Loss nan #360
Description
Loss is nan at the beginning.
I do nothing but follow the Readme.
I use single 1650desktop to run it. Batch size is set to 16 and accumulate gradient of 2 batch.
{{ message }}
Loss is nan at the beginning.
I do nothing but follow the Readme.
I use single 1650desktop to run it. Batch size is set to 16 and accumulate gradient of 2 batch.