search

hubertsiuzdak / snac

Multi-Scale Neural Audio Codec (SNAC) compresses audio into discrete codes at a low bitrate
https://hubertsiuzdak.github.io/snac/
MIT License
439 stars 26 forks source link

Training with attention or not #18

Open Naminwang opened 5 months ago

Naminwang commented 5 months ago

Hi, when i see the config on hugging face for model predict, the attn_window_size is null, so i wonder if the attention is used in training state? And, can you share some training details, some thing like lr, the size of training data...