OOM problem for train stage

salu133445 / bmusegan

Code for “Convolutional Generative Adversarial Networks with Binary Neurons for Polyphonic Music Generation”

https://salu133445.github.io/bmusegan/

MIT License

58 stars 13 forks source link

OOM problem for train stage #4

Closed ericosmic closed 6 years ago

ericosmic commented 6 years ago

When I run the train stage (the second stage) of musegan/bmusegan , it would jump out and remind me the OOM problem 2018-11-05 10-17-58 But at same time , my GPU memory >10G for free, it is weird. Do you know what is problem? 2018-11-05 10-32-57

salu133445 commented 6 years ago

You need to reduce the batch size used in the second-stage training. We used 32 for the first stage and 16 for the second stage.