ENH: stats.entropy, special.{entr, rel_entr}: add array API support #20673
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mdhaber
commented
May 9, 2024
| # Test for PR-479 | ||
| assert_almost_equal(stats.entropy([0, 1, 2]), 0.63651416829481278, |
There was a problem hiding this comment.
Every line is changed, so I suppose git can't tell, but it's a very straightforward conversion. Consider using a diff tool locally.
|
CI failure is just mypy complaints. LMK what else is needed and I'll silence them at the same time. |
lucascolley
approved these changes
May 11, 2024
| pk = xp.asarray([[0.1, 0.2], [0.6, 0.3], [0.3, 0.5]]) | ||
| qk = xp.asarray([[0.2, 0.1], [0.3, 0.6], [0.5, 0.3]]) | ||
| xp_assert_close(stats.entropy(pk, qk, axis=1), | ||
| xp.asarray([0.23104906, 0.23104906, 0.12770641])) |
There was a problem hiding this comment.
I'm seeing new GPU CI failures locally here and in a few other places. I'll just paste the output below for inspection since I'm about to head out (you can ignore the baseline FFT thread fails).
_______________________________________________________________________________ TestFFTThreadSafe.test_fft[torch] ________________________________________________________________________________
[gw6] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/fft/tests/test_basic.py:436: in test_fft
self._test_mtsame(fft.fft, a, xp=xp)
a = tensor([[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j,...+0.j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j]],
device='cuda:0', dtype=torch.complex128)
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d616ea10>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/fft/tests/test_basic.py:430: in _test_mtsame
q.get(timeout=5), expected,
args = (tensor([[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j....j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j]],
device='cuda:0', dtype=torch.complex128),)
expected = tensor([[200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[200.+0.j, 0.+0.j, 0.+0.j, ..... [200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]],
device='cuda:0', dtype=torch.complex128)
func = <uarray multimethod 'fft'>
i = 15
q = <queue.Queue object at 0x7fb5c6baff50>
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d616ea10>
t = [<Thread(Thread-33 (worker), stopped 140418285024832)>, <Thread(Thread-34 (worker), stopped 140418531137088)>, <Thread...8)>, <Thread(Thread-37 (worker), stopped 140418285024832)>, <Thread(Thread-38 (worker), stopped 140418531137088)>, ...]
worker = <function TestFFTThreadSafe._test_mtsame.<locals>.worker at 0x7fb5be284ae0>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
../../../../../spack/opt/spack/linux-ubuntu22.04-skylake/gcc-11.3.0/python-3.11.2-4qncg2nqev5evxfmamde3e6rnb34b4ls/lib/python3.11/queue.py:179: in get
raise Empty
E _queue.Empty
block = True
endtime = 6496614.722814094
remaining = -0.00012497138231992722
self = <queue.Queue object at 0x7fb5c6baff50>
timeout = 5
During handling of the above exception, another exception occurred:
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/_pytest/runner.py:340: in from_call
result: Optional[TResult] = func()
cls = <class '_pytest.runner.CallInfo'>
duration = 5.113782258704305
excinfo = <ExceptionInfo PytestUnhandledThreadExceptionWarning('Exception in thread Thread-33 (worker)\n\nTraceback (most recent... to set the primary context... (Triggered internally at ../aten/src/ATen/native/cuda/SpectralOps.cpp:313.)\n') tblen=9>
func = <function call_and_report.<locals>.<lambda> at 0x7fb5be284fe0>
precise_start = 6496609.609832196
precise_stop = 6496614.723614454
reraise = (<class '_pytest.outcomes.Exit'>, <class 'KeyboardInterrupt'>)
result = None
start = 1715465665.421142
stop = 1715465670.5349245
when = 'call'
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/_pytest/runner.py:240: in <lambda>
lambda: runtest_hook(item=item, **kwds), when=when, reraise=reraise
item = <Function test_fft[torch]>
kwds = {}
runtest_hook = <HookCaller 'pytest_runtest_call'>
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/pluggy/_hooks.py:501: in __call__
return self._hookexec(self.name, self._hookimpls.copy(), kwargs, firstresult)
firstresult = False
kwargs = {'item': <Function test_fft[torch]>}
self = <HookCaller 'pytest_runtest_call'>
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/pluggy/_manager.py:119: in _hookexec
return self._inner_hookexec(hook_name, methods, kwargs, firstresult)
firstresult = False
hook_name = 'pytest_runtest_call'
kwargs = {'item': <Function test_fft[torch]>}
methods = [<HookImpl plugin_name='runner', plugin=<module '_pytest.runner' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/...=None>>, <HookImpl plugin_name='logging-plugin', plugin=<_pytest.logging.LoggingPlugin object at 0x7fb5e4e21dd0>>, ...]
