Metadata-Version: 2.2
Name: kowalski
Version: 0.0.2
Summary: A suite of standard metrics for assessing learned feature representations, in PyTorch.
Author-email: Viet Anh Khoa Tran <v.tran@fz-juelich.de>
License: MIT
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: torch
Requires-Dist: einops
Requires-Dist: matplotlib
Requires-Dist: jaxtyping

# Kowalski
This library provides a suite of standard metrics for assessing learned feature representations, in PyTorch.

## Installation
```
pip install kowalski
```

### Example
```python
from kowalski import to_per_class_list
from kowalski.neural_collapse import class_distance_normalized_variance as cdnv
import torch
features = torch.randn(100, 128)
labels = torch.randint(0, 10, (100,))

print(cdnv(to_per_class_list(features, labels)))
```
