Max H. Gerlach
Max H. Gerlach
You don't need to use a docker image. You can directly follow instructions under https://www.tensorflow.org/install/install_sources on your machine. The build took a couple of hours in my case.
I might provide such a wheel if I need it myself one day, but that's not sure and it would still be quite a while.
> AttributeError: module 'torch' has no attribute '__version__' This sounds weird. Can you properly import (and test) torch in an interactive python session in the environment that you want to...
With `jinja2` _not_ pinned to version `3.0.*` I get this error message: ``` horovod/docs$ make html Traceback (most recent call last): File "/Users/max/deepl-git/horovod/docs/venv/bin/sphinx-build", line 5, in from sphinx.cmd.build import main...
Fixed with the new nightly package: https://github.com/horovod/horovod/runs/7805639078?check_suite_focus=true
> Will this slow down tests? Not sure, let's compare run times with master when the CI is done. I mainly thought of this idea because there's a chance that...
> > Will this slow down tests? > > Not sure, let's compare run times with master when the CI is done. The variation in run times between different instances...
Hi @yundai424, there's a bunch of spark related failures in the latest CI run, which I think is just flakiness (have seen that in another unrelated PR), but there are...
> Also I noticed there're two Build and Test GPU tasks that appear Pending but in fact passed if I clicked Details.. Is that expected? 🧐 That's currently a bug...
Thanks for the approval!