A basic fork of official frontend module of ModelDB machine learning management system.
Visit the original project at: https://github.com/mitdbg/modeldb
This project goal is to try to add some additional features:
- Confusion Matrix
- Compute resources consumption
- OAuth integration
By using the script launch.sh you will can launch almost all operations (type ./launch.sh help or see contents of the file for more details).
Besides, there are some variables to setup the script (all of them with MDB_ prefix):
| NAME | DESCRIPTION | Example |
|---|---|---|
| MDB_THRIFT_LOCAL | Execute thrift from distribuition (defaults to true) or use a docker container (false) | |
| MDB_THRIFT_VERSION | Thrift version (defaults to 0.10.0) | |
| MDB_THRIFT_MODEL_VERSION | Version of the ModelDB.thrift file (defaults to master) | |
| MDB_WAIT_BACKEND | Wait backend be reachable (defaults to true) | |
| MDB_BACKEND_HOST | Set backend host (defaults to localhost) | |
| MDB_DOCKER_NAME | Docker container name (defaults to modeldb-frontend) | |
| MDB_DOCKER_NAME_TAG | Docker tag for container (defaults to latest) | |
| MDB_DOCKER_OPTS | Default docker options (defaults to -d --net=host) |
First of all make sure that the ModelDB server (backend) is running and adjust the MDB_BACKEND_HOST variable.
Here you have a couple of examples about how to run it:
- Running on localhost, generating thrift sources by a docker container (by default) previously:
$ MDB_THRIFT_LOCAL=false ./launch.sh gen
$ ./launch.sh start- Running on localhost through a docker container:
$ ./launch.sh docker-build
$ ./launch.sh docker-startNOTE: Thrift generated code is passed as volume to the container in build time (directory thrift was created in first step of the above example). If you want to generate the thrift code within the same container then provide MDB_THRIFT_LOCAL="true" with the docker-build operation
In order to purge all created resources above enter this command ./launch.sh clean-all (included docker containers and images)
After all, the frontend should be available at http://localhost:3000