self = <_pytest.config.PytestPluginManager object at 0x7fb6ec2d5510>
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/_pytest/threadexception.py:87: in pytest_runtest_call
yield from thread_exception_runtest_hook()
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/_pytest/threadexception.py:77: in thread_exception_runtest_hook
warnings.warn(pytest.PytestUnhandledThreadExceptionWarning(msg))
E pytest.PytestUnhandledThreadExceptionWarning: Exception in thread Thread-33 (worker)
E
E Traceback (most recent call last):
E File "/home/treddy/github_projects/spack/opt/spack/linux-ubuntu22.04-skylake/gcc-11.3.0/python-3.11.2-4qncg2nqev5evxfmamde3e6rnb34b4ls/lib/python3.11/threading.py", line 1038, in _bootstrap_inner
E self.run()
E File "/home/treddy/github_projects/spack/opt/spack/linux-ubuntu22.04-skylake/gcc-11.3.0/python-3.11.2-4qncg2nqev5evxfmamde3e6rnb34b4ls/lib/python3.11/threading.py", line 975, in run
E self._target(*self._args, **self._kwargs)
E File "/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/fft/tests/test_basic.py", line 415, in worker
E q.put(func(*args))
E ^^^^^^^^^^^
E File "/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/fft/_backend.py", line 28, in __ua_function__
E return fn(*args, **kwargs)
E ^^^^^^^^^^^^^^^^^^^
E File "/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/fft/_basic_backend.py", line 60, in fft
E return _execute_1D('fft', _pocketfft.fft, x, n=n, axis=axis, norm=norm,
E ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
E File "/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/fft/_basic_backend.py", line 34, in _execute_1D
E return xp_func(x, n=n, axis=axis, norm=norm)
E ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
E UserWarning: Attempting to run cuFFT, but there was no current CUDA context! Attempting to set the primary context... (Triggered internally at ../aten/src/ATen/native/cuda/SpectralOps.cpp:313.)
cm = <_pytest.threadexception.catch_threading_exception object at 0x7fb5d6589790>
msg = 'Exception in thread Thread-33 (worker)\n\nTraceback (most recent call last):\n File "/home/treddy/github_projects/sp...Attempting to set the primary context... (Triggered internally at ../aten/src/ATen/native/cuda/SpectralOps.cpp:313.)\n'
thread_name = 'Thread-33 (worker)'
________________________________________________________________________________ TestFFTThreadSafe.test_ifft[cupy] ________________________________________________________________________________
[gw6] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/fft/tests/test_basic.py:440: in test_ifft
self._test_mtsame(fft.ifft, a, xp=xp)
a = array([[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+... [1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j]])
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d6174590>
xp = <module 'cupy' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/__init__.py'>
scipy/fft/tests/test_basic.py:429: in _test_mtsame
xp_assert_equal(
args = (array([[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1....[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j],
[1.+0.j, 1.+0.j, 1.+0.j, ..., 1.+0.j, 1.+0.j, 1.+0.j]]),)
expected = array([[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+... [1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]])
func = <uarray multimethod 'ifft'>
i = 13
q = <queue.Queue object at 0x7fb5bdf1c490>
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d6174590>
t = [<Thread(Thread-113 (worker), stopped 140418531137088)>, <Thread(Thread-114 (worker), stopped 140418293425728)>, <Thre...>, <Thread(Thread-117 (worker), stopped 140417737684544)>, <Thread(Thread-118 (worker), stopped 140417729291840)>, ...]
worker = <function TestFFTThreadSafe._test_mtsame.<locals>.worker at 0x7fb5bff2bf60>
xp = <module 'cupy' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/__init__.py'>
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/testing/_array.py:95: in assert_array_equal
numpy.testing.assert_array_equal(
err_msg = 'Function returned wrong value in multithreaded context'
kwargs = {}
strides_check = False
verbose = True
x = array([[0.005+0.j, 0.005+0.j, 0.005+0.j, ..., 0.005+0.j, 0.005+0.j,
0.005+0.j],
[0.005+0.j, 0.005+0.j, ...0.005+0.j,
0.005+0.j],
[0.005+0.j, 0.005+0.j, 0.005+0.j, ..., 0.005+0.j, 0.005+0.j,
0.005+0.j]])
y = array([[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+... [1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]])
../../../../../spack/opt/spack/linux-ubuntu22.04-skylake/gcc-11.3.0/python-3.11.2-4qncg2nqev5evxfmamde3e6rnb34b4ls/lib/python3.11/contextlib.py:81: in inner
return func(*args, **kwds)
E AssertionError:
E Arrays are not equal
E Function returned wrong value in multithreaded context
E Mismatched elements: 160000 / 160000 (100%)
E Max absolute difference: 0.995
E Max relative difference: 0.995
E x: array([[0.005+0.j, 0.005+0.j, 0.005+0.j, ..., 0.005+0.j, 0.005+0.j,
E 0.005+0.j],
E [0.005+0.j, 0.005+0.j, 0.005+0.j, ..., 0.005+0.j, 0.005+0.j,...
