config.yaml

DATA_ROOT: '../datasets/'
LOGS_ROOT: '../logs/'


MODEL:
  # architecture
  generator: 'resnet_9blocks'
  discriminator: 'patchgan_3layers'
  spgan: False

  sync_bn: False
  samples_per_bn: 64


DATA:

  height: 256
  width: 128
  norm_mean: [0.5, 0.5, 0.5]
  norm_std: [0.5, 0.5, 0.5]

  TRAIN:
    # augmentation
    is_autoaug: False

    is_flip: True
    flip_prob: 0.5

    is_pad: False
    pad_size: 10

    is_blur: False
    blur_prob: 0.5

    is_erase: False
    erase_prob: 0.5

    # dual augmentation for MMT
    is_mutual_transform: False
    mutual_times: 2


TRAIN:
  seed: 1
  deterministic: True
  # mixed precision training for PyTorch>=1.6
  amp: False

  # datasets
  datasets: {'market1501': 'trainval', 'dukemtmcreid': 'trainval'}
  unsup_dataset_indexes: [1,]

  epochs: 50
  iters: 200

  LOSS:
    losses: {'gan_G': 1., 'recon': 10., 'ide': 0.5, 'gan_D': 1.}

  # validate
  val_freq: 1

  # sampler
  SAMPLER:
    num_instances: 0
    is_shuffle: True

  # data loader
  LOADER:
    samples_per_gpu: 8
    workers_per_gpu: 2

  # optim
  OPTIM:
    optim: 'adam'
    lr: 0.0002
    adam_beta1: 0.5
    weight_decay: 0

  SCHEDULER:
    lr_scheduler: 'linear'
    n_epochs_init: 25
    n_epochs_decay: 25


TEST:

  # data loader
  LOADER:
    samples_per_gpu: 32
    workers_per_gpu: 4