Detic / configs /Detic_LbaseCCcapimg_CLIP_R5021k_640b64_4x_ft4x_max-size.yaml
AK391
files
159f437
_BASE_: "Base-C2_L_R5021k_640b64_4x.yaml"
MODEL:
WITH_CAPTION: True
SYNC_CAPTION_BATCH: True
ROI_BOX_HEAD:
ADD_IMAGE_BOX: True # caption loss is added to the image-box
USE_ZEROSHOT_CLS: True
IMAGE_LABEL_LOSS: 'max_size'
WEIGHTS: "models/BoxSup-C2_Lbase_CLIP_R5021k_640b64_4x.pth"
SOLVER:
MAX_ITER: 90000
IMS_PER_BATCH: 64
BASE_LR: 0.0002
WARMUP_ITERS: 1000
WARMUP_FACTOR: 0.001
DATASETS:
TRAIN: ("lvis_v1_train_norare","cc3m_v1_train_tags")
DATALOADER:
SAMPLER_TRAIN: "MultiDatasetSampler"
DATASET_RATIO: [1, 4]
USE_DIFF_BS_SIZE: True
DATASET_BS: [8, 32]
DATASET_INPUT_SIZE: [640, 320]
USE_RFS: [True, False]
DATASET_INPUT_SCALE: [[0.1, 2.0], [0.5, 1.5]]
FILTER_EMPTY_ANNOTATIONS: False
MULTI_DATASET_GROUPING: True
DATASET_ANN: ['box', 'captiontag']
NUM_WORKERS: 8
WITH_IMAGE_LABELS: True