E y: array([[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
E [1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
E [1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],...
args = (<built-in function eq>, array([[0.005+0.j, 0.005+0.j, 0.005+0.j, ..., 0.005+0.j, 0.005+0.j,
0.005+0.j],
... [1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]]))
func = <function assert_array_compare at 0x7fb6e3f31f80>
kwds = {'err_msg': 'Function returned wrong value in multithreaded context', 'header': 'Arrays are not equal', 'strict': False, 'verbose': True}
self = <contextlib._GeneratorContextManager object at 0x7fb6e3f51c90>
________________________________________________________________________________ TestFFTThreadSafe.test_rfft[cupy] ________________________________________________________________________________
[gw6] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/fft/tests/test_basic.py:444: in test_rfft
self._test_mtsame(fft.rfft, a, xp=xp)
a = array([[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
...,
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.]])
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d6175d90>
xp = <module 'cupy' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/__init__.py'>
scipy/fft/tests/test_basic.py:429: in _test_mtsame
xp_assert_equal(
args = (array([[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
....,
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.]]),)
expected = array([[200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[200.+0.j, 0.+0.j, 0.+0.j, ..., ... 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]])
func = <uarray multimethod 'rfft'>
i = 0
q = <queue.Queue object at 0x7fb5c4832d10>
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d6175d90>
t = [<Thread(Thread-177 (worker), stopped 140418531137088)>, <Thread(Thread-178 (worker), stopped 140418285033024)>, <Thre...>, <Thread(Thread-181 (worker), stopped 140418531137088)>, <Thread(Thread-182 (worker), stopped 140418285033024)>, ...]
worker = <function TestFFTThreadSafe._test_mtsame.<locals>.worker at 0x7fb5be0fbf60>
xp = <module 'cupy' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/__init__.py'>
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/testing/_array.py:95: in assert_array_equal
numpy.testing.assert_array_equal(
err_msg = 'Function returned wrong value in multithreaded context'
kwargs = {}
strides_check = False
verbose = True
x = array([[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[0.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 1.+0.j, 0.+... [0.+0.j, 0.+0.j, 1.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[0.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]])
y = array([[200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[200.+0.j, 0.+0.j, 0.+0.j, ..., ... 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]])
../../../../../spack/opt/spack/linux-ubuntu22.04-skylake/gcc-11.3.0/python-3.11.2-4qncg2nqev5evxfmamde3e6rnb34b4ls/lib/python3.11/contextlib.py:81: in inner
return func(*args, **kwds)
E AssertionError:
E Arrays are not equal
E Function returned wrong value in multithreaded context
E Mismatched elements: 1200 / 80800 (1.49%)
E Max absolute difference: 200.
E Max relative difference: 1.
E x: array([[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
E [0.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 1.+0.j, 0.+0.j],
E [0.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],...
E y: array([[200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
E [200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
E [200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],...
args = (<built-in function eq>, array([[1.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[0.+0.j, 0.+0.j, 0.+0.j,...0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j],
[200.+0.j, 0.+0.j, 0.+0.j, ..., 0.+0.j, 0.+0.j, 0.+0.j]]))
func = <function assert_array_compare at 0x7fb6e3f31f80>
kwds = {'err_msg': 'Function returned wrong value in multithreaded context', 'header': 'Arrays are not equal', 'strict': False, 'verbose': True}
self = <contextlib._GeneratorContextManager object at 0x7fb6e3f51c90>
_________________________________________________________________ TestEntropy.test_entropy_with_axis_0_is_equal_to_default[torch] _________________________________________________________________
[gw26] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:91: in test_entropy_with_axis_0_is_equal_to_default
xp_assert_close(stats.entropy(pk, qk, axis=0),
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0')
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f6034d7d710>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f6042f80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f6042f809a0>
kwd_samples = []
kwds = {'axis': 0}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f6042f2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f6042f80900>
temp = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
tuple_to_result = <function <lambda> at 0x7f6042f2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = 0
base = None
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0')
sum_kwargs = {'axis': 0, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7f6032c64e00>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
array_args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f6035634a60>
arg = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f60592394d0>
overloaded_args = [tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7f6108994cc0>
relevant_args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
func = <function Tensor.__array__ at 0x7f6108994cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f60592394d0>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
____________________________________________________________________________ TestEntropy.test_entropy_positive[torch] _____________________________________________________________________________
[gw27] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:18: in test_entropy_positive
eself = stats.entropy(pk, pk)
pk = tensor([0.5000, 0.2000, 0.3000], device='cuda:0')
qk = tensor([0.1000, 0.2500, 0.6500], device='cuda:0')
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f65d83d4490>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([0.5000, 0.2000, 0.3000], device='cuda:0'), tensor([0.5000, 0.2000, 0.3000], device='cuda:0'))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f65e6c80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f65e6c809a0>
kwd_samples = []
kwds = {}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f65e6c2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f65e6c80900>
temp = tensor([0.5000, 0.2000, 0.3000], device='cuda:0')
tuple_to_result = <function <lambda> at 0x7f65e6c2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = 0
base = None
pk = tensor([0.5000, 0.2000, 0.3000], device='cuda:0')
qk = tensor([0.5000, 0.2000, 0.3000], device='cuda:0')
sum_kwargs = {'axis': 0, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([0.5000, 0.2000, 0.3000], device='cuda:0'), tensor([0.5000, 0.2000, 0.3000], device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7f65d6f71300>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([0.5000, 0.2000, 0.3000], device='cuda:0'), tensor([0.5000, 0.2000, 0.3000], device='cuda:0'))
array_args = (tensor([0.5000, 0.2000, 0.3000], device='cuda:0'), tensor([0.5000, 0.2000, 0.3000], device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f65d643c5b0>
arg = tensor([0.5000, 0.2000, 0.3000], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([0.5000, 0.2000, 0.3000], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([0.5000, 0.2000, 0.3000], device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
overloaded_args = [tensor([0.5000, 0.2000, 0.3000], device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7f66ac398cc0>
relevant_args = (tensor([0.5000, 0.2000, 0.3000], device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([0.5000, 0.2000, 0.3000], device='cuda:0'),)
func = <function Tensor.__array__ at 0x7f66ac398cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([0.5000, 0.2000, 0.3000], device='cuda:0')
______________________________________________________________________________ TestEntropy.test_entropy_base[torch] _______________________________________________________________________________
[gw27] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:31: in test_entropy_base
S = stats.entropy(pk, qk)
S = tensor(4., device='cuda:0')
pk = tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.],
device='cuda:0')
qk = tensor([2., 2., 2., 2., 2., 2., 2., 2., 1., 1., 1., 1., 1., 1., 1., 1.],
device='cuda:0')
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f65dbb2d350>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.],
device='cuda:0'), tensor([2., 2., 2., 2., 2., 2., 2., 2., 1., 1., 1., 1., 1., 1., 1., 1.],
device='cuda:0'))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f65e6c80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f65e6c809a0>
kwd_samples = []
kwds = {}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f65e6c2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f65e6c80900>
temp = tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.],
device='cuda:0')
tuple_to_result = <function <lambda> at 0x7f65e6c2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = 0
base = None
pk = tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0')
qk = tensor([0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.0833, 0.0417,
0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417],
device='cuda:0')
sum_kwargs = {'axis': 0, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.062...0833, 0.0833, 0.0833, 0.0417,
0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417],
device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7f65d7381bc0>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.062...0833, 0.0833, 0.0833, 0.0417,
0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417],
device='cuda:0'))
array_args = (tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.062...0833, 0.0833, 0.0833, 0.0417,
0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417, 0.0417],
device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f65d5ea2ce0>
arg = tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
overloaded_args = [tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7f66ac398cc0>
relevant_args = (tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0'),)
func = <function Tensor.__array__ at 0x7f66ac398cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625,
0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625, 0.0625],
device='cuda:0')
___________________________________________________________________________ TestEntropy.test_entropy_transposed[torch] ____________________________________________________________________________
[gw26] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:104: in test_entropy_transposed
xp_assert_close(stats.entropy(pk.T, qk.T),
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0')
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f6034d7fe10>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([[0.1000, 0.6000, 0.3000],
[0.2000, 0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.3000, 0.5000],
[0.1000, 0.6000, 0.3000]], device='cuda:0'))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f6042f80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f6042f809a0>
kwd_samples = []
kwds = {}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f6042f2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f6042f80900>
temp = tensor([[0.1000, 0.6000, 0.3000],
[0.2000, 0.3000, 0.5000]], device='cuda:0')
tuple_to_result = <function <lambda> at 0x7f6042f2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = 0
base = None
pk = tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0')
qk = tensor([[0.6667, 0.3333, 0.6250],
[0.3333, 0.6667, 0.3750]], device='cuda:0')
