metadata

license: apache-2.0
datasets:
  - lambada
language:
  - en
library_name: transformers
pipeline_tag: text-generation
tags:
  - text-generation-inference
  - causal-lm
  - int8
  - ONNX
  - PostTrainingStatic
  - Intel® Neural Compressor
  - neural-compressor

INT8 GPT-J 6B

GPT-J 6B is a transformer model trained using Ben Wang's Mesh Transformer JAX. "GPT-J" refers to the class of model, while "6B" represents the number of trainable parameters.

This int8 ONNX model is generated by neural-compressor and the fp32 model can be exported with below command:

python -m transformers.onnx --model=EleutherAI/gpt-j-6B onnx_gptj/ --framework pt --opset 13 --feature=causal-lm-with-past

Test result

	INT8	FP32
Lambada Acc	0.7944	0.7954
Model size (GB)	6	23

How to use

Download the model and script by cloning the repository:

git clone https://huggingface.co/Intel/gpt-j-6B-int8-static

Then you can do inference based on the model and script 'evaluation.ipynb'.