Metadata-Version: 2.4
Name: comfi_fast_grnn_torch
Version: 0.0.3
Summary: A PyTorch implementation of Comfi-FastGRNN
Author-email: Nicolas Arrieta Larraza <NIAL@bang-olufsen.dk>, Niels de Koeijer <NEMK@bang-olufsen.dk>
License: MIT
Project-URL: Homepage, https://github.com/narrietal/Fast-ULCNet
Requires-Python: >=3.10
Description-Content-Type: text/markdown
Requires-Dist: torch>=2.5.1
Requires-Dist: libsegmenter==1.0.4
Requires-Dist: torchinfo==1.8.0
Requires-Dist: CRM_pytorch==0.1.0

# comfi-fast-grnn-torch
Implements Comfi-FastGRNN in torch.

## Usage

The `ComfiFastGRNN` module is designed to be a drop-in replacement for standard PyTorch RNN layers (like `nn.LSTM` or `nn.GRU`), but with added support for low-rank factorization and complementary filtering.

### Basic Implementation
Here is how to use the layer with default settings in a standard training loop:

```python
import torch
from comfi_fast_grnn_torch import ComfiFastGRNN 

comfi_fgrnn = ComfiFastGRNN(
    input_size=32, 
    hidden_size=64, 
    num_layers=1
)
```
