Repository navigation
MKL unexpected behaviour in numpy.linalg.inv returning different values in successive calls with same input data #11023
Description
Checklist
- I added a descriptive title
- I searched for other issues and couldn't find a solution or duplication
- I already searched in Google and didn't find any good information or help
- I looked at the docs and didn't see anything to help
What happened?
MKL unexpected behavior in numpy.linalg.inv returning different values in successive calls with same input data
When calling numpy.linalg.inv in a conda environment with numpy installed via conda, I get a matrix with some nan elements.
Besides that, immediate call to the same function with the same input data returns a non nan matrix.
If I test the same script in an environment where numpy is installed via pip this doesn't happen.
Issue created for Numpy in: numpy/numpy#20233
Conda info
No response
Conda config
conda config --show
add_anaconda_token: True
add_pip_as_python_dependency: True
aggressive_update_packages:
- ca-certificates
- certifi
- openssl
allow_conda_downgrades: False
allow_cycles: True
allow_non_channel_urls: False
allow_softlinks: False
always_copy: False
always_softlink: False
always_yes: None
anaconda_upload: None
auto_activate_base: True
auto_stack: 0
auto_update_conda: True
bld_path:
changeps1: True
channel_alias: https://conda.anaconda.org
channel_priority: flexible
channels:
- defaults
client_ssl_cert: None
client_ssl_cert_key: None
clobber: False
conda_build: {}
create_default_packages: []
croot: C:\Users\lbieczynski\conda-bld
custom_channels:
pkgs/main: https://repo.anaconda.com
pkgs/r: https://repo.anaconda.com
pkgs/msys2: https://repo.anaconda.com
pkgs/pro: https://repo.anaconda.com
custom_multichannels:
defaults:
- https://repo.anaconda.com/pkgs/main
- https://repo.anaconda.com/pkgs/r
- https://repo.anaconda.com/pkgs/msys2
local:
debug: False
default_channels:
- https://repo.anaconda.com/pkgs/main
- https://repo.anaconda.com/pkgs/r
- https://repo.anaconda.com/pkgs/msys2
default_python: 3.8
default_threads: None
deps_modifier: not_set
dev: False
disallowed_packages: []
download_only: False
dry_run: False
enable_private_envs: False
env_prompt: ({default_env})
envs_dirs:
- C:\Users\lbieczynski\.conda\envs
- C:\ProgramData\Anaconda3\envs
- C:\Users\lbieczynski\AppData\Local\conda\conda\envs
error_upload_url: https://conda.io/conda-post/unexpected-error
execute_threads: 1
extra_safety_checks: False
force: False
force_32bit: False
force_reinstall: False
force_remove: False
ignore_pinned: False
json: False
local_repodata_ttl: 1
migrated_channel_aliases: []
migrated_custom_channels: {}
non_admin_enabled: True
notify_outdated_conda: True
offline: False
override_channels_enabled: True
path_conflict: clobber
pinned_packages: []
