search

hyperopt / hyperopt-convnet

Convolutional computer vision architectures that can be tuned by hyperopt.
70 stars 20 forks source link

single precision float #7

Open tdomhan opened 10 years ago

tdomhan commented 10 years ago

I just wanted to note that I needed to add THEANO_FLAGS='floatX=float32' in order to run the test successfully.

Because if I didn't tensor.sqrt would return a double precision number and then the assert in line 492 of file pyll_slm.py would fail.

jaberg commented 10 years ago

Good catch! The code should assert that explicitly... On Nov 16, 2013 12:18 PM, "Tobias Domhan" notifications@github.com wrote:

I just wanted to note that I needed to add THEANO_FLAGS='floatX=float32'in order to run the test successfully.

Because if I didn't tensor.sqrt would return a double precision number and then the assert in line 492 of file pyll_slm.py would fail.

— Reply to this email directly or view it on GitHubhttps://github.com/hyperopt/hyperopt-convnet/issues/7 .