Created by: justinmeibi

1.yolov3_darknet.yml coco 数据集加载与训练模型训练yolo_v3，map 一直非常低，4卡batch8迭代20000步,map大概0.14。请问预训练模型是否有问题，或者我这边哪里操作有问题?

2.yolov4_coco 数据集量化训练 CUDA_VISIBLE_DEVICES=2 python slim/quantization/train.py --eval -c ./configs/yolov4/yolov4_cspdarknet_voc.yml -o max_iters=1000 save_dir=./output_v4 LearningRate.base_lr=0.0001 LearningRate.schedulers="[!PiecewiseDecay {gamma: 0.1, milestones: [10000]}]" pretrain_weights=https://paddlemodels.bj.bcebos.com/object_detection/yolov4_cspdarknet.pdparams
已将bn 改为：bn 报错 212-WARNING: recv endsignal from outq with errmsg[consumer[consumer-951-6] failed to map with error:[target0 not in samples]]