【mmdetection实践】(五)理解train的过程
文章目录 之前的几篇文章已经分别理解了: 如何定义自己的数据集 如何训练自己的网络 dataset和model是怎么构造的 本文就再详细的看一下,在构造好了dataset和model是如何训练的。从tools/train.py中 # mmdetection/tools/train.py train_detector ( model , datasets , cfg , distributed = distributed , validate = args . validate , timestamp = timestamp , meta = meta ) 可以进入在mmdet/apis/train.py中 # mmdetection/mmdet/apis/train.py runner = Runner ( model , batch_processor , optimizer , cfg . work_dir , logger = logger , meta = meta ) . . . runner . run ( data_loaders , cfg . workflow , cfg . total_epochs ) 再查看runner,可以看到, # mmcv/runner/runner.py class Runner : def train ( self , data