An open API service indexing awesome lists of open source software.

https://github.com/neuro-ml/thunder

đŸŒŠī¸ The Deep Learning framework based on Lightning
https://github.com/neuro-ml/thunder

deep-learning experiments machine-learning python pytorch pytorch-lightning

Last synced: 10 months ago
JSON representation

đŸŒŠī¸ The Deep Learning framework based on Lightning

Awesome Lists containing this project

README

          

[![docs](https://img.shields.io/badge/-docs-success)](https://neuro-ml.github.io/thunder/)
![License](https://img.shields.io/github/license/neuro-ml/thunder)
[![codecov](https://codecov.io/gh/neuro-ml/thunder/branch/master/graph/badge.svg)](https://codecov.io/gh/neuro-ml/thunder)
[![pypi](https://img.shields.io/pypi/v/thunder?logo=pypi&label=PyPi)](https://pypi.org/project/thunder/)

> _You saw the lightning. Now it's time to hear the thunder_ đŸŒŠī¸

# Thunder đŸŒŠī¸

The Deep Learning framework based on [Lightning](https://github.com/Lightning-AI/lightning).

## Install

```bash
pip install thunder
```

:warning:
> Currently thunder is not published on pypi. Install it via git clone.

## Start experimenting

It's as simple as 1, 2, 3:

1. Create a config (e.g. `base.config`):
```python
from myproject import MyDataset, MyModule
from lightning import Trainer
from torch.utils.data import DataLoader

# these 3 fields are required
train_data = DataLoader(MyDataset())
module = MyModule()
trainer = Trainer()
```

2. Build the experiment:
```shell
thunder build base.config /path/to/some/folder
```

3. Run it
```shell
thunder run /path/to/some/folder
```

Also, 2 and 3 can be combined into a single command:
```shell
thunder build-run base.config /path/to/some/folder
```

## More advanced stuff

See [our docs](https://neuro-ml.github.io/thunder/) for a full list of neat things `thunder`đŸŒŠī¸ can do for you