🐛 Bug

when I train a cnn-rnn-ctc text recognize model, I meet nan loss after some iters, but it's ok at pytorch 0.4 with warpctc

To Reproduce

Steps to reproduce the behavior:

1. download the code from https://github.com/WenmuZhou/crnn.pytorch
2. change the ctc loss from warpctc to nn.CTCloss()
3. run

• PyTorch Version (e.g., 1.0): 1.0.0.dev20181115
• OS (e.g., Linux): ubuntu 16.04
• How you installed PyTorch (conda, pip, source): pip3
• Build command you used (if compiling from source):
• Python version: 3.5.2
• CUDA/cuDNN version: 8.0/6.0
• GPU models and configuration: 1080ti
• Any other relevant information: