changes for python3.6

[Whadup]

Hi there,
to run on my mac with python 3.6, I had to do these small changes.
apparently, the api for collections has changed from iteritems() to items() and typechecking for numpy integers is more strict (int divided by 2 gives float)
Hope this helps,
Cheers
Lukas

[EigenPro]

Thanks Lukas! This neat modification is compatible with Python 2.7. So I would like to just merge it. But could you revert the change for the assert (L 40) in this commit.? I want to keep this for a while to avoid the relatively disappointing performance on CPU.

[Whadup]

I will post another pull request with my changes once I’ve written up my results using your algorithm. Until then, maybe use this patch? From f071377 Mon Sep 17 00:00:00 2001 From: Lukas Pfahler <lukas.pfahler@tu-dortmund.de> Date: Tue, 5 Dec 2017 10:21:04 +0100 Subject: [PATCH] changes for python3.6

…

--- run_expr.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/run_expr.py b/run_expr.py index f2ba50e..c66e35d 100644 --- a/run_expr.py +++ b/run_expr.py @@ -128,7 +128,7 @@ trainers['Kernel EigenPro'] = Trainer(model=model, # Assemble SGD trainer. rff_weights = np.float32( # for Gaussian kernel np.sqrt(2. / (2 * 5 ** 2)) # s = 5 - * np.random.randn(D, d/2)) + * np.random.randn(D, d>>1)) input_shape = (D,) x = Input(shape=input_shape, dtype='float32', name='feat') rf_f = RFF(rff_weights, input_shape=input_shape) @@ -159,7 +159,7 @@ trainers['EigenPro with random Fourier feature'] = Trainer( model=model, x_train = x_train, x_test=x_test) # Start training. -for name, trainer in trainers.iteritems(): +for name, trainer in trainers.items(): print("") initial_epoch=0 np.random.seed(1) # Keras uses numpy random number generator -- 2.13.1

On 10. Dec 2017, at 07:58, EigenPro ***@***.***> wrote: @EigenPro requested changes on this pull request. Thanks Lukas! This neat modification is compatible with Python 2.7. So I would like to just merge it. But could you revert the change for the assert (L 40) in this commit.? I want to keep this for a while to avoid the relatively disappointing performance on CPU. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (review)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACM4F64-bHhVlBzDbc_n47eJQu4pTFaeks5s-4EjgaJpZM4Q1-on>.

file merged