https://github.com/captain-pool/hydra-wandb-sweeper

WandB sweeps integration with Hydra sweeper
https://github.com/captain-pool/hydra-wandb-sweeper

Last synced: 4 months ago
JSON representation

WandB sweeps integration with Hydra sweeper

• Host: GitHub
• URL: https://github.com/captain-pool/hydra-wandb-sweeper
• Owner: captain-pool
• Created: 2022-02-15T22:26:37.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2024-01-26T16:01:04.000Z (5 months ago)
• Last Synced: 2024-01-26T17:24:17.619Z (5 months ago)
• Language: Python
• Homepage:
• Size: 28.3 KB
• Stars: 37
• Watchers: 4
• Forks: 4
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Lists

• awesome-stars - captain-pool/hydra-wandb-sweeper - WandB sweeps integration with Hydra sweeper (Python)

README

        
## WandB sweeper integration with Hydra

### Installing

```bash

$ pip install git+https://github.com/captain-pool/hydra-wandb-sweeper.git

```

## Usage

Using this plugin is very simple, as one can just invoke the sweeper by overriding `hydra/sweeper` with `wandb`.

```python

@hydra.main(config_path=..., config_name=...)

def main(cfg: DictConfig):

  with wandb.init(id=run_id) as run:

    ...

    wandb.log({"loss": loss})

```

## Hydra Overrides

### Passing Distributions for continuous parameters

Sweeping over continuous params (for ex. learning rates in ML pipelines) can be done by passing the upper limit and lower with the hydra's `interval(, )` override. For example:

```

$ python3 /path/to/trainer/file hydra/sweeper=wandb model.learning_rate="interval(0.1, 0.2)"

```

By default, the `interval()` sweep uses the uniform distribution while making parameter suggestions. To use a different distribution add the name of the supported distribution (`uniform`, `log_uniform`, `q_uniform`, `q_log_uniform`, `q_normal`, `log_normal`, `q_log_normal`) as a [`tag`](https://hydra.cc/docs/advanced/override_grammar/extended/#tag) to the [`interval()`](https://hydra.cc/docs/advanced/override_grammar/extended/#interval-sweep) sweep. For example, to use a normal distribution with `mean=0` and `variance=1` to suggest a value for `learning_rate` use the following:

```

$ python3 /path/to/trainer/file hydra/sweeper=wandb model.learning_rate="tag(normal, interval(0, 1))"

```

### Categorial parameters

Sweeping over categorical items is as simple as passing a list of items to sweep over thereby directly using Hydra' [`choice()`](https://hydra.cc/docs/advanced/override_grammar/extended/#choice-sweep) sweep feature. for example, to sweep over a categorical list of `batch_size`,

```

$ python3 /path/to/trainer/file hydra/sweeper=wandb model.batch_size=8,10,12,14

```

Equivalently, the `choice()` command can also be expliclty used:

```

$python3 /path/to/trainer/file hydra/sweep=wandb model.batch_size="choice(8,10,12,14)".

```

Categorical sweep can also be done using hydra's [`range()`](https://hydra.cc/docs/advanced/override_grammar/extended/#range-sweep) sweep. The previous task can be equivalently achieved as:

```

$ python3 /path/to/trainer/file hydra/sweeper=wandb model.batch_size="range(8, 15, step=2)"

```