# Toy ML models and experiments

This repo contains some experiments with ML I am working on to learn how they work.

## Setup

### Debian 13 (Trixie) Python3 setup

```bash
$ sudo apt install python3-full python3-venv
python3-full is already the newest version (3.13.5-1).
python3-venv is already the newest version (3.13.5-1).
```

### Get the code and init the venv

```bash
git clone http://the.git.repo/user/mltoys.git mltoys && cd mltoys
python3 -m venv .venv
source .venv/bin/activate
```

### Installing python deps

For instructions installing pytorch refer to the [PyTorch Home page]

```bash
pip3 install numpy
# use the nvidia CUDA or CPU only packages if required.
# I use the ROCm packages, so the repo uses the ROCm packages.
pip3 install torch torchvision --index-url https://download.pytorch.org/whl/rocm6.4
pip3 install pandas matplotlib
```

## Running the code

### Training the model

```bash
(.venv) ➜ python3 pairwise_compare.py train
```

### Running inference

```bash
(.venv) ➜ python3 pairwise_compare.py infer
```

### Some General notes

- This module logs to pairwise_compare.log using the python logging module
- Saved model state (training data) is saved at model.pth
- output_graphs.py generates two graphs to inspect the training process, copy the step log into the script to generate graphs.

[PyTorch Home page]: https://pytorch.org