[Dist] Disable -lto option.

Some Tizen distributors use -lto, which breaks Tensorflow-Lite.
Override such an option by adding -lno-lto.

Change-Id: Ib3bda7c020be0480d2603b124519a7ebab9904d1
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Remove tf-normal dependencies.

Enable tf-lite only, remove tf-normal build deps.

Change-Id: I4edcdc644a1ae353215715611a210bdb5b1d44d4
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Build TF-Lite only

1. That's what we need in Tizen 5.5 for now.
2. TF CMake Build is too much broken. I've resolved all build & runtime-link errors; however, now I'm seeing much harder runtime memory issues with CMake TF build. Until we get BAZEL+OpenJDK8 in Tizen, let's hold tensorflow.

Change-Id: Iaba1de07f541a603c832ef6ba7a360858e69307a
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Handling double-quote in debian-way

It has a strange parsing rule.

Change-Id: I46ff25d87d85122a8d80839cb1fe3cf03d5f9382
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Add cflags for debian tf-lite build

TF Lite wans some additional build flags.

Change-Id: I32ba9243a42257040c83b9be3b30a14bdd84279c
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Fix incorrect path

Path to neon2sse is incorrect.

Change-Id: Iba6e3c7f49e9d30c3a48f6570db32e49392baa50
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Add -fpermissive

It appears that it does not get flags from command line.

Change-Id: Ic49fc56784a9c65aeac0ea048f51407bd3b5a914
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] array_ops.h is generated by tf_cc_ops

Add tf_cc_ops in the dependency lists.

Change-Id: Ic491e94d433e0e3ec274cbcf2c15ca343da8cd00
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] Another array_ops error after shared lib

[ 1814s] /home/abuild/rpmbuild/BUILD/tensorflow-v1.13.1/tensorflow/core/grappler/optimizers/data/vectorization/cwise_op_vectorizer.cc:18:41: fatal error: tensorflow/cc/ops/array_ops.h: No such file or directory
[ 1814s]  #include "tensorflow/cc/ops/array_ops.h"
[ 1814s]                                          ^
[ 1814s] compilation terminated.
[ 1814s] CMakeFiles/tf_core_cpu.dir/build.make:3062: recipe for target 'CMakeFiles/tf_core_cpu.dir/home/abuild/rpmbuild/BUILD/tensorflow-v1.13.1/tensorflow/core/grappler/optimizers/data/vectorization/cwise_op_vectorizer.cc.o' failed
[ 1814s] make[2]: *** [CMakeFiles/tf_core_cpu.dir/home/abuild/rpmbuild/BUILD/tensorflow-v1.13.1/tensorflow/core/grappler/optimizers/data/vectorization/cwise_op_vectorizer.cc.o] Error 1
[ 1814s] make[2]: *** Waiting for unfinished jobs....
[ 1819s] CMakeFiles/Makefile2:2692: recipe for target 'CMakeFiles/tf_core_cpu.dir/all' failed
[ 1819s] make[1]: *** [CMakeFiles/tf_core_cpu.dir/all] Error 2
[ 1819s] Makefile:129: recipe for target 'all' failed
[ 1819s] make: *** [all] Error 2
[ 1819s] error: Bad exit status from /var/tmp/rpm-tmp.38NlR1 (%build)
message("STOPPING
[ 1819s]

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] Fix undefined symbol of SessionRef

Fix:
(gst-plugin-scanner:3296): GStreamer-WARNING **: Failed to load plugin '/home/abuild/rpmbuild/BUILD/nnstreamer-0.2.1/build/ext/nnstreamer/tensor_filter/libnnstreamer_filter_tensorflow.so': /lib64/libpywrap_tensorflow_internal.so: undefined symbol: _ZN10tensorflow10SessionRefC1EPNS_7SessionE

Change-Id: I46762a797798b8415d8f3f6af596904bfb6b6820
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Debian/Ubuntu] Avoid unicode build error

[ 3752s] /usr/src/packages/BUILD/tensorflow/core/kernels/unicode_ops.cc:65:39: error: invalid conversion from 'const UChar* {aka const short unsigned int*}' to 'const char16_t*' [-fpermissive]
[ 3752s]      const char16_t* buf = in.getBuffer();
[ 3752s]                                        ^
[ 3753s] CMakeFiles/tf_core_kernels.dir/build.make:13670: recipe for target 'CMakeFiles/tf_core_kernels.dir/usr/src/packages/BUILD/tensorflow/core/kernels/unicode_ops.cc.o' failed

