ADCME.jl stuck on Tensorflow Installation [MacOS] #64
Comments
|
Can you try using PyCall
tf = pyimport("tensorflow")Does it get stuck as well? |
|
The command shows no response. Do I need to install TensorFlow manually? |
|
It's really weird I ended up checking whether python installed in the The snippet below works without an issue $ ~/.julia/adcme/bin/python
...
>>> import tensorflow as tf
>>>I'm out of ideas... |
|
Unfortunately, this is a known issue long ago. It might be an incompatible version of LLVM on Mac. So I suggest you try another Julia version or machine. |
|
I got everything installed on a linux cluster, but will try with an older Julia version |
|
Maybe add a label to your issues 'known issues'? I think that would help users! Thank you so much for your help either way, I'm going to try it out on the cluster for now. |
|
Sorry for the inconvenience and thanks for the suggestion. We are going to look into the issue and try to find a workaround in the future. |
|
Any update on this? |
|
I also meet with the same problem on Mac OS Big Sur 11.2.1 |
|
Changing to Julia 1.3 has fixed this issue for me.
Building ADCME ─→ `~/.julia/packages/ADCME/7qIYe/deps/build.log`
[ Info: Your Julia version is 1.3.1, current ADCME version is 0.7.0, ADCME dependencies installation path: /Users/lizhehao/.julia/adcme
[ Info: --------------- (1/6) Install Tensorflow Dependencies ---------------
[ Info: ADCME dependencies have already been installed.
[ Info: Installing conda dependencies...
Collecting package metadata (repodata.json): done
Solving environment: -
The environment is inconsistent, please check the package plan carefully
The following packages are causing the inconsistency:
- conda-forge/noarch::tensorflow-probability==0.8.0=py_0
- defaults/osx-64::tensorflow-base==1.15.0=mkl_py37h032239d_0
- defaults/noarch::tensorflow-estimator==1.15.1=pyh2649769_0
- defaults/osx-64::tensorflow==1.15.0=mkl_py37hb249377_0
done
==> WARNING: A newer version of conda exists. <==
current version: 4.8.4
latest version: 4.9.2
Please update conda by running
$ conda update -n base conda
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
#
# To activate this environment, use
#
# $ conda activate base
#
# To deactivate an active environment, use
#
# $ conda deactivate
[ Info: --------------- (2/6) Check Python Version ---------------
Building Conda ─→ `~/.julia/packages/Conda/tJJuN/deps/build.log`
Resolving package versions...
Building PyCall → `~/.julia/packages/PyCall/tqyST/deps/build.log`
Resolving package versions...
┌ Info: PyCall Python version: /Users/lizhehao/.julia/adcme/bin/python
└ Conda Python version: /Users/lizhehao/.julia/adcme/bin/python
[ Info: --------------- (3/6) Looking for TensorFlow Dynamic Libraries ---------------
[ Info: --------------- (4/6) Preparing Custom Operator Environment ---------------
[ Info: Downloading dependencies to /Users/lizhehao/.julia/adcme/lib/Libraries...
[ Info: --------------- (5/6) Skipped: Installing GPU Dependencies ---------------
[ Info: --------------- (6/6) Write Dependency Files ---------------
[ Info: --------------- Finished: /Users/lizhehao/.julia/packages/ADCME/7qIYe/deps/deps.jl --------------- |
Hi,
I have been struggling with the installation of
ADCME.jlfor almost an entire afternoon, when I found inADSeismic.jlthat it may take up to 20 min. So, I let the installation run with build and the verbose flag. It turns out I can't get past the TensorFlow Library building.I'm using a clean installation of Julia 1.5.2 (
rm -rf ~/.julia) and thenThe build log is pasted below
Thanks!
Lucas
PS: I ended cancelling the installation a bunch of times since I always thought "it can't take thaat long". I even cleared out my entire
Juliabecause I thought I had some old package that didn't install correctly! It would be nice, if it said how long the installation approximately takes since it's longer than the average Julia package!Verbose Build Output
The text was updated successfully, but these errors were encountered: