lasagne - 3 test failures; fixed upstream by
https://github.com/Lasagne/Lasagne/pull/836
Latest upstream (37ca134; only packaging change was to drop
remove-deprecated.patch) doesn't have these failures. There is a
warning that cuda_convnet is no longer available with Theano 1.0, but
that has
setuptools instead
of distutils as recommended by the PyPA?
Le mer. 21 mars 2018 à 20:45, Rebecca N. Palmer <rebecca_pal...@zoho.com> a
écrit :
Source: sympy
Severity: serious
Control: tags -1 patch
X-Debbugs-Cc: debian-pyt...@lists.debian.org
python3-distutils has been moved out of pyt
Source: sympy
Severity: serious
Control: tags -1 patch
X-Debbugs-Cc: debian-pyt...@lists.debian.org
python3-distutils has been moved out of python3.6 (as of 3.6.5~rc1-2),
so if you need it, please build-depend on it. (Or python3-setuptools,
given that this looks like it might prefer that.)
Source: theano
Severity: wishlist
(Posting here rather than on debian-release as the formal transition
process is rarely used, and its tools mostly useless, for non-compiled
cases.)
Theano 1.0 (currently in Salsa) contains some API-breaking changes:
Package: python3-xarray
Version: 0.9.2-1
Severity: serious
Control: tags -1 upstream
Two netcdf tests fail in current sid (see log below).
This is known upstream as https://github.com/pydata/xarray/issues/1721 ,
according to which the actual problem is that scipy has been writing
netcdf files
Control: tags -1 upstream fixed-upstream
This appears to be 2 separate bugs, both triggered by using pandas 0.20
and fixed upstream.
groupby_bins:
https://github.com/pydata/xarray/issues/1386
https://github.com/pydata/xarray/pull/1390
test_sel: this is really a pandas bug (fixed in 0.21, but
Control: tags -1 patch fixed-upstream pending
Ready for upload (the GPU tests again haven't been run, but this
shouldn't touch those parts).
--
debian-science-maintainers mailing list
debian-science-maintainers@lists.alioth.debian.org
raise nose.SkipTest("known failure of test_stata on non-little endian")
E NameError: name 'nose' is not defined
You need an 'import nose' first, if the test doesn't already have one.
--
debian-science-maintainers mailing list
debian-science-maintainers@lists.alioth.debian.org
Package: python-theano
Version: 0.9.0+dfsg-1
Severity: serious
Suspect the problem is theano/theano/tensor/signal/pool.py:650, which
effectively does int32 = *(int64 *)(pointer_to_some_int_type) - if
some_int_type is int32, that works on little-endian but not big-endian.
--
I intend to file this upstream after investigating further (with a patch
if I can); the main purpose of this Debian bug is to explain why I can't
fully test the theano package I recently pushed.
--
debian-science-maintainers mailing list
debian-science-maintainers@lists.alioth.debian.org
Package: libclblas2
Version: 2.12-1
Control: tags -1 upstream
Control: affects -1 beignet-opencl-icd
Some clblas operations use '0.0' (a double-precision literal) not '0.0f'
(a single-precision literal) even when processing single-precision arrays.
This causes it to crash on GPUs that don't
buildds:
don't let it fail the whole build.
Author: Rebecca N. Palmer <rebecca_pal...@zoho.com>
Bug-Debian: https://bugs.debian.org/835531
Forwarded: not-needed
diff --git a/theano/sparse/tests/test_basic.py b/theano/sparse/tests/test_basic.py
index 8c183b9..03d79f1 100644
--- a/theano/sparse
and negative stride handling
Cast values, not pointers, from int64 to int32.
Remember that first-in-index order (numpy) and
first-in-memory-order (BLAS) are not always the same thing.
Bump c_code_cache_version to make sure existing installs use the fixes.
Author: Rebecca N. Palmer <rebecca_
Package: theano
Version: 0.8.2-3
Severity: important
Control: tags -1 upstream
Because Numpy arrays' data pointer is the first element in *index* order
but BLAS expects the first element in *memory* order, simply calling a
BLAS function as in SamplingDotCSR
Found the problem: the affected functions (
https://sources.debian.net/src/theano/0.8.2-4/theano/sparse/opt.py/#L862
,
https://sources.debian.net/src/theano/0.8.2-4/theano/sparse/opt.py/#L1902
) cast a pointer-to-intptr_t (64 bit) to a pointer-to-int (32-bit).
Which isn't just broken on
Control: tags -1 patch
Three patches because this is logically three bugs, though it's filed as
two upstream:
https://github.com/Theano/Theano/issues/5494
https://github.com/Theano/Theano/issues/5396
First patch taken from upstream
Control: forwarded -1 https://github.com/Theano/Theano/issues/5498
Control: tags -1 patch
I don't think this is a regression - it's Python 3 specific
(numpy.array(list of longs, which this test uses on Python 2)=int64
array, but numpy(list of Python 3 ints)=int_nativesize array; see above
libopencv-dev doesn't pull in the Java libraries; I don't know if the
appropriate fix is that it should, or that the cmake script shouldn't be
looking for them when building C(++).
--
debian-science-maintainers mailing list
debian-science-maintainers@lists.alioth.debian.org
18 matches
Mail list logo