HeavenWalker

报错1: RuntimeError: storage has wrong size: expected 768 got 2359296报错2: Unable to load weights from pytorch checkpoint file报错场景: 使用torch.nn.parallel.DistributedDataParallel进行单机多卡模型训练, 保存的bert模型在读取的时候报上述错误, 原因是保存模型的时候使用多线程保存, 猜想是多卡同时写入, 文件在保存过程中出了问题解决

import mtcnn 报错ImportError: cannot import name ‘get_config’ from 'tensorflow.python.eager.context’环境: tensorflow2.2, tensorflow2.3, mtcnn0.1.0>>> import mtcnnD:\anaconda\lib\site-packages\numpy\_distributor_init.py:30: UserWarning: loaded more