Open timpitman opened 4 years ago
This is still on the to-do list. Unfortunately, it has not been as easy to build for macOS for this version as previous ones.
Thanks @csoja, to give context - this is preventing Conda users of our StellarGraph graph machine learning library from installing on MacOS via the conda package.
I am on this issue. I will get to osx version soon.
This is getting to be an issue for us as well with maintaining our CaImAn neuroscience suite:
(In our case, it's also an issue that we're locked on python 3.7 on OSX because there are no recent builds)
Yes, I understand your issue pretty well, but until now Tensorflow resisted pretty well to be built on MacOSX using clang. I spend several hours on this issue, but sadly couldn't resolve it. Some parts of tensorflow seems to stick to a gcc compiler and it was impossible to build it by using clang.
Of course help is most welcome on this. I can push my changes to the custom-compier settings I did for it.
Is installing/using a real gcc on OSX an option?
Yes, it is at least worth a try. We could build a gcc 10 using specific 10.14 SDK as sysroot ... I am in doubt that current crosstool-ng supports native osx. It isn't hard to build a gcc for OSX, nevertheless we would need to take care to distribute runtime libraries and target-libraries for it, as we would do for other targets too.
We now have a tensorflow
package on conda-forge using conda-forge's compiler stack, feel free to merge this / take a look at it.
tensorflow version 2.1.0 is not showing up as an option on the conda-forge channel
tensorflow version 2.1.0 is not showing up as an option on the conda-forge channel
2.4.1 is available on conda-forge; is 2.1.0 specifically needed, or can you use a newer version?
I tried installing 2.4.1 to test if it would work, but did not find that package version in my anaconda navigator for the specific environment I installed it in. Perhaps it is an issue on my end.
$ conda install -c conda-forge tensorflow
produced:
Collecting package metadata (current_repodata.json): done
Solving environment: done
## Package Plan ##
environment location: /opt/anaconda3/envs/SySe_VulnerabilityDetection
added / updated specs:
- tensorflow
The following packages will be downloaded:
package | build
---------------------------|-----------------
ca-certificates-2021.5.30 | h033912b_0 136 KB conda-forge
certifi-2021.5.30 | py37hf985489_0 141 KB conda-forge
openssl-1.1.1k | h0d85af4_0 1.9 MB conda-forge
python_abi-3.7 | 2_cp37m 4 KB conda-forge
------------------------------------------------------------
Total: 2.2 MB
The following NEW packages will be INSTALLED:
python_abi conda-forge/osx-64::python_abi-3.7-2_cp37m
The following packages will be SUPERSEDED by a higher-priority channel:
ca-certificates pkgs/main::ca-certificates-2021.7.5-h~ --> conda-forge::ca-certificates-2021.5.30-h033912b_0
certifi pkgs/main::certifi-2021.5.30-py37hecd~ --> conda-forge::certifi-2021.5.30-py37hf985489_0
openssl pkgs/main::openssl-1.1.1k-h9ed2024_0 --> conda-forge::openssl-1.1.1k-h0d85af4_0
Proceed ([y]/n)? y
Downloading and Extracting Packages
ca-certificates-2021 | 136 KB | ############################################################################## | 100%
python_abi-3.7 | 4 KB | ############################################################################## | 100%
openssl-1.1.1k | 1.9 MB | ############################################################################## | 100%
certifi-2021.5.30 | 141 KB | ############################################################################## | 100%
Preparing transaction: done
Verifying transaction: | WARNING conda.core.path_actions:verify(962): Unable to create environments file. Path not writable.
environment location: /Users/crystalcontreras/.conda/environments.txt
done
Executing transaction: \ WARNING conda.core.envs_manager:register_env(50): Unable to register environment. Path not writable or missing.
environment location: /opt/anaconda3/envs/SySe_VulnerabilityDetection
registry file: /Users/crystalcontreras/.conda/environments.txt
done
Actual Behavior
TensorFlow 2.1.0 is available for linux-64 and win-64 platforms from https://anaconda.org/anaconda/tensorflow, but not for osx-64. The issue has been reported inside https://github.com/tensorflow/tensorflow/issues/35754, however that issue is closed.
Expected Behavior
Tensorflow 2.1.0 should be available for Mac (osx-64)
Steps to Reproduce
conda install tensorflow=2.1.0
Anaconda or Miniconda version:
Anaconda 2020.02
Operating System:
osx-64