GNU bug report logs - #39562
python-keras build fails: test_selu: Not equal to tolerance rtol=1e-07, atol=0

Previous Next

Package: guix;

Reported by: Pierre Neidhardt <mail <at> ambrevar.xyz>

Date: Tue, 11 Feb 2020 14:42:02 UTC

Severity: normal

Done: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>

Bug is archived. No further changes may be made.

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 39562 in the body.
You can then email your comments to 39562 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

View this report as an mbox folder, status mbox, maintainer mbox


Report forwarded to bug-guix <at> gnu.org:
bug#39562; Package guix. (Tue, 11 Feb 2020 14:42:02 GMT) Full text and rfc822 format available.

Acknowledgement sent to Pierre Neidhardt <mail <at> ambrevar.xyz>:
New bug report received and forwarded. Copy sent to bug-guix <at> gnu.org. (Tue, 11 Feb 2020 14:42:03 GMT) Full text and rfc822 format available.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Pierre Neidhardt <mail <at> ambrevar.xyz>
To: bug-guix <at> gnu.org
Cc: Ricardo Wurmus <rekado <at> elephly.net>, Marius Bakke <mbakke <at> fastmail.com>
Subject: python-keras build fails: test_selu: Not equal to tolerance
 rtol=1e-07, atol=0
Date: Tue, 11 Feb 2020 11:49:45 +0100
[Message part 1 (text/plain, inline)]
Tests fail with:

--8<---------------cut here---------------start------------->8---
=================================== FAILURES ===================================
__________________________________ test_selu ___________________________________

    def test_selu():
        x = K.placeholder(ndim=2)
        f = K.function([x], [activations.selu(x)])
        alpha = 1.6732632423543772848170429916717
        scale = 1.0507009873554804934193349852946

        positive_values = get_standard_values()
        result = f([positive_values])[0]
        assert_allclose(result, positive_values * scale, rtol=1e-05)

        negative_values = np.array([[-1, -2]], dtype=K.floatx())

        result = f([negative_values])[0]
        true_result = (np.exp(negative_values) - 1) * scale * alpha

>       assert_allclose(result, true_result)
E       AssertionError:
E       Not equal to tolerance rtol=1e-07, atol=0
E
E       Mismatch: 50%
E       Max absolute difference: 1.1920929e-07
E       Max relative difference: 1.0726715e-07
E        x: array([[-1.111331, -1.520167]], dtype=float32)
E        y: array([[-1.111331, -1.520167]], dtype=float32)

tests/keras/activations_test.py:226: AssertionError
--8<---------------cut here---------------end--------------->8---

See attached log.

-- 
Pierre Neidhardt
https://ambrevar.xyz/
[signature.asc (application/pgp-signature, inline)]
[keras-build.log (application/octet-stream, attachment)]

Information forwarded to bug-guix <at> gnu.org:
bug#39562; Package guix. (Tue, 11 Feb 2020 19:26:01 GMT) Full text and rfc822 format available.

Message #8 received at 39562 <at> debbugs.gnu.org (full text, mbox):

From: Gábor Boskovits <boskovits <at> gmail.com>
To: Pierre Neidhardt <mail <at> ambrevar.xyz>
Cc: 39562 <at> debbugs.gnu.org
Subject: Re: bug#39562: python-keras build fails: test_selu: Not equal to
 tolerance rtol=1e-07, atol=0
Date: Tue, 11 Feb 2020 20:25:31 +0100
[Message part 1 (text/plain, inline)]
This resembles to the babl issue we are having. Is it possible that this is
hardware related?

Pierre Neidhardt <mail <at> ambrevar.xyz> ezt írta (időpont: 2020. febr. 11., Ke
16:19):

