Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

MusicGen

Model basics

Reasoning traces

Not supported

Context length

2K tokens

Max output length

No data

Model type

基础大模型

Release date

2023-06-08

Model file size

6.51GB

MoE architecture

Total params / Active params

33.0B / N/A

Knowledge cutoff

No data

Inference modes

No mode data

MusicGen

Open source & experience

Code license

MIT License

Weights license

CC-BY-NC 4.0- 不可以商用

GitHub repo

https://github.com/facebookresearch/audiocraft

Hugging Face

https://huggingface.co/facebook/musicgen-large

Live demo

No live demo

MusicGen

Official resources

Paper

Simple and Controllable Music Generation

DataLearnerAI blog

No blog post yet

MusicGen

API details

API speed

No data

No public API pricing yet.

MusicGen

Benchmark Results

No benchmark data to show.

MusicGen

Publisher

Facebook AI研究实验室

View publisher details

MusicGen

Model Overview

MusicGen是一个自回归tansformer模型，通过32kHz EnCodec Tokenizer训练，4个codebooks在50Hz采样。与现有的方法不同，如MusicGen MusicGen不需要自我监督的语义表示，它在一次传递中生成所有4个码本。

MusicGen代码开源协议MIT协议，可商用。但是预训练结果的开源协议是CC-BY-NC 4.0，不可商用！

DataLearner 官方微信

欢迎关注 DataLearner 官方微信，获得最新 AI 技术推送