Skip to content

Latest commit

 

History

History
40 lines (24 loc) · 1.09 KB

README.MD

File metadata and controls

40 lines (24 loc) · 1.09 KB

Sequential Attention (SeqAttn)

The official repo for the paper Discovering Music Relations with Sequential Attention.

Blog Post & Demos

https://music-x-lab.github.io/SeqAttn/

Pre-trained Models

You can get all the pre-trained models in the paper here:

https://drive.google.com/drive/folders/1FydcZKpzvvpjMaY5DOCL5k3P_1I78MdC?usp=sharing

After downloading, put the ".sdict" files in the "cache_data" folder.

Running Training

We have included the processed nottingham dataset and chpop dataset in the data folder. Please uncompress the data.zip file to the data folder first.

You can re-train the models with the following example code:

seq_attention_model.py nottingham 4
seq_attention_bidirectional_model.py nottingham 4

Model Evaluation

See comparison_result.py for details.

Model Generation

You can run generation using pre-trained models with the example code below:

seq_attention_bidirectional_model_generation.py nottingham 4 1.0
seq_attention_bidirectional_model_generation.py chpop 4 1.0
seq_attention_bidirectional_model_generation.py chpop 16 1.0