Skip to content

soeque1/bert_pytorch_onnx

Repository files navigation

bert_pytorch_onnx

Prepare

(MAC) brew install libomp
pip install -r requirements.txt

config.json

According to the test of pytorch github, the test config, BertModelTester's initializer is used.

Main

mkdir onnx
python bert_to_onnx_fixed_seq.py
python bert_to_onnx_dynamic_seq.py

(Prepare) Server

git clone https://github.com/microsoft/onnxruntime.git --recursive
docker build -t mcr.microsoft.com/azureml/onnxruntime:latest -f onnxruntime/dockerfiles/Dockerfile.server onnxruntime/dockerfiles/

(RUN) Server (HTTP)

docker run -p 8001:8001 -v $PWD/:/usr/server mcr.microsoft.com/azureml/onnxruntime:latest --log_level verbose --model_path=/usr/server/onnx/torch_bert_fixed.onnx --model_name=bert --model_version=1

(Test) Client

PYTHONPATH=./tutorials/tutorials python test_client.py

or

curl -X POST -d "@xxx.json" -H "Content-Type: application/json" http:https://0.0.0.0:8001/v1/models/bert/versions/1:predict

Tests

python -m pytest tests

TODO:

  • gRPC

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Languages