search

NVIDIA / tacotron2

Tacotron 2 - PyTorch implementation with faster-than-realtime inference
BSD 3-Clause "New" or "Revised" License
5.06k stars 1.38k forks source link

Is it a good idea to continue training from where I left off after updating my dataset? #518

Closed Aayzie closed 3 years ago

Aayzie commented 3 years ago

Let's say I'm at 50 epoch, and then I updated my dataset by adding new lines, is it better to continue from my last checkpoint before I updated the dataset, or is it better to start over from the pretrained LJ model? Thanks :)

creator-kevinovich commented 3 years ago

It depends on the specific algorithm/framework you are using to train your model. Some of them support incremental learning. If this is the case, you can add data to your model afterwards to update your model weights.

Aayzie commented 3 years ago

It depends on the specific algorithm/framework you are using to train your model. Some of them support incremental learning. If this is the case, you can add data to your model afterwards to update your model weights.

Okay Pog thanks