sum_kwargs = {'axis': 0, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0'), tensor([[0.6667, 0.3333, 0.6250],
[0.3333, 0.6667, 0.3750]], device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7f6032cf7100>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0'), tensor([[0.6667, 0.3333, 0.6250],
[0.3333, 0.6667, 0.3750]], device='cuda:0'))
array_args = (tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0'), tensor([[0.6667, 0.3333, 0.6250],
[0.3333, 0.6667, 0.3750]], device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f6032f05f60>
arg = tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f60592394d0>
overloaded_args = [tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7f6108994cc0>
relevant_args = (tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0'),)
func = <function Tensor.__array__ at 0x7f6108994cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f60592394d0>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([[0.3333, 0.6667, 0.3750],
[0.6667, 0.3333, 0.6250]], device='cuda:0')
__________________________________________________________________________ TestEntropy.test_entropy_broadcasting[torch] ___________________________________________________________________________
[gw26] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:112: in test_entropy_broadcasting
res = stats.entropy(x, y, axis=-1)
rng = Generator(PCG64) at 0x7F60334110E0
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f6034d88f90>
x = tensor([0.1155, 0.4569, 0.2974], device='cuda:0', dtype=torch.float64)
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
y = tensor([[0.4619],
[0.4314]], device='cuda:0', dtype=torch.float64)
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([0.1155, 0.4569, 0.2974], device='cuda:0', dtype=torch.float64), tensor([[0.4619],
[0.4314]], device='cuda:0', dtype=torch.float64))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f6042f80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f6042f809a0>
kwd_samples = []
kwds = {'axis': -1}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f6042f2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f6042f80900>
temp = tensor([0.1155, 0.4569, 0.2974], device='cuda:0', dtype=torch.float64)
tuple_to_result = <function <lambda> at 0x7f6042f2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = -1
base = None
pk = tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64)
qk = tensor([[0.3333, 0.3333, 0.3333],
[0.3333, 0.3333, 0.3333]], device='cuda:0', dtype=torch.float64)
sum_kwargs = {'axis': -1, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64), tensor([[0.3333, 0.3333, 0.3333],
[0.3333, 0.3333, 0.3333]], device='cuda:0', dtype=torch.float64))
f = <function get_array_special_func.<locals>.f at 0x7f6032cf0400>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64), tensor([[0.3333, 0.3333, 0.3333],
[0.3333, 0.3333, 0.3333]], device='cuda:0', dtype=torch.float64))
array_args = (tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64), tensor([[0.3333, 0.3333, 0.3333],
[0.3333, 0.3333, 0.3333]], device='cuda:0', dtype=torch.float64))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f603529af80>
arg = tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f60592394d0>
overloaded_args = [tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64)]
public_api = <function Tensor.__array__ at 0x7f6108994cc0>
relevant_args = (tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64),)
func = <function Tensor.__array__ at 0x7f6108994cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f60592394d0>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([[0.1328, 0.5253, 0.3419],
[0.1328, 0.5253, 0.3419]], device='cuda:0', dtype=torch.float64)
_______________________________________________________________________________ TestEntropy.test_entropy_2d[torch] ________________________________________________________________________________
[gw27] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:46: in test_entropy_2d
xp_assert_close(stats.entropy(pk, qk),
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0')
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f65d8276450>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f65e6c80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f65e6c809a0>
kwd_samples = []
kwds = {}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f65e6c2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f65e6c80900>
temp = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
tuple_to_result = <function <lambda> at 0x7f65e6c2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = 0
base = None
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0')
sum_kwargs = {'axis': 0, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7f65d6ee6520>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
array_args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f65e7b244f0>
arg = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
overloaded_args = [tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7f66ac398cc0>
relevant_args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
func = <function Tensor.__array__ at 0x7f66ac398cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
_____________________________________________________________________________ TestEntropy.test_entropy_2d_zero[torch] _____________________________________________________________________________
[gw27] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:53: in test_entropy_2d_zero
xp_assert_close(stats.entropy(pk, qk),
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.0000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0')
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f65d8277710>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.0000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f65e6c80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f65e6c809a0>
kwd_samples = []
kwds = {}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f65e6c2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f65e6c80900>
temp = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
tuple_to_result = <function <lambda> at 0x7f65e6c2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = 0
base = None
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.0000, 0.1000],
[0.3750, 0.6000],
[0.6250, 0.3000]], device='cuda:0')
sum_kwargs = {'axis': 0, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.0000, 0.1000],
[0.3750, 0.6000],
[0.6250, 0.3000]], device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7f65d705a0c0>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.0000, 0.1000],
[0.3750, 0.6000],
[0.6250, 0.3000]], device='cuda:0'))
array_args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.0000, 0.1000],
[0.3750, 0.6000],
[0.6250, 0.3000]], device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f65db861360>
arg = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
overloaded_args = [tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7f66ac398cc0>
relevant_args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'),)
func = <function Tensor.__array__ at 0x7f66ac398cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
_______________________________________________________________________ TestEntropy.test_entropy_2d_nondefault_axis[torch] ________________________________________________________________________
[gw27] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/stats/tests/test_entropy.py:70: in test_entropy_2d_nondefault_axis
xp_assert_close(stats.entropy(pk, qk, axis=1),
pk = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
qk = tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0')
self = <scipy.stats.tests.test_entropy.TestEntropy object at 0x7f65d82a8a10>
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/stats/_axis_nan_policy.py:407: in axis_nan_policy_wrapper
return hypotest_fun_in(*args, **kwds)
_no_deco = False
args = (tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0'), tensor([[0.2000, 0.1000],
[0.3000, 0.6000],
[0.5000, 0.3000]], device='cuda:0'))
default_axis = 0
hypotest_fun_in = <function entropy at 0x7f65e6c80ae0>
is_too_small = <function _axis_nan_policy_factory.<locals>.is_too_small at 0x7f65e6c809a0>
kwd_samples = []
kwds = {'axis': 1}
msg = 'Use of `nan_policy` and `keepdims` is incompatible with non-NumPy arrays.'
