# philo

**Repository Path**: sing_jay_lee/philo

## Basic Information

- **Project Name**: philo
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: GPL-3.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-12-18
- **Last Updated**: 2021-12-18

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# philo

<!-- > Bringing modalities together since eternity

*Philo* (derived from [Philophrosyne](https://en.wikipedia.org/wiki/Philophrosyne), the goddess of friendship) is a module for adaptive fusion of multiple modalities.
It possess two special spells right now, namely, *AutoFusion* and *GanFusion*.
Both of them fall under high-level fusion-magic class of *Ada* (most spells under this class *adaptive* in nature).
Philo's *src* of power is PyTorch.


Leave a :star: if you find the code useful :smiley: -->

Repository for Adaptive Fusion Techniques for Multimodal Data (Python package to be released soon)

## Usage instructions:

Make sure that you are in the root directory before running the following commands:
* Modify the `src/models/config.py` file appropriately before running any commands
* If it's your first run (starting with nothing but the raw files), run `python src/process_raw.py`. This will create processed dataset files inside `data/processed`
* Once you have the processed files, run `python src/train.py` to train the network. It will take care of:
    - Preprocessing the text
    - Creating/Loading vocabulary
    - Generating/Loading embeddings
    - Training the model