ai_project_template/example_1

A simple AI project template for classification.

To run:

First, you need to run

conda env create -f conda.yaml

to make the environment and

conda activate aitest

to get into it. Then you need to prepare the data sample by running:

python 0-preparation.py

It downloads the data into data directory. Then you should run:

mlflow run . --no-conda

This will run 1-train.py with the default parameters to train a model.

In order to run the file with custom parameters, you can run it like:

mlflow run . \
-P run_name='test' \
-P batch_size=64 \
-P epochs=20 \
-P aug_rot=45 \
-P aug_w=0.05 \
-P aug_h=0.05 \
-P aug_zoom=0.05 \
-P model_path='../models' \
--no-conda

Once the code is finished executing, you can view the run's metrics, parameters, and details by running the command

mlflow ui

and navigating to http://localhost:5000.

After all and to predict with the trained model, you can run the User Interface by:

streamlit run 2-predict.py