> Tests fail with:
>
> --8<---------------cut here---------------start------------->8---
> =================================== FAILURES
> ===================================
> __________________________________ test_selu
> ___________________________________
>
>     def test_selu():
>         x = K.placeholder(ndim=2)
>         f = K.function([x], [activations.selu(x)])
>         alpha = 1.6732632423543772848170429916717
>         scale = 1.0507009873554804934193349852946
>
>         positive_values = get_standard_values()
>         result = f([positive_values])[0]
>         assert_allclose(result, positive_values * scale, rtol=1e-05)
>
>         negative_values = np.array([[-1, -2]], dtype=K.floatx())
>
>         result = f([negative_values])[0]
>         true_result = (np.exp(negative_values) - 1) * scale * alpha
>
> >       assert_allclose(result, true_result)
> E       AssertionError:
> E       Not equal to tolerance rtol=1e-07, atol=0
> E
> E       Mismatch: 50%
> E       Max absolute difference: 1.1920929e-07
> E       Max relative difference: 1.0726715e-07
> E        x: array([[-1.111331, -1.520167]], dtype=float32)
> E        y: array([[-1.111331, -1.520167]], dtype=float32)
>
> tests/keras/activations_test.py:226: AssertionError
> --8<---------------cut here---------------end--------------->8---
>
> See attached log.
>
> --
> Pierre Neidhardt
> https://ambrevar.xyz/
>
[Message part 2 (text/html, inline)]

Information forwarded to bug-guix <at> gnu.org:
bug#39562; Package guix. (Tue, 03 Mar 2020 09:25:02 GMT) Full text and rfc822 format available.

Message #11 received at 39562 <at> debbugs.gnu.org (full text, mbox):

From: Pierre Neidhardt <mail <at> ambrevar.xyz>
To: Gábor Boskovits <boskovits <at> gmail.com>
Cc: 39562 <at> debbugs.gnu.org
Subject: Re: bug#39562: python-keras build fails: test_selu: Not equal to
 tolerance rtol=1e-07, atol=0
Date: Tue, 03 Mar 2020 10:24:00 +0100
[Message part 1 (text/plain, inline)]
Patch sent: 39880 <at> debbugs.gnu.org.
Also see https://github.com/keras-team/keras/issues/13856.

-- 
Pierre Neidhardt
https://ambrevar.xyz/
[signature.asc (application/pgp-signature, inline)]

Reply sent to Maxim Cournoyer <maxim.cournoyer <at> gmail.com>:
You have taken responsibility. (Mon, 05 Oct 2020 05:53:02 GMT) Full text and rfc822 format available.

Notification sent to Pierre Neidhardt <mail <at> ambrevar.xyz>:
bug acknowledged by developer. (Mon, 05 Oct 2020 05:53:03 GMT) Full text and rfc822 format available.

Message #16 received at 39562-done <at> debbugs.gnu.org (full text, mbox):

From: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
To: Pierre Neidhardt <mail <at> ambrevar.xyz>
Cc: Gábor Boskovits <boskovits <at> gmail.com>,
 39562-done <at> debbugs.gnu.org
Subject: Re: bug#39562: python-keras build fails: test_selu: Not equal to
 tolerance rtol=1e-07, atol=0
Date: Mon, 05 Oct 2020 01:52:34 -0400
Pierre Neidhardt <mail <at> ambrevar.xyz> writes:

> Patch sent: 39880 <at> debbugs.gnu.org.
> Also see https://github.com/keras-team/keras/issues/13856.

I fixed it with a substitute* instead, and also applied another upstream
fix in commit 5f37d7e38e.

Thanks for the report!

Closing,

Maxim




Information forwarded to bug-guix <at> gnu.org:
bug#39562; Package guix. (Mon, 05 Oct 2020 08:32:01 GMT) Full text and rfc822 format available.

Message #19 received at 39562-done <at> debbugs.gnu.org (full text, mbox):

From: Pierre Neidhardt <mail <at> ambrevar.xyz>
To: Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Cc: Gábor Boskovits <boskovits <at> gmail.com>,
 39562-done <at> debbugs.gnu.org
Subject: Re: bug#39562: python-keras build fails: test_selu: Not equal to
 tolerance rtol=1e-07, atol=0
Date: Mon, 05 Oct 2020 10:31:48 +0200
[Message part 1 (text/plain, inline)]
Thanks!

-- 
Pierre Neidhardt
https://ambrevar.xyz/
[signature.asc (application/pgp-signature, inline)]

bug archived. Request was from Debbugs Internal Request <help-debbugs <at> gnu.org> to internal_control <at> debbugs.gnu.org. (Mon, 02 Nov 2020 12:24:06 GMT) Full text and rfc822 format available.

This bug report was last modified 3 years and 147 days ago.

Previous Next


GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.