SKDModel-miniImageNet--ravi-resnet12-Gen1-Table2.yaml

augment: true
augment_times: 5
augment_times_query: 1
backbone:
  kwargs:
    avg_pool: true
    is_flatten: true
    keep_prob: 0.0
    maxpool_last2: true
  name: resnet12
batch_size: 64
classifier:
  kwargs:
    alpha: 0.1
    cls_classifier_path: ./results/SKDModel-miniImageNet--ravi-resnet12-5-1-Aug-28-2021-07-01-15/checkpoints/cls_classifier_best.pth
    emb_func_path: ./results/SKDModel-miniImageNet--ravi-resnet12-5-1-Aug-28-2021-07-01-15/checkpoints/emb_func_best.pth
    feat_dim: 640
    gamma: 1.0
    is_distill: true
    kd_T: 2
    num_class: 64
  name: SKDModel
data_root: /data/wzy/miniImageNet--ravi
deterministic: true
device_ids: 3
episode_size: 1
epoch: 65
image_size: 84
includes:
- headers/data.yaml
- headers/device.yaml
- headers/misc.yaml
- headers/model.yaml
- headers/optimizer.yaml
- classifiers/SKD.yaml
- backbones/resnet12.yaml
log_interval: 100
log_level: info
log_name: null
log_paramerter: false
loss:
  kwargs: null
  name: CrossEntropyLoss
lr_scheduler:
  kwargs:
    T_max: 100
    eta_min: 1.0e-06
  name: CosineAnnealingLR
n_gpu: 1
optimizer:
  kwargs:
    lr: 0.001
    momentum: 0.9
    weight_decay: 0.0005
  name: SGD
  other:
    emb_func: 0.001
parallel_part:
- emb_func
- cls_classifier
- rot_classifier
pretrain_path: null
query_num: 15
result_root: ./results
resume: false
save_interval: 10
save_part:
- emb_func
- cls_classifier
- rot_classifier
seed: 0
shot_num: 1
tb_scale: 1.6666666666666667
test_episode: 600
test_epoch: 5
test_query: 15
test_shot: 1
test_way: 5
train_episode: 1000
use_loss_yaml: true
use_memory: false
way_num: 5