52 lines
792 B
YAML
52 lines
792 B
YAML
DATASET:
|
|
SELECTION_BATCH_SIZE: 50
|
|
SUBSAMPLE_CLASSES: base
|
|
|
|
|
|
DATALOADER:
|
|
RETURN_IMG0: true
|
|
TRAIN_X:
|
|
BATCH_SIZE: 1
|
|
TEST:
|
|
BATCH_SIZE: 256
|
|
NUM_WORKERS: 4
|
|
|
|
INPUT:
|
|
SIZE: (224, 224)
|
|
INTERPOLATION: "bicubic"
|
|
PIXEL_MEAN: [0.48145466, 0.4578275, 0.40821073]
|
|
PIXEL_STD: [0.26862954, 0.26130258, 0.27577711]
|
|
TRANSFORMS: ["random_resized_crop", "random_flip", "normalize"]
|
|
# CUTOUT_N: 1
|
|
# CUTOUT_LEN: 128
|
|
|
|
OPTIM:
|
|
NAME: "sgd"
|
|
LR: 0.0035
|
|
MAX_EPOCH: 5
|
|
LR_SCHEDULER: "cosine"
|
|
WARMUP_EPOCH: 1
|
|
WARMUP_TYPE: "constant"
|
|
WARMUP_CONS_LR: 1e-5
|
|
|
|
TRAIN:
|
|
PRINT_FREQ: 20
|
|
|
|
MODEL:
|
|
BACKBONE:
|
|
NAME: "ViT-B/16"
|
|
|
|
|
|
|
|
TEST:
|
|
PER_CLASS_RESULT: false
|
|
FINAL_MODEL: "best_val"
|
|
|
|
|
|
|
|
TRAINER:
|
|
MAPLE:
|
|
N_CTX: 2
|
|
CTX_INIT: "a photo of a"
|
|
PREC: "fp16"
|
|
PROMPT_DEPTH: 9 |