pip_interop_enabled: False
pkgs_dirs:
- C:\ProgramData\Anaconda3\pkgs
- C:\Users\lbieczynski\.conda\pkgs
- C:\Users\lbieczynski\AppData\Local\conda\conda\pkgs
proxy_servers: {}
quiet: False
remote_backoff_factor: 1
remote_connect_timeout_secs: 9.15
remote_max_retries: 3
remote_read_timeout_secs: 60.0
repodata_fns:
- current_repodata.json
- repodata.json
repodata_threads: None
report_errors: None
restore_free_channel: False
rollback_enabled: True
root_prefix: C:\ProgramData\Anaconda3
safety_checks: warn
sat_solver: pycosat
separate_format_cache: False
shortcuts: True
show_channel_urls: None
signing_metadata_url_base: None
solver_ignore_timestamps: False
ssl_verify: True
subdir: win-64
subdirs:
- win-64
- noarch
target_prefix_override:
track_features: []
unsatisfiable_hints: True
unsatisfiable_hints_check_depth: 2
update_modifier: update_specs
use_index_cache: False
use_local: False
use_only_tar_bz2: False
verbosity: 0
verify_threads: 1
whitelist_channels: []Conda list
conda list
# Name Version Build Channel
argon2-cffi 20.1.0 py38h2bbff1b_1 defaults
async_generator 1.10 pyhd3eb1b0_0 defaults
attrs 21.2.0 pyhd3eb1b0_0 defaults
backcall 0.2.0 pyhd3eb1b0_0 defaults
blas 1.0 mkl defaults
bleach 4.0.0 pyhd3eb1b0_0 defaults
brotli 1.0.9 ha925a31_2 defaults
ca-certificates 2021.10.26 haa95532_2 defaults
certifi 2021.10.8 py38haa95532_0 defaults
cffi 1.14.6 py38h2bbff1b_0 defaults
colorama 0.4.4 pyhd3eb1b0_0 defaults
cycler 0.10.0 py38_0 defaults
debugpy 1.4.1 py38hd77b12b_0 defaults
decorator 5.1.0 pyhd3eb1b0_0 defaults
defusedxml 0.7.1 pyhd3eb1b0_0 defaults
entrypoints 0.3 py38_0 defaults
fonttools 4.25.0 pyhd3eb1b0_0 defaults
freetype 2.10.4 hd328e21_0 defaults
icc_rt 2019.0.0 h0cc432a_1 defaults
icu 58.2 ha925a31_3 defaults
importlib-metadata 4.8.1 py38haa95532_0 defaults
importlib_metadata 4.8.1 hd3eb1b0_0 defaults
intel-openmp 2021.4.0 haa95532_3556 defaults
ipykernel 6.4.1 py38haa95532_1 defaults
ipython 7.29.0 py38hd4e2768_0 defaults
ipython_genutils 0.2.0 pyhd3eb1b0_1 defaults
ipywidgets 7.6.5 pyhd3eb1b0_1 defaults
jedi 0.18.0 py38haa95532_1 defaults
jinja2 3.0.2 pyhd3eb1b0_0 defaults
jpeg 9d h2bbff1b_0 defaults
jsonschema 3.2.0 pyhd3eb1b0_2 defaults
jupyter 1.0.0 py38_7 defaults
jupyter_client 7.0.1 pyhd3eb1b0_0 defaults
jupyter_console 6.4.0 pyhd3eb1b0_0 defaults
jupyter_core 4.8.1 py38haa95532_0 defaults
jupyterlab_pygments 0.1.2 py_0 defaults
jupyterlab_widgets 1.0.0 pyhd3eb1b0_1 defaults
kiwisolver 1.3.1 py38hd77b12b_0 defaults
libpng 1.6.37 h2a8f88b_0 defaults
libtiff 4.2.0 hd0e1b90_0 defaults
libwebp 1.2.0 h2bbff1b_0 defaults
lz4-c 1.9.3 h2bbff1b_1 defaults
m2w64-gcc-libgfortran 5.3.0 6 defaults
m2w64-gcc-libs 5.3.0 7 defaults
m2w64-gcc-libs-core 5.3.0 7 defaults
m2w64-gmp 6.1.0 2 defaults
m2w64-libwinpthread-git 5.0.0.4634.697f757 2 defaults