Change-Id: Iaaea748380a911a775e5c09fb1fa7a51bf31ecff
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Separate main tensorflow package

Separate tensorflow.deb from tensorflow-dev.deb

Change-Id: I3773d89f49469cc868e10a8096d3fefbd4f0403f
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] Aviod build error after build_shared enable

[ 1903s] /home/abuild/rpmbuild/BUILD/tensorflow-v1.13.1/tensorflow/core/grappler/optimizers/data/vectorization/cwise_op_vectorizer.cc:18:41: fatal error: tensorflow/cc/ops/array_ops.h: No such file or directory
[ 1903s]  #include "tensorflow/cc/ops/array_ops.h"
[ 1903s]                                          ^
[ 1903s] compilation terminated.

Change-Id: Ic492db50d2d74611bdff955b57678fab96c5505e
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] TF's .so is missing absl symbols

Add dependencies to avoid:

(gst-plugin-scanner:29481): GStreamer-WARNING **: Failed to load plugin '/home/abuild/rpmbuild/BUILD/nnstreamer-0.2.1/build/ext/nnstreamer/tensor_filter/libnnstreamer_filter_tensorflow.so': /lib64/libpywrap_tensorflow_internal.so: undefined symbol: _ZN4absl16variant_internal21ThrowBadVariantAccessEv

Change-Id: Ic1b1fe2ac58537678a6a0c317441497a73423acd
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Enable shared-lib option of tensorflow

Let's enable shared_lib mode.

Change-Id: I5d41bfd852f94b83dcd28bf29ea5ad2322f118bb
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Fix rule path error

There was a missing deliminator, /.

Change-Id: I8ea784929c187ea2947fc9831cda6aeab90386e9
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] Fix undefined symbol of TFE_OpSetAttrBoolList

Fix:

(gst-plugin-scanner:2272): GStreamer-WARNING **: Failed to load plugin '/home/abuild/rpmbuild/BUILD/nnstreamer-0.2.1/build/ext/nnstreamer/tensor_filter/libnnstreamer_filter_tensorflow.so': /lib64/libpywrap_tensorflow_internal.so: undefined symbol: TFE_OpSetAttrBoolList

(gst-plugin-scanner:2272): GStreamer-WARNING **: Failed to load plugin '/home/abuild/rpmbuild/BUILD/nnstreamer-0.2.1/build/gst/nnstreamer/libnnstreamer.so': /lib64/libpywrap_tensorflow_internal.so: undefined symbol: TFE_OpSetAttrBoolList

Change-Id: I78b357d00bbf5d2cd57fb666f9885db9d8d4e965
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Enable debian build for Ubuntu

Let it not use systemlib for protobuf/grpc.
Ubuntu xenial/bionic has old versions of them.

Change-Id: Icb3018993f2f54b74226b16c2daea6a7e089e996
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] Fix regression on non-systemlib build mode for protobuf/grpc

Fix regression on non-systemlib builds.

Change-Id: Ie7912aba09d30bbf14557019d5d56b932fa1161f
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Revert "[Dist] Remove grpc/protobuf/zlib from dist cache"

Ubuntu Xeinal/Bionic has older versions of GRPC/Protobuf that are
not compatible with tensorflow 1.13. Use these archives for them.

This reverts commit d5bd25d162d3b0834aaff7b0000963b2416715b0.

[Dist] Remove dependency on grpc-static

Let's use shared library of grpc in Tizen.

Change-Id: Ief4cf02f13ab24019a80e137e58cc3dc596d04d9
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] If GRPC is disabled, don't use it

TF_CORE_FRAMEWORK has been keep using it if it's NOT WIN32.

Change-Id: Ia3c9885459a1e906a8bffaad2ebdf0751ca62911
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Ubuntu] Fix rule file for pdebuild

With this update, GRPC dependency is resolved dynamically.
Fixed other potential issues and realigned install scripts.

Change-Id: I798649f1789650c01c64a867c736bf384ab6d16a
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[TFLite/Make] Fix shell portability

Explicitly request bash for bash syntax.

