Giters

theolepage / sslsv

Framework for training and evaluating self-supervised learning methods for speaker verification.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Supervised version

Subin-Kim46 opened this issue · comments

commented

Hi Théo,

I was impressed by your sslsv project.
In the README.md, self supervised version is well introduced. (Maybe, " python train.py configs/vicreg_b256.yml ")
If i want to use the supervised version, how can I use the supervised version of contrastive learning?
I think if dist.is_available() and dist.is_initialized() are true, it's for self supervised learning. And if Distributed is false, it's for supervised learning. Is it right?

I'll be waiting for the reply.

commented

Hi @Subin-Kim46,

Thank you!

Training a model in a supervised way is not directly implemented in the framework yet.

However, you can use the code in evaluate_label_efficient.py by calling the train method with supervised=True similarly to line 126 (as shown below).

sslsv/evaluate_label_efficient.py

Line 126 in 810926d

train(args, int(nb_labels), supervised=True)

The distributed mode does not refer to the supervision but is used to distribute the training with DistributedDataParallel.

commented

Thank you!