markupsafe 2.0.1 py38h2bbff1b_0 defaults
matplotlib 3.4.3 py38haa95532_0 defaults
matplotlib-base 3.4.3 py38h49ac443_0 defaults
matplotlib-inline 0.1.2 pyhd3eb1b0_2 defaults
mistune 0.8.4 py38he774522_1000 defaults
mkl 2021.4.0 haa95532_640 defaults
mkl-service 2.4.0 py38h2bbff1b_0 defaults
mkl_fft 1.3.1 py38h277e83a_0 defaults
mkl_random 1.2.2 py38hf11a4ad_0 defaults
msys2-conda-epoch 20160418 1 defaults
munkres 1.1.4 py_0 defaults
nbclient 0.5.3 pyhd3eb1b0_0 defaults
nbconvert 6.1.0 py38haa95532_0 defaults
nbformat 5.1.3 pyhd3eb1b0_0 defaults
nest-asyncio 1.5.1 pyhd3eb1b0_0 defaults
notebook 6.4.5 py38haa95532_0 defaults
numpy 1.21.2 py38hfca59bb_0 defaults
numpy-base 1.21.2 py38h0829f74_0 defaults
olefile 0.46 pyhd3eb1b0_0 defaults
openssl 1.1.1l h2bbff1b_0 defaults
packaging 21.0 pyhd3eb1b0_0 defaults
pandocfilters 1.4.3 py38haa95532_1 defaults
parso 0.8.2 pyhd3eb1b0_0 defaults
pickleshare 0.7.5 pyhd3eb1b0_1003 defaults
pillow 8.4.0 py38hd45dc43_0 defaults
pip 21.0.1 py38haa95532_0 defaults
prometheus_client 0.11.0 pyhd3eb1b0_0 defaults
prompt-toolkit 3.0.20 pyhd3eb1b0_0 defaults
prompt_toolkit 3.0.20 hd3eb1b0_0 defaults
pycparser 2.20 py_2 defaults
pygments 2.10.0 pyhd3eb1b0_0 defaults
pyparsing 3.0.4 pyhd3eb1b0_0 defaults
pyqt 5.9.2 py38ha925a31_4 defaults
pyrsistent 0.17.3 py38he774522_0 defaults
python 3.8.12 h6244533_0 defaults
python-dateutil 2.8.2 pyhd3eb1b0_0 defaults
pywin32 228 py38hbaba5e8_1 defaults
pywinpty 0.5.7 py38_0 defaults
pyzmq 22.2.1 py38hd77b12b_1 defaults
qt 5.9.7 vc14h73c81de_0 defaults
qtconsole 5.1.1 pyhd3eb1b0_0 defaults
qtpy 1.10.0 pyhd3eb1b0_0 defaults
scipy 1.7.1 py38hbe87c03_2 defaults
send2trash 1.8.0 pyhd3eb1b0_1 defaults
setuptools 58.0.4 py38haa95532_0 defaults
sip 4.19.13 py38ha925a31_0 defaults
six 1.16.0 pyhd3eb1b0_0 defaults
sqlite 3.36.0 h2bbff1b_0 defaults
terminado 0.9.4 py38haa95532_0 defaults
testpath 0.5.0 pyhd3eb1b0_0 defaults
tk 8.6.11 h2bbff1b_0 defaults
tornado 6.1 py38h2bbff1b_0 defaults
traitlets 5.1.0 pyhd3eb1b0_0 defaults
vc 14.2 h21ff451_1 defaults
vs2015_runtime 14.27.29016 h5e58377_2 defaults
wcwidth 0.2.5 pyhd3eb1b0_0 defaults
webencodings 0.5.1 py38_1 defaults
wheel 0.37.0 pyhd3eb1b0_1 defaults
widgetsnbextension 3.5.1 py38_0 defaults
wincertstore 0.2 py38haa95532_2 defaults
winpty 0.4.3 4 defaults
xz 5.2.5 h62dcd97_0 defaults
zipp 3.6.0 pyhd3eb1b0_0 defaults
zlib 1.2.11 h62dcd97_4 defaults
zstd 1.4.9 h19a0ad4_0 defaultsAdditional Context
No response
Activity
bnavigator commentedon Nov 4, 2021
See discussion in numpy/numpy#20233. Conda's main repo linalg libraries need a fix. MKL and OpenBLAS.
kenodegard commentedon Nov 19, 2021
This appears to be an issue with the
numpypackage, notcondaitself. I would encourage reporting this to https://github.com/ContinuumIO/anaconda-issues/.Thanks!