yolov8_l_500e_coco.yml

_BASE_: [
  '../datasets/coco_detection.yml',
  '../runtime.yml',
  '_base_/optimizer_500e_high.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_500e_coco/model_final


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


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