Change-Id: I7cb6d0b34c393258059d5ebd56415098ab5f8933
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Ubuntu] Disable GRPC in xenial. Update GRPC for bionic

GRPC in xenial is too old. Disable it.

Change-Id: Ie9659d76797f468168a18538c052b6751a74d53c
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Remove grpc/protobuf/zlib from dist cache

We now build with system's shared library for grpc/protobuf/zlib.

Change-Id: I9cd999363015f17bdf38d53f0fd1a1d8fd9bf2e4
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] x64/tf-lite breaks nnstreamer

There are erros at nnstreamer:

[   15s] c++  -o ext/nnstreamer/tensor_filter/libnnstreamer_filter_tensorflow-lite.so 'ext/nnstreamer/tensor_filter/584bea7@@nnstreamer_filter_tensorflow-lite@sha/tensor_filter_tensorflow_lite.c.o' 'ext/nnstreamer/tensor_filter/584bea7@@nnstreamer_filter_tensorflow-lite@sha/tensor_filter_tensorflow_lite_core.cc.o' -Wl,--no-undefined -Wl,--as-needed -shared -fPIC -Wl,--start-group -Wl,-soname,libnnstreamer_filter_tensorflow-lite.so -O2 -g2 -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector -Wformat-security -fmessage-length=0 -Wl,--as-needed -feliminate-unused-debug-types --param=ssp-buffer-size=4 -fdiagnostics-color=never -m64 -march=nehalem -msse4.2 -mfpmath=sse -fasynchronous-unwind-tables -fno-omit-frame-pointer -g -std=gnu++11 gst/nnstreamer/libnnstreamer.so /usr/lib64/libtensorflow-lite.a -ldl /usr/lib64/libglib-2.0.so /usr/lib64/libgstreamer-1.0.so /usr/lib64/libgobject-2.0.so /usr/lib64/libgstbase-1.0.so /usr/lib64/libgstcontroller-1.0.so /usr/lib64/libgstvideo-1.0.so /usr/lib64/libgstaudio-1.0.so -lm -ldl /usr/lib64/liborc-0.4.so -Wl,--end-group -pthread '-Wl,-rpath,$ORIGIN/../../../gst/nnstreamer' -Wl,-rpath-link,/home/abuild/rpmbuild/BUILD/nnstreamer-0.2.1/build/gst/nnstreamer
[   15s] /usr/lib64/gcc/x86_64-tizen-linux-gnu/6.2.1/../../../../x86_64-tizen-linux-gnu/bin/ld: /usr/lib64/libtensorflow-lite.a(interpreter.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a shared object; recompile with -fPIC
[   15s] /usr/lib64/gcc/x86_64-tizen-linux-gnu/6.2.1/../../../../x86_64-tizen-linux-gnu/bin/ld: /usr/lib64/libtensorflow-lite.a(register.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a shared object; recompile with -fPIC

Change-Id: I08b2bcb9920d430171325957ab9f70cabea40ccb
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Install absl headers. Exclude duplicated.

TF headers require ABSL headers. Install them.
Exclude duplicated hedares of tf-lite from tf's devel.

Change-Id: I7e4a255383484e886857d2d2e33250cf7f5f42f9
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

CMake Fix for GRPC Shared Build

It has other .pc files

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

WIP: Fix cmake errors at ld step.

Change-Id: Ic1fa92de1cd31d0a06e707042988c277dd69e2df
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

WIP

Change-Id: Ia9fc23681c675f59a18a6d3309db88a5750d5641
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Add build-deps, use systemlibs.

Apply changes from Tizen build

Change-Id: I7c05a236d0c4a593ad105b3b39db08ab4dc509d4
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] use Tizen's zlib, protobuf, grpc

Let's use the system's protobuf, grpc, and zlib.
And do not export headers of protobuf from tensorflow.

Change-Id: I6ad41fd1a15b6d2651a94f5a098114000520f456
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[CMake] Allow to use pre-installed protobuf

If the system already has protobuf, let's use it without
downloading one.

Change-Id: Iaa2e722f0a5e4170422d66579f56be29dee01249
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Pkgconf] Fix nnstreamer-tensorflow build error

Let it refer to /usr/include/tensorflow.

