v0.1.0-alpha.3 lachesis

• Adds SRE20-CTS recipe v1
• Fixes some config files in voxceleb/adv.v2

v0.1.0-alpha.2 lachesis

Fixed make_voxceleb2cat.pl, all speakers extracted from the same video were getting the same spkid, the script did not take into account that more than one speaker could be extracted from each video

v0.1.0-alpha.1 lachesis

• Fixed some bugs in adversarial attacks scripts
• Added VOiCES challenge recipe
• Added Spine2Net results
• Updated table of recipes

v.0.1.0-alpha lachesis

• Configuration files and command line arguments are handled using jsonarparse, not argparse anymore. This allow to use yaml files and override the values in the yaml file from the command line.
• First version using nn.DistributedDataParallel instead of nn.DataParallel
• Supports FairScale Sharded DataParallel, we didn't observer significant memory gains in our models using this so far
• Added SpineNet, Spine2Net and TSE-Spine2Net x-vector architectures from our IS21 paper
• New recipes:
  • Recipe for classifying adversarial attacks algorithms and threat models from our IS21 paper (voxceleb/adv.v2)
  • Recipe for adv attacks aginst spk verif renamed as voxceleb/adv.v1 and adv.v1.1, these recipes have been updated and cleaned up
  • Recipe for SRE19-AV Audio part with AHC diarization (sre19-av-a/v2.1)
  • Recipe for Chime5 speaker verification setup chime5/v1

v0.0

Last version using nn.DataParallel and argparse, next versions will use nn.DistributedDataParallel for multigpu and jsonargparse to parse all arguments and model hyperparameters

NVIDIA-Apex-last

Last version using NVIDIA Apex for automatic mixed-precision training before switching to pytorch1.6 native amp.

gard-eval1.0

Release for the first GARD eval on spk classification