I forgot about the trouble using '<' in pip dependencies on the CLI.
This time I verified the build works.

PiperOrigin-RevId: 318332763
Change-Id: I347aee8121464232222e72d89409c600159ee80c

See #40688.

PiperOrigin-RevId: 318122157
Change-Id: Ief46c5610f3aaf0cdd7d43ce1a10d6d87e8e8e01

PiperOrigin-RevId: 315483879
Change-Id: I4be29ffa3c425f01c28dc8c95250209a2b0bada4

PiperOrigin-RevId: 306572687
Change-Id: I49448295cdd088eb114a7d518dd0db1822f2d7a6

Resolves #38247

PiperOrigin-RevId: 305114966
Change-Id: I83d00699783f9827ab526533fdeaaa3012086e66

PiperOrigin-RevId: 305090774
Change-Id: If65c6d0957f3da84222eb86f9d38a9146b12c7ed

This completes the plan announced in https://groups.google.com/a/tensorflow.org/d/msg/developers/5ZZG7M5E6FQ/zKv6uFn1EgAJ, removing all Python 2 considerations from our Dockerfiles. There are no longer any -py3 tags, as all the images are running Python 3.

Resolves #36971 and #37422. We're not able to dedicate the time to maintaining multiple versions of Python 3 at this time, unfortunately.

PiperOrigin-RevId: 304650599
Change-Id: Ib2b21e809f242fe914fb407381d8eaaa20f2ef99

This is necessary to start using bazel's experimental cc_shared_library support.
This change is part of the refactoring described in https://github.com/tensorflow/community/pull/179

PiperOrigin-RevId: 294736824
Change-Id: I8d13c71545cf2f2edd47818b40fa413e3e9c3254

This is necessary to start using bazel's experimental cc_shared_library support.
This change is part of the refactoring described in https://github.com/tensorflow/community/pull/179

PiperOrigin-RevId: 288808507
Change-Id: Ie78fb4ff8dad128ebef280037cce4d3c4f42addc

There seems to be a regression (or some change) in cublas 10.2.2.89-1
(the latest as of this commit) which prevents cublas from initializing
properly. This change pins to a known-working version of libcublas10.

See https://github.com/tensorflow/tensorflow/issues/9489#issuecomment-562394257.
Resolves https://github.com/tensorflow/tensorflow/issues/35364.

Resolves the warning noted in https://github.com/tensorflow/tensorflow/issues/35364

TensorFlow 2.1 was been upgraded to CUDA 10.1. This change (finally)
updates the non-devel GPU dockerfiles as well. Additionally, it moves
the horovod partial specs into their own key, because TensorFlow's CI
uses the "ubuntu" keys in CI, when horovod packages are not available.

This change makes the dockerfile build tests pass. They have been broken
ever since CUDA 10.1, and I fixed them by removing dependency on the old
ci_build configure scripts and by updating TensorRT to version 6.

Unfortunately, TensorRT builds seem to be broken at head (see
https://github.com/tensorflow/tensorflow/issues/35115), so I've had to
disable TensorRT for the build tests. However, this should make the
devel images able to build regularly again.

the implementation of Keras Applications across tf.keras and multi-backend
Keras.

PiperOrigin-RevId: 285302893
Change-Id: I06084c6827df651c68547c488e9fcad908d36796

the implementation of Keras Applications across tf.keras and multi-backend
Keras.

PiperOrigin-RevId: 285107804
Change-Id: I8cb04dc0bb665d2b51c3f41a23c2226d2ac4a29f

the implementation of Keras Applications across tf.keras and multi-backend
Keras.

PiperOrigin-RevId: 285084218
Change-Id: I27f574cae7476dda0161759470663ee89f3b4df6

PiperOrigin-RevId: 279959797
Change-Id: Ic5785033bc735d388f368aebee891160781fb193

Besides upgrading the Bazel version, we also refactored all build scripts to use rbe options in .bazelrc file.
In order to migrate for https://github.com/bazelbuild/bazel/issues/7480, we have to specify the complete strategies list in .bazelrc file.

PiperOrigin-RevId: 275459466
Change-Id: Iaec997da7862245955a36ebb1018d901f61c591d

While working on some Kubeflow Pipelines automation
we noticed the gpu based tf images had curl installed
while the non-gpu didn't which was causing some issues.

This consistently make curl available on all images.

PiperOrigin-RevId: 261857381

as per README contributing guidelines

Python 2-based images have been failing to build, and I think it's
because 'future' is missing. This PR resolves that, safe_loads yaml to
quiet a warning about safety, and rewords a few things in
CONTRIBUTING.md.

Add Dockerfile partials to support Mkl + MPI + Horovod; Remove trailing whitespace from python.partial.Dockerfile

PiperOrigin-RevId: 249052565

PiperOrigin-RevId: 248608540

This upgrades all TF official docker images aside from custom-op
(which does not use this partial-related system) to use Ubuntu 18.04,
up from 16.04.

Resolves #28495, #22292.

PiperOrigin-RevId: 246887169

PiperOrigin-RevId: 246550182

PiperOrigin-RevId: 243017772

Besides upgrading the Bazel version, this change also does:
1. Rename --experimental_remote_platform_override to --remote_default_platform_properties
2. Use --experimental_windows_native_test_wrapper in Windows build because one test was timeout due to a bug in Bazel's shell test wrapper in 0.23.2. We will enable native test wrapper on Windows soon.

We have to make the above two changes in this CL because they are not supported with the current Bazel 0.19.2.

PiperOrigin-RevId: 240088268

PiperOrigin-RevId: 235418348

This disables TensorRT in the devel builds, which currently can't build
in Docker's environment for some reason, and fixes some additional typos
in the GPU devel build setup.

FYI @tjakob, @tfboyd.

Due to an incredibly obtuse quirk of the Dockerfile language, the GPU
images weren't getting built properly after the recent ARG changes.

Add support for ppc64le dockerfiles with newest assembler changes.