conv1 weights and biases become nan

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Follow the steps in Compiling, Data, and TrainingExample for ILSVRC2012
2. I used same parameters for running convnet except --train-freq=10


What is the expected output? What do you see instead?
Keep training with correct weights. (Top-5 error reaches 0.7 before 13th epochs)
After 14th epochs, conv1 weights and biases become nan and top-5 error becomes 
0.99...

What version of the product are you using? On what operating system?
latest version of cuda-convnet2, Ubuntu 14.04

Please provide any additional information below.

Original issue reported on code.google.com by lywoo...@gmail.com on 1 Dec 2014 at 1:07