ZeruiW / XAI-Service

Quick Review Demo Web Portal

http://34.73.250.83:5173/

Important Update

1. az_blob_connection_str.json and mongo.*.conf are no longer required. All configuration goes to .env.dev or .end.pred. For instance, the following environment variables are required:

   ENV=dev
   MONGO_CONF_STR=
   AZ_BLOB_STR=

2. For the frontend:

   Make sure you have the .env file in frontend-x:

   VITE_ENV=dev
   VITE_BASE_URL=http://127.0.0.1:5006/

Local Dev with Flask

Before you start, please have your .env.dev file ready with the required environment variables.

1. Start mongo container

   docker compose -f ./backend/docker-compose.yml -f ./backend/docker-compose-dev.yml --project-directory . up mongo --build

2. Start Central:

   pip install -q backend/central_dev/. && flask --app 'backend/central:create_app("dev")' run -p 5006

3. Start Azure blob service

   flask --app backend/db_service/azure_blob run -p 5009

4. Start ResNet50:

   flask --app backend/model_service/resnet50 run -p 5001

5. Start Grad-CAM XAI:

   flask --app 'backend/xai_service/pytorch_cam:create_app(cam_method="grad-cam")' run -p 5003

6. Start Eval Service:

   flask --app backend/evaluation_service run -p 5004

7. Start Frontend:

   docker compose -f frontend-x/docker-compose.yml --project-directory . -p frontend up fex --build

Check this link for API and use case:

https://www.postman.com/youyinnn/workspace/concordia/collection/2019955-72d3c5f3-2070-4bba-97de-d5990085b20e?action=share&creator=2019955&active-environment=2019955-c8be28eb-2739-48db-89b7-a74fc752029c

1. activate the central;
2. register db, xai, model, evaluation service;
3. create xai tasksheet, available config for demo:

   {
     "method_name": "grad-cam",
     "data_set_name": "imagenet1000",
     "data_set_group_name": "g0",
     "model_name": "resnet50",
     "executor_config": {
       "use_pytorch_multiprocess": true
     }
   }

4. run the task;
5. check the result;

Local Dev with Docker and Docker-compose

Before you start, make sure (1) the .env.dev is ready and (2) you are in the location of the project root.

docker compose -f ./backend/docker-compose.yml -f ./backend/docker-compose-dev.yml --project-directory . up

If you want to push your changes to the docker image, use

docker compose -f ./backend/docker-compose.yml -f ./backend/docker-compose-dev.yml --project-directory . up --build

Then:

Please use http://host.docker.internal` to replace all localhost or 127.0.0.1 for your service registration and demonstration. For instance:

and

If you want to build a single service:

docker build . -f ./backend/central/Dockerfile -t central