Skip to content

SevenZhan/Pytorch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

35 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Tips:

  1. TORCH_HOME: used to custom model dir when downloading pretrained models from model zoos;

Datasets:

  1. Split2TnV: used to create train and validation set in the form of txt file;
  2. ItemList: taken the txt files create from Split2TnV and used to create pytorch datasets which then can be load with pytorch dataloader;

Losses:

  1. SVSoftmax: Support Vector Guided Softmax Loss for Face Recognition: https://arxiv.org/pdf/1812.11317.pdf;
  2. Arcface: ArcFace: Additive Angular Margin Loss for Deep Face Recognition: https://arxiv.org/pdf/1801.07698.pdf;

Schedulers:

  1. CyclicalLR: Cyclical Learning Rates for Training Neural Networks: https://arxiv.org/pdf/1506.01186.pdf;
  2. CosineLR: SGDR: STOCHASTIC GRADIENT DESCENT WITH WARM RESTARTS: https://arxiv.org/pdf/1608.03983.pdf;

CosineAnnealingWarmRestarts(1.3.0):

  1. T_0: cycle length;
  2. T_mult: cycle length multiplier; cosine_annealing_warm_restarts

OneCycleLR(1.3.0):

  1. max_lr: maximum learning rate value;
  2. total_steps: cycle length, can also be set by using epochs & steps_per_epoch; one_cycle