Open adowling2 opened 1 year ago
@dguittet @radhakrishnatg @esrawli @nareshsusarla @Xinhe-Chen To recap, our goal is to have the scripts/notebooks used to generate key results from the reports included in the final release. If the scripts/notebooks are not in the release candidate, please work with @lbianchi-lbl to open and merge a PR.
@lbianchi-lbl I am following these instructions on an M1 macOS (with a recent branch on @nareshsusarla's fork) and got this error:
INFO: pip is looking at multiple versions of dispatches[surrogates] to determine which version is compatible with other requirements. This could take a while.
ERROR: Could not find a version that satisfies the requirement tensorflow==2.10.0 (from dispatches[surrogates]) (from versions: 2.13.0rc0, 2.13.0rc1, 2.13.0rc2, 2.13.0)
ERROR: No matching distribution found for tensorflow==2.10.0
I decided to do pip install --editable .
because I want to see if some scripts in @nareshsusarla's fork run on my computer (and tensorflow
is not needed), so this is unimportant.
@lbianchi-lbl I am following these instructions on an M1 macOS (with a recent branch on @nareshsusarla's fork) and got this error:
INFO: pip is looking at multiple versions of dispatches[surrogates] to determine which version is compatible with other requirements. This could take a while. ERROR: Could not find a version that satisfies the requirement tensorflow==2.10.0 (from dispatches[surrogates]) (from versions: 2.13.0rc0, 2.13.0rc1, 2.13.0rc2, 2.13.0) ERROR: No matching distribution found for tensorflow==2.10.0
Is this in a fresh Conda environment? If so, the Python version could be an issue - for various reasons DISPATCHES is only compatible with 3.8 or 3.9, with the latter being the one used for the instructions above.
Running pip install --editable .
is a good alternative to pip install -r requirements-dev.txt
, modulo some possible test failures due to e.g. addheader
, but if things work, then no need to change.
Yes, a new conda environment with Python 3.9. I wonder if that version of tensorflow is not available for macOS arm.
Yes, a new conda environment with Python 3.9. I wonder if that version of tensorflow is not available for macOS arm.
Oh, right, sorry, for some reason I skipped the part about the M1 Mac part in your message. Yes, I'd say it's likely, especially considering that 2.10 is relatively old. From a quick check, it looks like the arm64
build is only available (on PyPI) since Tensorflow 2.13: https://pypi.org/project/tensorflow/2.13.0/#files
Goal
Verify the key results in each case study report are reproducible with the final DISPATCHES release candidate
Tasklist
RE case study (generic)
NE case study (generic) (#210)
*TensorFlow will likely be pinned to 2.12 (or 2.10)
FE case study (PR #209 )
Others
Instructions
Set up environment for verification
The output of the last two commands should look something like this: