A repository containing resources for pre-training, fine-tuning, and evaluating musical (MIDI) transformer models.
Note that this project is under active development
The main goal of the gpt-aria project is to create a suite of powerful pre-trained generative (symbolic) music models. We want to investigate how modern training (pre-training & fine-tuning) techniques can be used to improve the quality/usefulness of such models. Alongside this we are building various data (MIDI) preprocessing tools, allowing you to easily fine-tune our models on your own data.
If you are new to symbolic music models, a good place to start are the following projects/blogposts by Google Magenta and OpenAI:
Long story short: Transformer + MIDI + GPUs = 🎵 x ∞
Make sure you are using Python 3.10+. Note that I haven't explicitly developed this project for anything other than Linux. If you are using Windows, things might not work properly. In this case I suggest installing using WSL.
git clone https://github.com/eleutherai/aria
cd aria
pip install -e .
You can find preliminary checkpoints at the following locations
Finetuned piano-only checkpoints (improved robustness):
large - https://storage.googleapis.com/aria-checkpoints/large-abs-inst.safetensors
Pretrained checkpoints:
large - https://storage.googleapis.com/aria-checkpoints/large-abs-pt.bin
medium - https://storage.googleapis.com/aria-checkpoints/medium-abs-pt.bin
small - https://storage.googleapis.com/aria-checkpoints/small-abs-pt.bin
You can then sample using the cli:
aria sample \
-m large \
-c <path-to-checkpoint> \
-p <path-to-midifile> \
-var <num-variations-to-generate> \
-trunc <seconds-in-to-truncate-prompt> \
-l <number-of-tokens-to-generate> \
-temp 0.95 \
-e
You can use aria sample -h to see a full list of options. If you wish to sample from a pretrained checkpoint, please use the -pt flag.