"ERROR: Cannot find './util/python/python_lib'" #2703
Comments
|
Did you run configure? |
|
Hello, Yes I did run configure as it is written in the "Steps to reproduce" section of my previous post :-) |
|
@jplu: I saw that, but one can never be too careful when triaging hundreds of issues. :) Can you add a |
|
@girving: No worries I understand :) I put a
But nothing is displayed: |
|
@jplu You might need a flag to bazel to get it to not suppress stdout. Not sure which that is, though; |
|
The redirect command did the thing, I had to think about this solution before... Anyway, here the path:
|
|
@damienmg We have a https://github.com/tensorflow/tensorflow/blob/master/util/python/BUILD#L14 On the other hand, @martinwicke, I suppose if we add |
|
FWIW: the path was in the output :) The test that check a symlink is printing a more comprehensive message about it not being a symlink, or did I read the script wrong? |
|
(and nothing special required with symlinks, but remember that when running, most files will actually be symlinks while directories will be recreated). |
|
Can we just use all_files as a data depedency for python_check? That will not fail (since it's just a glob), and if nothing is there, we'll still see it's not there and get our nice message. |
|
@martinwicke Seems reasonable. Want to assign it to someone? |
|
This used to work out of the box, no? Sounds like a bug was recently introduced. |
|
Olivia, can you change the BUILD file in util/python to have a filegroup containing all files in the directory, like this: filegroup( (syntax may be screwy), and add that as a data dependency to the python_check target? And then make sure it works... I hadn't seen this before. |
|
Sorry for the delay I was not able to use this computer until now. Unfortunately the problem still remains :-( I have done a fresh new clone corresponding to the commit 84225a2 |
|
@caisq Could you take a look at this please? |
|
I had this same issue in the closed tickets referenced above, @jmchen-g thanks for re-opening this one. |
|
Hello, I have tested with the commit fefe52c that contains the fix proposed by @plutoshe. Unfortunately the issue still remains :( Don't hesitate to ask me if you need other information to help you to debug. |
|
what's your bazel version? |
|
I'm using Bazel 0.2.3 |
|
I'm using bazel 0.2.2 with the same error |
|
Which python did you specify when running |
|
Python specified is /usr/local/bin/python (version Python 2.7.11) Ran /bin/bash before the install instructions with no change in result. |
|
The specified Python is: I also can run |
|
Ok. I'm just stabbing at the dark -- I cannot reproduce this and I'm wondering what the difference between our installations might be. Can you try running using docker just to see if that works? |
|
It works via Docker: |
|
When I run: I get the same error. |
|
Once the run of |
|
Ok, I found the issue, the script
The problem is that the path Everything is working like a charm if I set the symlink A possible fix could be to ask another question in the configure to set where the packages for Python are installed. WDYT? |
|
Interesting. We actually ask distutils for this path (see utils/python/python_config.sh), so it appears distutils is misconfigured. We could switch to an alternative method, but I'm not at all sure it's better in general. |
|
The main problem is that people might have multiple location for their Python's packages. So why not checking the |
|
Let's see if I understood: Let's say |
|
Yes you have understood. If a Python lib is not in the standard location or in For example for my case, my If I run the interactive python and do So I think, by just checking this environment variable we should be safe, but we never know. Nevertheless, checking the I get only those two, and not the others contained in my
|
|
@itsmeolivia can you look into modifying the python setup to include the workflow outlined by @jplu? |
…on path from a list of possible known locations Change: 129778433
|
Perfect! Everything is compiling like a charm with this fix Thanks a lot! |
CHANGELOG ========= d4ae542ed - Fix nullptr dereference issue in triangular product. 7769eb1b2 - Fix problems with recent changes and Tensorflow in Power ba1cb6e45 - Fixes #2703 by adding max_digits10 function 9995c3da6 - Fix -Wmaybe-uninitialized in SVD 4e9e493b4 - Fix -Waggressive-loop-optimizations 6e7abeae6 - fix arm build warnings 81fe2d424 - Fix more gcc compiler warnings / sort-of bugs 21cd3fe20 - Optimize check_rows_cols_for_overflow 9297aae66 - Fix AVX512 nomalloc issues in trsm. 1a2bfca8f - Fix annoying warnings 63dcb429c - Fix use of arg function in CUDA. 8f927fb52 - Altivec: fix compilation with C++20 and higher d4b05454a - Fix argument for _mm256_cvtps_ph imm parameter 15ac3765c - Fix ivcSize return type in IndexedViewMethods.h 3791ac8a1 - Fix supportsMMA to obey EIGEN_ALTIVEC_MMA_DYNAMIC_DISPATCH compilation flag and compiler support. bc57b926a - Add Quaternion constructor from real scalar and imaginary vector 31cd2ad37 - Ensure EIGEN_HAS_ARM64_FP16_VECTOR_ARITHMETIC is always defined on arm. 7465b7651 - Disable FP16 arithmetic for arm32. b3267f693 - Remove unused variable in test/svd_common.h. PiperOrigin-RevId: 552130715
Uh oh!
There was an error while loading. Please reload this page.
Hello,
I have the following error when I try to compile the pip package from the source code.
Environment info
Operating System: Ubuntu 14.04
Installed version of CUDA and cuDNN:
(please attach the output of
ls -l /path/to/cuda/lib/libcud*):If installed from sources, provide the commit hash: ea9e00a
Steps to reproduce
Logs or other output that would be helpful
But, I do have this folder locally in the Tensorflow repository. Did I do something wrong? Or it is a bug coming from Tensorflow?
Thanks in advance for any help!
The text was updated successfully, but these errors were encountered: