Segmentation fault caused by: f_grad_shared(qi_i, D_gt_id, is_train)

[johnnyflame]

Hi there,

When running the program, it appears to crash due to segmentation fault at line 827 in run.py:
out = f_grad_shared(qi_i, D_gt_id, is_train)

The program is running on CentOS, with Theano 0.9.

Any help/insight would be greatly appreciated. Cheers.

[rodrigonogueira4]

How many RAM do you have? I had segmentation fault problems when running on machines with less than 32 GB of RAM. If that is your case, try using smaller batch_size_train and batch_size_pred (for example 32 or 16).

[johnnyflame]

I have 56 GB of RAM, and have experimented with different batch sizes (32,16 and 8). Unfortunately it made no differences, still produces the same error.