Skip to content

Latest commit

 

History

History
115 lines (80 loc) · 4.42 KB

model_card.md

File metadata and controls

115 lines (80 loc) · 4.42 KB

Model card

This is a generative model from the paper "Byte Pair Encoding for Symbolic Music" (EMNLP 2023). The model has been trained with Byte Pair Encoding (BPE) on the Maestro dataset to generate classical piano music with the REMI tokenizer.

Model Details

Model Description

It has a vocabulary of 20k tokens learned with Byte Pair Encoding (BPE) using MidiTok.

Model Sources

Uses

The model is designed for autoregressive music generation. It generates the continuation of a music prompt.

How to Get Started with the Model

Use the code below to get started with the model. You will need the miditok, transformers and torch packages to make it run, that can be installed with pip.

You will also need to manually download the tokenizer.conf file from the repo files.

import torch
from transformers import AutoModelForCausalLM
from miditok import REMI
from miditoolkit import MidiFile

torch.set_default_device("cuda")
model = AutoModelForCausalLM.from_pretrained("Natooz/Maestro-REMI-bpe20k", trust_remote_code=True, torch_dtype="auto")
tokenizer = REMI(params="tokenizer.conf")
input_midi = MidiFile("path/to/file.mid")
input_tokens = tokenizer(input_midi)

generated_token_ids = model.generate(input_tokens.ids, max_length=200)
generated_midi = tokenizer(generated_token_ids)
generated_midi.dump("path/to/continued.mid")

Training Details

Training Data

The model has been trained on the Maestro dataset. The dataset contains about 200 hours of classical piano music. The tokenizer is trained with Byte Pair Encoding (BPE) to build a vocabulary of 20k tokens.

Training Procedure

  • Training regime: fp16 mixed precision on V100 PCIE 32GB GPUs
  • Compute Region: France

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0001
  • train_batch_size: 64
  • eval_batch_size: 96
  • seed: 444
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine_with_restarts
  • lr_scheduler_warmup_ratio: 0.3
  • training_steps: 100000

Environmental impact

We cannot estimate reliably the amount of CO2eq emitted, as we lack data on the exact power source used during training. However, we can highlight that the cluster used is mostly powered by nuclear energy, which is a low carbon energy source ensuring a reduced direct environmental impact.

Citation

BibTeX:

@inproceedings{bpe-symbolic-music,
    title = "Byte Pair Encoding for Symbolic Music",
    author = "Fradet, Nathan  and
      Gutowski, Nicolas  and
      Chhel, Fabien  and
      Briot, Jean-Pierre",
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/2301.11975",
}