search

samsad35 / VQ-MAE-S-code

A Vector Quantized Masked AutoEncoder for speech emotion recognition
https://ieeexplore.ieee.org/document/10193151
GNU Affero General Public License v3.0
16 stars 1 forks source link

ft vqvae #2

Open WWWWxp opened 10 months ago

WWWWxp commented 10 months ago

Dear author, in your fine-tuning code, I found that vqvae is not used. Why?

samsad35 commented 10 months ago

In fact, we store all encoder output indices in an HDF5 file to improve training and finetuning speed. To do this, see the file: HDF5 creation

WWWWxp commented 10 months ago

Looking forward to your release of the pre-trained model VQVAE

samsad35 commented 9 months ago

@Fwinner-wxp Hi, I apologize for missing your message earlier. I've now shared the weights for the VQVAE model (Link).