Change-Id: I3a4f691cf2a059e4483b8e374125f00b31bd6834
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Cmake/external/png fix install path inconsistency

use the same path for lib install paths

Change-Id: I4fc42af2745c9e3eaeaf628c34b005d686268184
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist/Debian] Enable 1.13 for debian build

WIP

Change-Id: Ieed6c60cd957d42a5a1229eabe1ea255e9c400fd
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Update pkgconfig data

Fix pkgconfig data for tensorflow dev users.

Change-Id: I05718d5a733e2628a330a1226df6c74ecaf5b406
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Tf-lite devel package update

Fix build errors of nnstreamer/tizen

Change-Id: If561a7b720f874286f3ac4bb52edc04aaa54ae61
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[DIST] Rewrite install scripts for 1.13

Tf-lite install checked with i586

Fixes for TF w/ other arch will follow

Change-Id: I1fe831c63373d4bd669dac59231ab0bddd3c52a4
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Fix tensorflow-lite build

GBS successfully builds tensorflow-lite with this commit

Next TODO: rewrite install scripts for packaging

Change-Id: I3232d291a22ee47892f0a1396422420912da3f14
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] enable i586.

TODO: Find out tf-lite only build for 1.13

Change-Id: If751ffc3e2ab1bd1a974506d1c05692385d64b5c
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] remove find -delete, which is no more necessary

Change-Id: Iba76d3344c75333fda42b8c84cd04d6e9996adcf
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Fix libpng link error

Fix error with 64bit systems:
[ 1088s] make[2]: *** No rule to make target 'png/install/lib64/libpng16.a', needed by 'proto_text'.  Stop.

Change-Id: Iaff40bd2819b3395939c9601da3cee2190d0e2a3
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Archive clean up & dep update

OPENSSL should not be required; but due to tensorflow
bug (build-related), it still requires openssl/rsa.h

Change-Id: I68a4c808c2c415100498856f2f421d2c09a6d06c
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

CMake: clean up archive files with cmake

There are dirty cmake caches in archives.
Clean them up.

Change-Id: Iafdbe246499afc4e75f920c43a4eaa60e044c91b
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Archive Update with make clean, removed eigen patch.

1. Eigen is pre-patched.
2. Archives cleaned up (WIP)

Change-Id: I99d85304957b13250c7ac6b56bedfd130d7f05e7
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

WIP. Experimenting with GBS

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Added build-dependency on ICU

Tensorflow uses ICU for unicode processing.

Change-Id: I3746811b7a3c25205cdcb172055e42fabb09fee7
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Added cmake dependency

There was a missing dependency that incurs
build errors stochastically.

tensorflow/core/grappler/optimizers/data/vectorization/cwise_op_vectorizer.cc:18:10: fatal error: tensorflow/cc/ops/array_ops.h: No such file or directory

Change-Id: I91b5923a90aab8080ea9cd559eb66e7c0b62f414
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Added archive files for tensorflow static build

Tensorflow still does the primitive and PITA of
downloading all external library source codes
and embeded them statically.

As we don't intend to refactor the whole build
structure of tensorflow, we added the archives.

Change-Id: I7cddfa8bf8b1d191b24ac13554b94ed44ef978f7
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

CMake Build Fixed. It Works!

CMake build success with network support

How to do:
$ pushd build
$ cmake ../tensorflow/contrib/cmake  -Dtensorflow_ENABLE_SSL_SUPPORT=OFF -Deigen_PATCH_FILE=/source/tizen/tensorflow/packaging/eigen.patch; make -j4

TODO
- Archive downloaded packages for GBS
- Package binary files properly
- Code clean and reformulate commits

Change-Id: I1f8d175f0b5518701e1e8c22077f2318e6c6ec58
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

CMake Dependency fix

Add array_ops dependency, which was missing
Removed a duplicated file inclusion, creating build errors.