n_outputs = 1
n_samples = <function <lambda> at 0x7f65e6c2c7c0>
override = {'nan_propagation': True, 'vectorization': False}
paired = True
result_to_tuple = <function <lambda> at 0x7f65e6c80900>
temp = tensor([[0.1000, 0.2000],
[0.6000, 0.3000],
[0.3000, 0.5000]], device='cuda:0')
tuple_to_result = <function <lambda> at 0x7f65e6c2c5e0>
scipy/stats/_entropy.py:155: in entropy
vec = special.rel_entr(pk, qk)
axis = 1
base = None
pk = tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0')
qk = tensor([[0.6667, 0.3333],
[0.3333, 0.6667],
[0.6250, 0.3750]], device='cuda:0')
sum_kwargs = {'axis': 1, 'keepdims': True}
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0'), tensor([[0.6667, 0.3333],
[0.3333, 0.6667],
[0.6250, 0.3750]], device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7f65d75ec900>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0'), tensor([[0.6667, 0.3333],
[0.3333, 0.6667],
[0.6250, 0.3750]], device='cuda:0'))
array_args = (tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0'), tensor([[0.6667, 0.3333],
[0.3333, 0.6667],
[0.6250, 0.3750]], device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7f65e158b850>
arg = tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
overloaded_args = [tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7f66ac398cc0>
relevant_args = (tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0'),)
func = <function Tensor.__array__ at 0x7f66ac398cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7f66fbf56b50>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
dtype = None
self = tensor([[0.3333, 0.6667],
[0.6667, 0.3333],
[0.3750, 0.6250]], device='cuda:0')
_______________________________________________________________________________ TestFFTThreadSafe.test_ihfft[cupy] ________________________________________________________________________________
[gw6] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/fft/tests/test_basic.py:456: in test_ihfft
self._test_mtsame(fft.ihfft, a, xp=xp)
a = array([[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
...,
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.]])
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d617a5d0>
xp = <module 'cupy' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/__init__.py'>
scipy/fft/tests/test_basic.py:429: in _test_mtsame
xp_assert_equal(
args = (array([[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
....,
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.],
[1., 1., 1., ..., 1., 1., 1.]]),)
expected = array([[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],
[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-... [1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],
[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j]])
func = <uarray multimethod 'ihfft'>
i = 0
q = <queue.Queue object at 0x7fb5bf082e50>
self = <scipy.fft.tests.test_basic.TestFFTThreadSafe object at 0x7fb5d617a5d0>
t = [<Thread(Thread-369 (worker), stopped 140417647507008)>, <Thread(Thread-370 (worker), stopped 140417630721600)>, <Thre...>, <Thread(Thread-373 (worker), stopped 140418531137088)>, <Thread(Thread-374 (worker), stopped 140418293425728)>, ...]
