mangoHoli

我的label如果过长的话，loss会变成nan或者inf

有空的话，路径问题和预训练模型能不能贴出埋坑步骤？太难了，我下载了checkpoints但里面没有ckpt呀我一脸mb。哭唧唧

I used python3.6 and tensorflow 1.14.0. It works!

> @mangoHoli I met the same error, have you solved it? I forgot how i solved it. But you can check if you got the following code in ‘WORK_PATH/crnn/models/utils.py’: #!/usr/bin/python...

> I cant find data_parallel in utils. Only me? I got this in utils. -------------- #!/usr/bin/python # encoding: utf-8 import torch.nn as nn import torch.nn.parallel def data_parallel(model, input, ngpu): if...

I got the same problem, I used tensorflow 1.14.0. I changed ckpt_path and tensorflow to 1.7.0. Cannot fix it. I think the author deleted the pretrained model as checkpoint. Can...