Models directory

# Naming convention

Model released ML model should have version name following the pattern `tYYMM_tag`

where:
* 't' - constant prefix indicating TADAS model
* YY - year of release
* MM - month of release
* tag - model keyword - only letters and digits, undescores are allowed

Each released model should be stored in a separate directory named after the model version.

# Structure of a model directory

* Makefile - build file. Should have targets "run", "publish", "clean"
* cli.py - CLI entry point
* model.pkl - [optional] trained model file
* model_config.py - model specific configuration

# CLI usage

1. check source data availability
2. collect source data
3. run the model
4. store results

```shell
python -m tadas.models.cli --model=<model_version> inference [--date=YYYY-MM-DD] --report=<countries|global>
python -m tadas.models.cli --model=<model_version> sync-to-snowflake --report=<countries|global>
```