Traceback with running with large networks about char-rnn

[hughperkins]

Hi. As you say, the memory allocation failure is associated with your GPU, not related to your system main memory.

[croxis]

Sad, but good to know, thank you.

[hughperkins]

By the way, do you mind posting the parameters you are hoping to use for training, so I can use them in my testing. It wont solve your memory issue, but just out of my own personal curiosity really. Edit, oh I see you've included them, in your command output ;-)

[hughperkins]

Do you mind giving me some details about your dataset, ie how large it is, the various dimensions and any other characteristics and so on please?

[croxis]

I was going to target something along the lines of:
-eval_val_every 500 -num_layers 6 -rnn_size 600 -dropout 0.5 -seq_length 100
I don't know how to evaluate the dataset, so I just zipped it up (1 meg zipped): https://www.dropbox.com/s/77uiqp1zbhfgut4/mtg.zip?dl=0

[hughperkins]

Thanks!