Change-Id: I5c7de3d69d1c9d7fd1056f8c170344b42f4b8426
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Let CMake Load ICU

ICU is required by unicode-related codes

Change-Id: Ib01def46bb8318563be0fbfe6d84601f42db369c
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Eigen fix 2

use
build$ cmake ../tensorflow/contrib/cmake  -Dtensorflow_ENABLE_SSL_SUPPORT=OFF -Deigen_PATCH_FILE=${ABSOLUTEPATH}/packaging/eigen.patch

Change-Id: I5e56260e9318f14df72009365b18d9a9870f8f6c
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Update Eigen Fix for Reusability

use
build$ cmake ../tensorflow/contrib/cmake  -Dtensorflow_ENABLE_SSL_SUPPORT=OFF -Deigen_PATCH_FILE=../packaging/eigen.patch

Change-Id: Iaab17b57a3af1737173c306a5351c5972e20450d
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

WIP: Fix Eigen Errors

Apply https://bitbucket.org/eigen/eigen/commits/88fc23324517/

Change-Id: I7ca808e5f91ed7e65d9c1a7d80cb943002c6d682
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

WIP: ABSL Error and Array_ops Error Fixed

Change-Id: I0e94da6128d841ca9c86cbbf264006cc2cb7035f
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

WIP2 Fixingup ABSL

Change-Id: Icb539d8789ee87f426e544a248a6ab21d3b1f927
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

WIP

Change-Id: Ifad90cf9964e63277c57fa9c2fd27c4eab17f4e6
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

[Dist] Rebased porting patches for 1.13.1

Selectively rebased and merged.

Change-Id: I77ac6ece2f9496adfa99b6dafa6d3489b828cf93
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Imported Upstream version 1.13.1

Imported Upstream version 1.12.0

Imported Upstream version 1.9.0

Merge pull request #20455 from lygztq/r1.9

fix bug in maxout function

Merge branch 'r1.9' into r1.9

Merge pull request #20593 from av8ramit/update_version_final

Updating the version to 1.9.0 official.

Merge branch 'r1.9' into update_version_final

Merge pull request #20596 from av8ramit/cp_disable_gcp

Remove unused gcp and hdfs config flags, as these are on by default now.

Remove unused gcp and hdfs config flags, as these are on by default now.

PiperOrigin-RevId: 202753310

Merge pull request #20535 from lamberta/r1.9-cherrypick-docs-no-merge

r1.9-cherry-pick-request: Docs updates for tutorials, guides, and notebooks

Merge branch 'r1.9' into r1.9

Updating the version to 1.9.0 official.

fix rc2

Update docs_src in 1.9 to match master

Update eager notebooks in 1.9 to match master

Merge pull request #20437 from bhack/patch-1

Advise batch_normalization with model_to_estimator

Merge pull request #20489 from tensorflow/yifeif-patch-2

Match for -path instead of -name

Match for path instead of name

fix bug in maxout function

The line "shape[axis] = -1" will make the shape wrong when dealing with batches with arbitrary sizes.

Advise batch_normalization with model_to_estimator

Exclude test sources from stream executor builds. (#20423)

PiperOrigin-RevId: 202423156

Merge pull request #20390 from case540/eigen_pip_cherry_pick

Fix install location for Eigen headers in pip build.

Potential fix for how pip installs headers used for custom ops.

These headers were recently moved from site-packages/external into
site-packages/tensorflow/include/external. Need to update setup.py
to reflect that.

Merge pull request #20351 from case540/update_version_19_rc2

Update version strings for TF 1.9.0-rc2.

Merge pull request #20324 from lamberta/r1.9-cherrypicks

1.9-cherry-pick-request: Doc updates for get_started page and nav

Update version strings for TF 1.9.0-rc2.

Fix checkpoints link in keras guide

Merge branch 'r1.9' into r1.9-cherrypicks

Merge pull request #20323 from case540/statusor_fix

Fix XLA header issue for pip build.

Merge pull request #20235 from rmlarsen/cherrypicks_5M8WT

Update Eigen version to commit e5e305a158a029f5b5f837bf821411a51439a970.

Fix leftnav for get_started

Remove section links that don't go anywhere.

Merge branch 'r1.9' into cherrypicks_5M8WT

Update eigen version to a fixed version for ppc64.

Fix small typo in RELEASE.md

Moving StatusOr from XLA to stream_executor.

PiperOrigin-RevId: 202179928

Merge pull request #20310 from petewarden/r1.9

 r1.9-rc2 cherry-pick request: Fix for RPi OpenBLAS compile issues, by pinning to known good version

Merge pull request #20290 from lamberta/r1.9-cherrypicks

1.9-cherry-pick-request: Docs updates for getting started and guide rename

Removed Python 3.5 updates for RPi