Mustafa Kamal issues

Repositories
Issues
Comments

Results 3 issues of


                                            Mustafa Kamal

Slow training

I am training IR-152 on MsCeleb1M. The gpu i am training on is NVIDIA RTX A4000. It takes about 11 hours per epoch. And if i want to train the...

Output size

what would be the output size of curricularface? would it return a scalar value or a tensor of dimension [batch_size,no_of_classes]?

Clarification on Varying Prediction Lengths Across Datasets

I noticed that the prediction lengths used in the paper vary depending on the dataset. Could you please explain the rationale behind selecting different prediction lengths for different datasets? Are...