ paper.
GPTQ calibration data consisted of 128 random 2048 token segments from the C4 dataset.
The grouping size used for quantization is equal to 128.
How to use this GPTQ model from Python code
Install the necessary packages
Requires: Transformers 4.33.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later.
pip3 install --upgrade transformers optimum
# If using PyTorch 2.1 + CUDA 12.x:
pip3 install --upgrade auto-gptq
# or, if using PyTorch 2.1 + CUDA 11.x:
pip3 install --upgrade auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/
If you are using PyTorch 2.0, you will need to install AutoGPTQ from source. Likewise if you have problems with the pre-built wheels, you should try building from source:
pip3 uninstall -y auto-gptq
git clone https://github.com/PanQiWei/AutoGPTQ
cd AutoGPTQ
git checkout v0.5.1
pip3 install .
You can then use the following code
from transformers import AutoTokenizer, TextGenerationPipeline,AutoModelForCausalLM
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
pretrained_model_dir = "iproskurina/opt-350m-gptq-4bit"
tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True)
model = AutoGPTQForCausalLM.from_quantized(pretrained_model_dir, device="cuda:0", model_basename="model")
pipeline = TextGenerationPipeline(model=model, tokenizer=tokenizer)
print(pipeline("auto-gptq is")[0]["generated_text"])