Skip to main content

Running the Babble Trainer

Having file/directory issues on Linux? Did your model training fail, or do you just want to run things manually? You've come to the right place!

This page covers how to download and run the source code for the Babble Trainer.

1) Prerequisites

To get started, you will need to install:

Windows

If you're on on Windows, you'll also need to download:

macOS/Linux

For macOS/Linux:

2) Downloading The Babble Trainer

Navigate to a folder where you would like to install the Babble Trainer. We'll be using the folder BabbleTrainer.

Then, open a terminal in your folder. Make sure its current directory matches the folder!

Now, run this command to download the Babble Trainer to the folder:

git clone https://github.com/Project-Babble/BabbleTrainer 

Downloading Python

Now, we need to install a version of Python the Babble Trainer is compatible with.

For reference, here is a compatibility table of the Babble Trainer and different versions of Python.
< 3.73.83.93.103.113.123.13
⚠️⚠️
note

Python 3.13 support is planned once key dependencies are built against 3.13, namely onnxruntime-directml and onnxruntime-gpu.

While not supported, nightly builds of the above dependencies exist that enable this. Microsoft outlines how to set this up here.

info

We recommend Python version 3.12.8.

In the same terminal, run:

pyenv init

If applicable, follow the instructions the above command outputs.

Once you have done that, in the same terminal as above, run:

pyenv install 3.12.8
pyenv global 3.12.8

This will set the global version of Python in your machine. If you ever need to change it, you can run the last command above with a different version.

3) Setting up

Now, in the same terminal as above, run:

Windows

cd BabbleTrainer
python -m venv venv
./venv/Scripts/activate.ps1

macOS/Linux

cd BabbleTrainer
python -m venv venv
source venv/bin/activate

This creates a virtual environment which is used to store the Babble Trainer's dependencies. Now run:

pip install -r requirements.txt

And wait for pip to install everything. This might take some time!

After it's done, navigate into the babble_data folder and run python setup.py install to build the model data loader. You may be prompted to install a C++ workload to build this, do so if needs be.

4) Running the Babble Trainer

Finally, you are ready to run the Babble Trainer! Run:

python main.py /path/to/user_cal.bin output_onnx_file.onnx

5) Subsequent Runs

For subsequent runs, you will need to activate the virtual environment prior to running the Babble Trainer.

You can do this by running the last line in Step 3 in tandem with the command from Step 4, making sure your terminal window is in the same location as it was before.

info

Make sure you are in your Babble Trainer folder prior to running these, or they will not work!

Windows

./venv/Scripts/activate.ps1
python main.py /path/to/user_cal.bin output_onnx_file.onnx

macOS/Linux

source venv/bin/activate
python main.py /path/to/user_cal.bin output_onnx_file.onnx

That's it!