-
Notifications
You must be signed in to change notification settings - Fork 110
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
TF version 1.14.0 (almost works locally) #512
base: master
Are you sure you want to change the base?
Conversation
I just noticed there was an upgrade. Possibly some more is needed, since this is a major(?) upgrade, why I didn't put in "skip ci". I'm just going by last change I saw, that was actually a (temporary) downgrade. I don't need this myself. OCaml wrapper is up to date, and I see it took a larger jump (and seemingly no changes needed, only one what seems to be for OCaml not TF): |
Thanks! Looks like CI failed, though - maybe something with the packaging of the upstream TensorFlow binaries has changed. |
"maybe something with the packaging of the upstream TensorFlow binaries has changed." To compare I looked at commits for OCaml (I would have thought less used the Julia). there only the download links in the README where updated; but you have to get the build system and compile yourself. It seems they use C++ [API] (I know you do not; as other recommended), while not obvious where (I see C++ only in "Use TensorFlow C++ based gradient computation"): |
Less urgent than #513 |
Any update? Looks like the CI failed because it cannot find the shared library. |
Not from me (should be simple though, or to support 1.15.2), see full answer at #513 |
I do also expect 1.15.2 to fail. Just changed PR to check. |
Hi, @oxinabox (and @chengchingwen) is 12.x really needed or 14.0 (or something newer), what I seemingly got to work, ok for you? I got it to work locally, with:
1.14 (and 1.15.2) do not work here on GitHub, as the library is missing, which I found out how to fix, but only locally, and so far not newer than |
pip3 install -U pip seemingly works (a TF 2 requirement to have pip >19) for conda to download TF 2.1.0 however, in the code this link gets generated and fails https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-cpu-linux-x86_64-2.1.0.tar.gz neither does this one work But slightly older I seemingly get to build as it downloads: |
1.14.0 is preferred over 1.12.x if you can get it to work. |
I found out why only the build.jl worked. Before there was only a .so file of each, now it's a links (and links to links) of both: /home/pharaldsson_sym/.julia/dev/TensorFlow/deps/downloads/lib/libtensorflow_framework.so |
No description provided.