worker = <function TestFFTThreadSafe._test_mtsame.<locals>.worker at 0x7fb5be0f9c60>
xp = <module 'cupy' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/__init__.py'>
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/cupy/testing/_array.py:95: in assert_array_equal
numpy.testing.assert_array_equal(
err_msg = 'Function returned wrong value in multithreaded context'
kwargs = {}
strides_check = False
verbose = True
x = array([[0.005-0.j, 0. -0.j, 0. -0.j, ..., 0. -0.j, 0. -0.j,
0. -0.j],
[0.005-0.j, 0. -0.j, ...0. -0.j,
0. -0.j],
[0.005-0.j, 0. -0.j, 0. -0.j, ..., 0. -0.j, 0. -0.j,
0. -0.j]])
y = array([[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],
[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-... [1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],
[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j]])
../../../../../spack/opt/spack/linux-ubuntu22.04-skylake/gcc-11.3.0/python-3.11.2-4qncg2nqev5evxfmamde3e6rnb34b4ls/lib/python3.11/contextlib.py:81: in inner
return func(*args, **kwds)
E AssertionError:
E Arrays are not equal
E Function returned wrong value in multithreaded context
E Mismatched elements: 800 / 80800 (0.99%)
E Max absolute difference: 0.995
E Max relative difference: 0.995
E x: array([[0.005-0.j, 0. -0.j, 0. -0.j, ..., 0. -0.j, 0. -0.j,
E 0. -0.j],
E [0.005-0.j, 0. -0.j, 0. -0.j, ..., 0. -0.j, 0. -0.j,...
E y: array([[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],
E [1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],
E [1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],...
args = (<built-in function eq>, array([[0.005-0.j, 0. -0.j, 0. -0.j, ..., 0. -0.j, 0. -0.j,
0. -0.j],
... [1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j],
[1.-0.j, 0.-0.j, 0.-0.j, ..., 0.-0.j, 0.-0.j, 0.-0.j]]))
func = <function assert_array_compare at 0x7fb6e3f31f80>
kwds = {'err_msg': 'Function returned wrong value in multithreaded context', 'header': 'Arrays are not equal', 'strict': False, 'verbose': True}
self = <contextlib._GeneratorContextManager object at 0x7fb6e3f51c90>
____________________________________________________________________ test_support_alternative_backends[f_name_n_args15-torch] _____________________________________________________________________
[gw5] linux -- Python 3.11.2 /home/treddy/python_venvs/py_311_scipy_dev/bin/python
scipy/special/tests/test_support_alternative_backends.py:31: in test_support_alternative_backends
@given(data=strategies.data())
f = <function given.<locals>.run_test_as_given.<locals>.wrapped_test at 0x7fce7d977ba0>
f_name_n_args = ('rel_entr', 2)
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/special/tests/test_support_alternative_backends.py:55: in test_support_alternative_backends
res = f(*args_xp)
args_np = [array(0., dtype=float32), array(0., dtype=float32)]
args_xp = [tensor(0., device='cuda:0'), tensor(0., device='cuda:0')]
data = data(...)
dtype = 'float32'
dtype_np = <class 'numpy.float32'>
dtype_xp = torch.float32
elements = {'allow_subnormal': False, 'max_value': 10.0, 'min_value': -10.0}
f = <function support_alternative_backends.<locals>.wrapped at 0x7fce9e5d84a0>
f_name = 'rel_entr'
f_name_n_args = ('rel_entr', 2)
final_shape = ()
mbs = mutually_broadcastable_shapes(num_shapes=2)
n_args = 2
ref = array(0., dtype=float32)
shapes = ((), ())
xp = <module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:50: in wrapped
return f(*args, **kwargs)
args = (tensor(0., device='cuda:0'), tensor(0., device='cuda:0'))
f = <function get_array_special_func.<locals>.f at 0x7fce61506520>
f_name = 'rel_entr'
kwargs = {}
n_array_args = 2
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in f
array_args = [np.asarray(arg) for arg in array_args]
args = (tensor(0., device='cuda:0'), tensor(0., device='cuda:0'))
array_args = (tensor(0., device='cuda:0'), tensor(0., device='cuda:0'))
f_scipy = <ufunc 'rel_entr'>
kwargs = {}
n_array_args = 2
other_args = ()
xp = <module 'scipy._lib.array_api_compat.torch' from '/home/treddy/github_projects/scipy/build-install/lib/python3.11/site-packages/scipy/_lib/array_api_compat/torch/__init__.py'>
scipy/special/_support_alternative_backends.py:35: in <listcomp>
array_args = [np.asarray(arg) for arg in array_args]
.0 = <tuple_iterator object at 0x7fce693c3310>
arg = tensor(0., device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1060: in __array__
return handle_torch_function(Tensor.__array__, (self,), self, dtype=dtype)
dtype = None
self = tensor(0., device='cuda:0')
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/overrides.py:1604: in handle_torch_function
result = mode.__torch_function__(public_api, types, args, kwargs)
args = (tensor(0., device='cuda:0'),)
kwargs = {'dtype': None}
mode = <torch.utils._device.DeviceContext object at 0x7fcea01310d0>
overloaded_args = [tensor(0., device='cuda:0')]
public_api = <function Tensor.__array__ at 0x7fcf4f984cc0>
relevant_args = (tensor(0., device='cuda:0'),)
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/utils/_device.py:77: in __torch_function__
return func(*args, **kwargs)
args = (tensor(0., device='cuda:0'),)
