Python: Tensorflow 2.5 requires grpcio 1.34

[gitpushdashf]

Tensorflow 2.5.0 requires grpcio 1.34, which prevents installation with some other packages that require 1.37 and later.

Related: #48109, f1a51f0#diff-f526feeafa1000c4773410bdc5417c4022cb2c7b686ae658b629beb541ae9112

It seems like the above commit should've landed for 2.5.0, but it appears to not be that way in pypi.

Log output from Poetry:

(snip)
And because tensorflow (2.5.0) depends on grpcio (>=1.34.0,<1.35.0),
(snip)

[ymodak]

On master I see

tensorflow/tensorflow/tools/ci_build/release/common_win.bat

Line 56 in 5393115

%PY_EXE% -m pip install "grpcio >= 1.37.0, < 2.0"

[mihaimaruseac]

This is the only relevant direct dependency on grpcio on TF:

tensorflow/tensorflow/tools/pip_package/setup.py

Line 130 in 1595ee1

REQUIRED_PACKAGES.append('grpcio >= 1.37.0, < 2.0')

Can you post the output of pip list and details about the architecture you are on?

[mihaimaruseac]

Oh, sorry, you meant the 2.5 release. There we indeed pin to 1.34.

tensorflow/tensorflow/tools/pip_package/setup.py

Lines 126 to 130 in 15d5b93

	# grpcio does not build correctly on big-endian machines due to lack of
	# BoringSSL support.
	# See https://github.com/tensorflow/tensorflow/issues/17882.
	if sys.byteorder == 'little':
	REQUIRED_PACKAGES.append('grpcio ~= 1.34.0')

Unfortunately, we won't be able to alter this without doing a patch release. We currently need to do patch releases for the vulns mentions in the 2.5 release notes and we're also very close to the 2.6 release. So let's wait until 2.6 and there this issue would be solved.

In the future, we will add automation to ensure dependencies are updated before release (CC @pak-laura FYI)

[gitpushdashf]

That sounds good. Thank you!

[google-ml-butler]

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

[gitpushdashf]

Why was this closed? I don't see a 2.6.0 release.

[ymodak]

TF 2.6 has now released https://github.com/tensorflow/tensorflow/releases/tag/v2.6.0
Perhaps we can close this thread now.

[gitpushdashf]

Working for me, thank you!

[google-ml-butler]

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