forecast-keiba

競馬予想のプロジェクトです。 Python, Docker, Kedro, MLflowを使っています。

Getting Started for Developer

# set google-chrome-stable_current_amd64.deb
https://dl.google.com/linux/direct/google-chrome-stable_current_amd64.deb
->
docker/pytorch_1_4/google-chrome-stable_current_amd64.deb

# build container
$ docker build ./ -t forecast-keiba

# run container
$ sh run.sh docker

# [optional]run container which use gpu(nvidia-docker)
$ sh run.sh nvidia-docker

# [optional]install jupyter extensions
---
pip3 install jupyter-tabnine
jupyter nbextension install --py jupyter_tabnine
jupyter nbextension enable --py jupyter_tabnine
jupyter serverextension enable --py jupyter_tabnine
---

# start jupyter notebook
$ kedro jupyter notebook --allow-root --port=8888 --ip=0.0.0.0 &

# exec pipeline
$ kedro run

# setup mlflow
$ mlflow ui --host 0.0.0.0 --backend-store-uri logs/mlruns

# set credentials
$ vim conf/local/credentials.yml
---
dev_line:
  access_token: <line access token>
---

Description

• OS version

  $ cat /etc/issue
  Ubuntu 18.04.3

• Python version

  $ python -V
  Python 3.7.4

• PyTorch version

  >>> torch.__version__
  '1.4.0'

Reference

• 競馬予想で始めるデータ分析・機械学習