Mistral-7B-v0.1-onnx

This is an optimized version of the Mistral 7B model, available on this repository: https://huggingface.co/mistralai/Mistral-7B-v0.1 and under the license on such repository. Microsoft permits you to use, modify, redistribute, and create derivatives of Microsoft's contributions to the optimized version subject to the restrictions and disclaimers of warranty and liability in license agreement.

Mistral-7b for ONNX Runtime

Introduction

This repository hosts the optimized versions of Mistral-7B-v0.1 to accelerate inference with ONNX Runtime CUDA execution provider.

See the usage instructions for how to inference this model with the ONNX files hosted in this repository.

Model Description

• Developed by: MistralAI
• Model type: Pretrained generative text model
• License: Apache 2.0 License
• Model Description: This is a conversion of the Mistral-7B-v0.1 for ONNX Runtime inference with CUDA execution provider.

Performance Comparison

Latency for token generation

Below is average latency of generating a token using a prompt of varying size using NVIDIA A100-SXM4-80GB GPU, taken from the ORT benchmarking script for Mistral

Usage Example

Following the benchmarking instructions. Example steps:

1. Clone onnxruntime repository.

git clone https://github.com/microsoft/onnxruntime
cd onnxruntime

1. Install required dependencies

python3 -m pip install -r onnxruntime/python/tools/transformers/models/llama/requirements-cuda.txt

1. Inference using manual model API, or use Hugging Face's ORTModelForCausalLM

from optimum.onnxruntime import ORTModelForCausalLM
from onnxruntime import InferenceSession
from transformers import AutoConfig, AutoTokenizer

sess = InferenceSession("Mistral-7B-v0.1.onnx", providers = ["CUDAExecutionProvider"])
config = AutoConfig.from_pretrained("mistralai/Mistral-7B-v0.1")

model = ORTModelForCausalLM(sess, config, use_cache = True, use_io_binding = True)

tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-v0.1")

inputs = tokenizer("Instruct: What is a fermi paradox?\nOutput:", return_tensors="pt")

outputs = model.generate(**inputs)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))