zhmxu

SSD_loss 训练公开数据集在几个 epoch 后 loss 突然变 nan

7

 使用 object_detection 中的SSD 训练 deepfashion 数据集时，在经过几个 epoch 后会随机出现 loss nan 如图，（多次调整learning_rate, batch_size都会随机出现nan ），数据跑过一轮，应该不是数据标注问题， 猜测可能是 ssd_loss 中数值边界处理的问题 

_caffe.cpp libprotobuf.lib(generated_message_util.obj) : fatal error LNK1112: module machine type 'x64' conflicts with target machine type 'X86'

4

When I "\python\caffe_caffe.cpp" compiled into "_caffe.pyd", there have been some problems: libprotobuf.lib(generated_message_util.obj) : fatal error LNK1112: module machine type 'x86' conflicts with target machine type 'X64' what's wrong with me?...