tensorflow-1.15.0 manylinux wheel on PyPi is invalid

[fhoering]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): CentOS Linux release 7.7.1908 (Core)
• Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device:
• TensorFlow installed from (source or binary): binary
• TensorFlow version: 1.15.0
• Python version: 3.6
• Installed using virtualenv? pip? conda?: virtualenv pip
• Bazel version (if compiling from source): N/A
• GCC/Compiler version (if compiling from source): N/A
• CUDA/cuDNN version: N/A
• GPU model and memory: N/A

Describe the problem

The published manylinux wheel of tensorflow 1.15.0 on Pypi is invalid:
https://files.pythonhosted.org/packages/3f/98/5a99af92fb911d7a88a0005ad55005f35b4c1ba8d75fba02df726cd936e6/tensorflow-1.15.0-cp36-cp36m-manylinux2010_x86_64.whl

(1.14.0 & 2.0.0 manylinux wheels work fine)

Provide the exact sequence of commands / steps that you executed before running into the problem

[f.horing] ~/tmp $python3.6 -m venv venv
[f.horing] ~/tmp $. venv/bin/activate
(venv) [f.horing] ~/tmp $pip install wheel
Collecting wheel
  Downloading http://build-nexus.prod.crto.in/repository/pypi/packages/00/83/b4a77d044e78ad1a45610eb88f745be2fd2c6d658f9798a15e384b7d57c9/wheel-0.33.6-py2.py3-none-any.whl
Installing collected packages: wheel
Successfully installed wheel-0.33.6
You are using pip version 9.0.3, however version 19.3.1 is available.
You should consider upgrading via the 'pip install --upgrade pip' command.
(venv) [f.horing] ~/tmp $python
Python 3.6.8 (default, Aug  7 2019, 17:28:10)
[GCC 4.8.5 20150623 (Red Hat 4.8.5-39)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from wheel.wheelfile import WheelFile
>>> wf = WheelFile("tensorflow-1.15.0-cp36-cp36m-manylinux2010_x86_64.whl")
>>> wf.extractall("/tmp")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib64/python3.6/zipfile.py", line 1524, in extractall
    self._extract_member(zipinfo, path, pwd)
  File "/usr/lib64/python3.6/zipfile.py", line 1579, in _extract_member
    shutil.copyfileobj(source, target)
  File "/usr/lib64/python3.6/shutil.py", line 79, in copyfileobj
    buf = fsrc.read(length)
  File "/usr/lib64/python3.6/zipfile.py", line 872, in read
    data = self._read1(n)
  File "/usr/lib64/python3.6/zipfile.py", line 962, in _read1
    self._update_crc(data)
  File "/home/f.horing/tmp/venv/lib64/python3.6/site-packages/wheel/wheelfile.py", line 91, in _update_crc
    raise WheelError("Hash mismatch for file '{}'".format(native(ef_name)))
wheel.cli.WheelError: Hash mismatch for file 'tensorflow-1.15.0.dist-info/METADATA'
>>> wf = WheelFile("tensorflow-2.0.0-cp36-cp36m-manylinux2010_x86_64.whl")
>>> wf.extractall("/tmp")
>>> wf = WheelFile("tensorflow-1.14.0-cp36-cp36m-manylinux1_x86_64.whl")
>>> wf.extractall("/tmp")
>>>

Any other info / logs
Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.

We would like to upgrade to tf 1.15 first before upgrading to 2.0.0

[mihaimaruseac]

Strange, because pip install tensorflow==1.15.0 works.

[fhoering]

Yes. But the hash of metadata is still wrong. pip install doesn't seem to check that.

My use case is that I try to embed the tensorflow wheel in a pex file and that fails.

[av8ramit]

@goldiegadde any idea on what changed between 2.0.0 and 1.15?

[mihaimaruseac]

1.15 has the single pip package for CPU/GPU and 2.0 doesn't. As we needed to change pip files for this, it might be the reason for this.

[fhoering]

Can you fix this ?
We would like to test this single package on GPU/CPU machines before moving on to 2.0.
Otherwise it means we need to upgrade directly to 2.0 which will be more risky.

[mihaimaruseac]

2.0 doesn't have a single pip package. @yifeif is working on enabling it for 2.1

[fhoering]

OK. thanks for the info. I wasn't aware of this. That also explains why tf 2.0.0 wheel is so small.

Can you nevertheless fix this ? I suppose we will have the same issue with 2.1 and we currently use both packages tensorflow-gpu and tensorflow and would like to use only one.

[mihaimaruseac]

Yes, 2.1 should not have this issue again.

[yifeif]

Thanks for reporting @fhoering! Yes we should and will fix this.

[tensorflow-bot]

Are you satisfied with the resolution of your issue?
Yes
No

[ronsoohyeong]

Hi Just change the latter part of the file name to py3-none-any.whl then it should work. 2020년 4월 10일 (금) 04:35, zpymyyn <notifications@github.com>님이 작성:

…

wheel pack tensorflow_gpu-1.15.0 Hi, thanks for the instructions. I managed to go to step 4, yet my METADATA file is empty, thus can't pack it. Do you mind sharing the METADATA file here? Thanks. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#33569 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AE7D4657KIQ73EN6QXZ2OFTRLYPQNANCNFSM4JC4WDUA> .

[pengyuan-zhou]

Hi Just change the latter part of the file name to py3-none-any.whl then it should work. 2020년 4월 10일 (금) 04:35, zpymyyn notifications@github.com님이 작성:
…
wheel pack tensorflow_gpu-1.15.0 Hi, thanks for the instructions. I managed to go to step 4, yet my METADATA file is empty, thus can't pack it. Do you mind sharing the METADATA file here? Thanks. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#33569 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE7D4657KIQ73EN6QXZ2OFTRLYPQNANCNFSM4JC4WDUA .

I think I deleted this comment right after posted.
Anyway, thanks.