Potentially Incorrect Tensorflow 2.3 Builds for Windows in Anaconda Repository

[ZOUG]

There are a few unusual builds in the Anaconda repository for Tensorflow (TF) 2.3 that seem to make the tensorflow-gpu 2.3 installation unable to function properly (see here).

For TF 2.1, conda search tensorflow=2.1 yields the following results:

# Name                       Version           Build  Channel
tensorflow                     2.1.0 eigen_py36hdbbabfe_0  pkgs/main
tensorflow                     2.1.0 eigen_py37hd727fc0_0  pkgs/main
tensorflow                     2.1.0 gpu_py36h3346743_0  pkgs/main
tensorflow                     2.1.0 gpu_py37h7db9008_0  pkgs/main
tensorflow                     2.1.0 mkl_py36h31ad7c1_0  pkgs/main
tensorflow                     2.1.0 mkl_py37ha977152_0  pkgs/main

In contrast, conda search tensorflow=2.3 yields the following:

# Name                       Version           Build  Channel
tensorflow                     2.3.0 mkl_py37h04bc1aa_0  pkgs/main
tensorflow                     2.3.0 mkl_py37h10aaca4_0  pkgs/main
tensorflow                     2.3.0 mkl_py37h3bad0a6_0  pkgs/main
tensorflow                     2.3.0 mkl_py37h48e11e3_0  pkgs/main
tensorflow                     2.3.0 mkl_py37h856240d_0  pkgs/main
tensorflow                     2.3.0 mkl_py37h936c3e2_0  pkgs/main
tensorflow                     2.3.0 mkl_py37h952ae9f_0  pkgs/main
tensorflow                     2.3.0 mkl_py37he40ee82_0  pkgs/main
tensorflow                     2.3.0 mkl_py37he70e3f7_0  pkgs/main
tensorflow                     2.3.0 mkl_py38h1fcfbd6_0  pkgs/main
tensorflow                     2.3.0 mkl_py38h37f7ee5_0  pkgs/main
tensorflow                     2.3.0 mkl_py38h3c6dea5_0  pkgs/main
tensorflow                     2.3.0 mkl_py38h46e32b0_0  pkgs/main
tensorflow                     2.3.0 mkl_py38h637f690_0  pkgs/main
tensorflow                     2.3.0 mkl_py38h8557ec7_0  pkgs/main
tensorflow                     2.3.0 mkl_py38h8c0d9a2_0  pkgs/main
tensorflow                     2.3.0 mkl_py38ha39cb68_0  pkgs/main
tensorflow                     2.3.0 mkl_py38hd19cc29_0  pkgs/main

These TF 2.3 builds do not follow the naming conventions of previous tensorflow builds (i.e., mkl_ prefix is only used for mkl versions). Moreover, for some of these tensorflow builds, their dependent _tflow_select options and tensorflow-base builds do not seem to match. Details of selected suspicious builds are shown in the following.

tensorflow 2.3.0 mkl_py37h10aaca4_0
-----------------------------------
file name   : tensorflow-2.3.0-mkl_py37h10aaca4_0.conda
name        : tensorflow
version     : 2.3.0
build       : mkl_py37h10aaca4_0
...
md5         : e20e4d0681f40a24cf19fc43d8b844ff
timestamp   : 2020-12-01 11:47:04 UTC
dependencies:
  - _tflow_select 2.3.0 gpu
  - python 3.7.*
  - tensorboard >=2.3.0
  - tensorflow-base 2.3.0 eigen_py37h17acbac_0
  - tensorflow-estimator >=2.3.0

tensorflow 2.3.0 mkl_py38h8557ec7_0
-----------------------------------
file name   : tensorflow-2.3.0-mkl_py38h8557ec7_0.conda
name        : tensorflow
version     : 2.3.0
build       : mkl_py38h8557ec7_0
...
md5         : 18c2e5d3ac7c2c15a7ec773fb24cf63f
timestamp   : 2020-12-01 11:41:24 UTC
dependencies:
  - _tflow_select 2.3.0 gpu
  - python 3.8.*
  - tensorboard >=2.3.0
  - tensorflow-base 2.3.0 eigen_py38h75a453f_0
  - tensorflow-estimator >=2.3.0

[katietz]

Yes, tensorflow 2.3.0 was broken in some aspects. We might revisit it, but I did now the upgrade to 2.4.1 and fixed along this all the issues we had for older tensorflow versions.
I uploaded to my private channel the rc for linux-64 for testing. Getting some feedback here would be nice.

[ZOUG]

Yes, tensorflow 2.3.0 was broken in some aspects. We might revisit it, but I did now the upgrade to 2.4.1 and fixed along this all the issues we had for older tensorflow versions.
I uploaded to my private channel the rc for linux-64 for testing. Getting some feedback here would be nice.

@katietz Great ! Where is your private channel? Will there be builds available for Windows?

[katietz]

@ZOUG Sorry, missed that. It is the channel 'ktietz'.