Getting Started

ts_diffusion is a community project investigating cutting-edge diffusion models for time series.

Prepare Dev Environment

0. Prerequisites

• pre-commit: set up git commit hooks
• whichever python virtual env management tool, pyenv, conda, ...
• poetry: package and env management

Python Environment

Here is a one-pager about python environment and pre-commit.

1. Setup pre-commit

1. Install pre-commit.
2. Clone this repository.
3. Install git hooks using pre-commit install.
   1. This has to be done ASAP, before committing any changes.
   2. If one forgets to do this before git commit, run pre-commit run --all-files then commit again.

2. Setup Python Environment

1. Install poetry.
2. Use your python environment management tool to create a environment with a specific python version,
3. e.g., conda create -n ts-diffusion python=3.10 pip if using conda, or
4. pyenv shell 3.10.9 if using pyenv.
5. Tell poetry we want to use this python environment: poetry env use 3.10.8.
6. Install dependencies: poetry install

'I hate poetry!'

If you really hate poetry, please use the requirements.txt to install dependencies.

However, please note that this file is automatically updated by poetry though the poetry-export hook. Any manual changes to this requirements.txt file will be overwritten by our automated pre-commit hooks.

Datasets

We can download datasets and load them through ts-bolt.

All contents in the datasets folder is excluded from git.

Download datasets

This repository ships a datasest downloading tool called "bolt".

poetry run bolt download --name ecb_exchange_rate --target datasets

List and Inspect Datasets

List datasets:

poetry run bolt list

Inspect the details of a dataset:

poetry run bolt list --name=ecb_exchange_rate