func = <function Tensor.__array__ at 0x7fcf4f984cc0>
kwargs = {'dtype': None}
self = <torch.utils._device.DeviceContext object at 0x7fcea01310d0>
types = (<class 'torch.Tensor'>,)
/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/_tensor.py:1062: in __array__
return self.numpy()
E TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
E Falsifying example: test_support_alternative_backends(
E xp=<module 'torch' from '/home/treddy/python_venvs/py_311_scipy_dev/lib/python3.11/site-packages/torch/__init__.py'>,
E f_name_n_args=('rel_entr', 2),
E data=data(...),
E )
E Draw 1: BroadcastableShapes(input_shapes=((), ()), result_shape=())
E Draw 2: 'float32'
E Draw 3: array(0., dtype=float32)
E Draw 4: array(0., dtype=float32)
E
E You can reproduce this example by temporarily adding @reproduce_failure('6.82.0', b'AXicY2DABwAAHgAB') as a decorator on your test case
dtype = None
self = tensor(0., device='cuda:0')
===================================================================================== short test summary info =====================================================================================
FAILED scipy/fft/tests/test_basic.py::TestFFTThreadSafe::test_fft[torch] - pytest.PytestUnhandledThreadExceptionWarning: Exception in thread Thread-33 (worker)
FAILED scipy/fft/tests/test_basic.py::TestFFTThreadSafe::test_ifft[cupy] - AssertionError:
FAILED scipy/fft/tests/test_basic.py::TestFFTThreadSafe::test_rfft[cupy] - AssertionError:
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_with_axis_0_is_equal_to_default[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_positive[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_base[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_transposed[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_broadcasting[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_2d[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_2d_zero[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/stats/tests/test_entropy.py::TestEntropy::test_entropy_2d_nondefault_axis[torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
FAILED scipy/fft/tests/test_basic.py::TestFFTThreadSafe::test_ihfft[cupy] - AssertionError:
FAILED scipy/special/tests/test_support_alternative_backends.py::test_support_alternative_backends[f_name_n_args15-torch] - TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
============================================================ 13 failed, 51728 passed, 11250 skipped, 157 xfailed, 13 xpassed in 58.81s ============================================================
There was a problem hiding this comment.
I can try to help iterate a bit though, if you ping me to check stuff, etc.
There was a problem hiding this comment.
Huh it looks like the _axis_nan_policy decorator is not getting skipped. It should always get an early exit. I tested locally with CuPy and didn't see this, which is surprising. I should be able to debug locally though. Thanks!
There was a problem hiding this comment.
Actually this is not unexpected; I misread before. torch doesn't have a rel_entr function, so I knew that for CPU it would evaluate that part with NumPy and then change the result back to torch. I think this is in the same category of tests that we expect to pass only on the CPU. Until we can use from_dlpack to do device transfers, we are OK with such tests failing on GPU. Can we skip torch GPU only?
There was a problem hiding this comment.
The alternative is to write a rel_entr function that relies only on array API calls. Usually we can't do that in any reasonable way for special functions, but in this case it would be pretty easy. I'd prefer for that to be a follow-up enhancement, though.
There was a problem hiding this comment.
Never mind, I think I can add it here.
There was a problem hiding this comment.
I don't have an opinion on it apart from thinking we'd prefer to skip rather than carrying the failures if it came to that.
There was a problem hiding this comment.
Could you try out the latest commit? I added a generic implementation of rel_entr.
mdhaber
commented
May 12, 2024
|
Hey Matt, I only see 2 failures left locally with |
|
That's a shortcoming of import numpy as np
from scipy import special
x = np.asarray([2., 3., 4.])
y = np.asarray(np.finfo(np.float64).tiny)
special.rel_entr(x, y) # array([1418.17913143, 2128.48509246, inf])
x * (np.log(x) - np.log(y)) # array([1418.17913143, 2128.48509246, 2839.13085157])It must be using the naive expression that takes the ratio Since it's not failing in CI, let's not take any action toward that here, but I can file an issue. Thanks @tylerjereddy! @lucascolley I think this is ready for another look. Thank you! |
lucascolley
approved these changes
May 14, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reference issue
Towards gh-20544
What does this implement/fix?
Adds array API support to
scipy.stats.entropy,special.entr,special.rel_entr.Additional information
torchdoesn't haverel_entr, so it falls back to back and forth conversion. One could implement arel_entrcalculation as an enhancement.