yolov8_l_100e_oiv7.yml

_BASE_: [
  '../../datasets/openimagesv7_detection.yml',
  '../../runtime.yml',
  '../_base_/optimizer_100e.yml',
  '../_base_/yolov8_cspdarknet.yml',
  '../_base_/yolov8_reader_high_aug.yml',
]
depth_mult: 1.0
width_mult: 1.0

log_iter: 50
snapshot_epoch: 10
weights: output/yolov8_l_100e_oiv7/model_final


YOLOv8CSPDarkNet:
  last_stage_ch: 512 # The actual channel is int(512 * width_mult), not int(1024 * width_mult) as in YOLOv5

YOLOv8Head:
  customized_c3: 256


TrainReader:
  batch_size: 16 # default 8 gpus, total bs = 128