train loss is nan

[harokell]

In the train log, loss is always nan.

[shaoyu-liu]

In the train log, loss is always nan.

+1

[shaoyu-liu]

In the train log, loss is always nan.

Have you solved this problem?