Open lyna-tlncgbnn opened 4 months ago
In the pre-training phase, all parameters are randomly initialized.
In the pre-training phase, all parameters are randomly initialized.
Yes, what I mean is: Do you have all the after pre trained weight parameters here, not just for the encoder part of downstream tasks。I want to continue pre-training with checkpoint for special dataset。
Thank you very much for your open source code. May I ask if there are weight parameters obtained from the decoder and projector sections during pre training?