Skip to content

Latest commit

 

History

History
17 lines (14 loc) · 923 Bytes

README.md

File metadata and controls

17 lines (14 loc) · 923 Bytes
Raw

Data Preprocessing for training EnCLAP

1. Make EnCodec Embeddings

python infer_encodec.py --data_path data/wavs --save_path data/encodec_embeddings

2. Make CLAP Embeddings

python infer_clap.py --data_path data/wavs --save_path data/clap_embeddings --clap_ckpt clap_ckpt_path

You may give different options to --enable_fusion and --audio_encoder flags based on the CLAP checkpoint you are using.

3. Make CSV files

You should make CSV files corresponding to train, validation, and test dataset. Examples are given in csv for AudioCaps and Clotho datasets.

  • Train csv files should have columns file_path and caption. If an audio file is labeled with multiple captions, they should be made listed in separate entries.
  • Validation and Test csv files should have columns file_path,caption_1, caption_2,...., caption_n if n captions are given for a single audio file.