Pytorch deployment with Flask

Deplying an image classifier with pytorch and Flask

pip install -r requirements.txt

Set up API

python src/flask_api.py --model_path model/model_checkpoint.pth --class_index_path model/class_index_dict.json --endpoint_name classifier --host localhost --port 5000

Make predictions

python src/predict.py --img_dir test_images --XAI False --endpoint_name classifier --host localhost --port 5000

The model can be trained using the repo: https://github.com/europeana/rd-img-classification-pilot

Run with docker

pull the image

docker pull axiom1123/torch_deployment:v4

as soon as you downloaded the image execute the command

docker run -t -p 0.0.0.0:3030:3030 torch_deployment:v4 python src/flask_api.py --model_path model/model_checkpoint.pth --class_index_path model/class_index_dict.json --endpoint_name classifier --host 0.0.0.0 --port 3030

the container is running now and the service is available

Run with Startup script

simple run the bash-script

bash start.sh