Yes againg boring coding things. But it’s useful for me, sorry. The github page is here. The installation section is kinde of meager. The have requirements, so let’s focus on them. I check my machine first:
(base) user@computer ~ $ > conda -V
conda 23.7.4
(base) user@computer ~ $ > python -V
Python 3.9.18
(base) user@computer ~ $ > nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Fri_Jan__6_16:45:21_PST_2023
Cuda compilation tools, release 12.0, V12.0.140
Build cuda_12.0.r12.0/compiler.32267302_0
There you go. So it should work. We git clone the directory (standard procedure), go in and follow instructions. But the first step already fails Like this:
(base) user@computer ~ $ > pip install tensorflow-gpu==2.12.0
Defaulting to user installation because
normal site-packages is not writeable
Looking in indexes: https://pypi.org/simple,
https://pypi.ngc.nvidia.com
Collecting tensorflow-gpu==2.12.0
Downloading tensorflow-gpu-2.12.0.tar.gz (2.6 kB)
Preparing metadata (setup.py) ... error
error: subprocess-exited-with-error
× python setup.py egg_info did not run successfully.
│ exit code: 1
╰─> [39 lines of output]
Traceback (most recent call last):
I try with different versions and google a little. Here we have a list of tensorflows available. It’s not clear is available for such a modern CUDA. But on this link they say “just try directly” and I try. And it works. Like this.
(base) user@computer ~ $ >pip install tensorflow
Defaulting to user installation because
normal site-packages is not writeable
Looking in indexes: https://pypi.org/simple,
https://pypi.ngc.nvidia.com
Collecting tensorflow
Obtaining dependency information for tensorflow from
https://files.pythonhosted.org/XXX/tensorflow-2.14.YYY
... here the download of needed packages...
Installing collected packages: libclang, flatbuffers,
zipp, wrapt, typing-extensions, termcolor,
tensorflow-io-gcs-filesystem, tensorflow-estimator,
tensorboard-data-server, pyasn1, protobuf, oauthlib,
numpy, MarkupSafe, keras, grpcio, google-pasta, gast,
cachetools, astunparse, absl-py, werkzeug, rsa,
requests-oauthlib, pyasn1-modules, opt-einsum, ml-dtypes,
importlib-metadata, h5py, markdown, google-auth,
google-auth-oauthlib, tensorboard, tensorflow
Attempting uninstall: numpy
Found existing installation: numpy 1.22.4
Uninstalling numpy-1.22.4:
Successfully uninstalled numpy-1.22.4
Successfully installed MarkupSafe-2.1.3
absl-py-2.0.0 astunparse-1.6.3 cachetools-5.3.2
flatbuffers-23.5.26 gast-0.5.4 google-auth-2.23.4
google-auth-oauthlib-1.0.0 google-pasta-0.2.0
grpcio-1.59.2 h5py-3.10.0 importlib-metadata-6.8.0
keras-2.14.0 libclang-16.0.6 markdown-3.5.1 ml-dtypes-0.2.0
numpy-1.26.2 oauthlib-3.2.2 opt-einsum-3.3.0 protobuf-4.25.0
pyasn1-0.5.0 pyasn1-modules-0.3.0 requests-oauthlib-1.3.1
rsa-4.9 tensorboard-2.14.1 tensorboard-data-server-0.7.2
tensorflow-2.14.0 tensorflow-estimator-2.14.0
ensorflow-io-gcs-filesystem-0.34.0 termcolor-2.3.0
typing-extensions-4.8.0 werkzeug-3.0.1
wrapt-1.14.1 zipp-3.17.0
Then we go to the IsoNet folder (that we get throug git clone) and install the dependencies.
IsoNet $ > pip install -r requirements.txt
Defaulting to user installation because normal
site-packages is not writeable
Looking in indexes: https://pypi.org/simple,
https://pypi.ngc.nvidia.com
...here the donwloads...
Successfully built scikit-image fire
Installing collected packages: PyQt5-Qt5, tifffile,
scipy, PyWavelets, PyQt5-sip, pyparsing, pillow,
networkx, mrcfile, kiwisolver, importlib-resources,
fonttools, fire, cycler, contourpy, PyQt5, matplotlib,
imageio, scikit-image
Successfully installed PyQt5-5.15.10 PyQt5-Qt5-5.15.2
PyQt5-sip-12.13.0 PyWavelets-1.4.1 contourpy-1.2.0
cycler-0.12.1 fire-0.5.0 fonttools-4.44.0 imageio-2.32.0
importlib-resources-6.1.1 kiwisolver-1.4.5 matplotlib-3.8.1
mrcfile-1.4.3 networkx-3.2.1 pillow-10.0.1 pyparsing-3.1.1
scikit-image-0.17.2 scipy-1.11.3 tifffile-2023.9.26
Well no errors also. Which is good. We open a new shell to test. This is my output
(base) user@computer ~ $ echo $PYTHONPATH
(base) user@computer ~ $ cd IsoNet/
(base) user@computer ~ $ source source-env.sh
| ENV: PYTHONPATH='/home/user:'
| ENV: PATH='/home/user/IsoNet/bin:XXXXX'
(base) user@computer ~ $ echo $PYTHONPATH
/home/user:
(base) user@computer ~ $ isonet.py check
IsoNet --version 0.2 installed
So I guess we are ready to go! NOTE: backposted because… because… because I feel like it’s an old item 🙂
Bits and Dragons 