Why "gradOutput" buffers are not freed after calling clearState()?

Element-Research / rnn

Recurrent Neural Network library for Torch7's nn

BSD 3-Clause "New" or "Revised" License

941 stars 313 forks source link

Why "gradOutput" buffers are not freed after calling clearState()? #425

Open gcheron opened 7 years ago

gcheron commented 7 years ago

Before saving my network containing "FastLSTM" modules, I call "clearState()" but the gradOutput buffers of the FastLSTM are not freed resulting in very big model file (and saving the model takes time). Any idea why? Thanks!