| Commit message (Collapse) | Author | Age |
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| | |
checkpointing.
Uses a more recursive/functional tracking style which avoids numbering layers. Maybe this is too magical and we should adapt tf.keras.Sequential first? Let me know what you think.
PiperOrigin-RevId: 190282346
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* Fix MKL build on MacOS
* Fix invalid return type, user_memory_->set_data_handle(buf) returns void.
* Fix type mismatch
* Implement downloading and linking of MKL repo on mac and windows
* Rework the way MKL libs are binded to avoid warnings and simplify code
* Switch to matching version of MKL-DNN
* Add mirror links
* Add renamed license file
* Add -fopenmp flag for linux mkl build as Slice op kernel is optimized using OpenMP
* Use clean_dep in the entire file for consistency
* Fix build. Constants are declrared as static in header outside the class scope now
* Reworked fix for `-openmp` flag using the suggestion from @gunan
Original version here https://github.com/gunan/tensorflow/commit/b4d39c5226e8bbafe084b1dfc35f4644f2d906c2
|
| |
| |
| |
| | |
PiperOrigin-RevId: 190161440
|
|\| |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Previously we only allowed to download clang when doing GPU builds.
The added skylark files use bazel's autoconf scripts, which were only added in
0.10.0. To provide nice error message for older versions of bazel
(i.e. 'version is less than 0.10' vs 'can't load @bazel_tools/cpp/...'), we
move the bazel version check into WORKSPACE file from workspace.bzl.
PiperOrigin-RevId: 190050798
|
|\ \
| | |
| | | |
SYCL with ComputeCpp: local_config_sycl has multiple matches
|
| | |
| | |
| | |
| | | |
PiperOrigin-RevId: 189945839
|
|\ \ \
| | |/
| |/| |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
By disabling openmp when building with clang.
If we want to enable openmp with clang, we'll probably have to have libomp as
an explicit dependency.
This fixes a breakage found by OS CI:
https://ci.tensorflow.org/view/Nightly/job/nightly-matrix-linux-gpu-clang/215/
PiperOrigin-RevId: 189644968
|
| | |
| | |
| | |
| | | |
PiperOrigin-RevId: 189258641
|
| | |
| | |
| | |
| | | |
PiperOrigin-RevId: 189231636
|
| | |
| | |
| | | |
i.e. without this modification Tensorflow under Nvidia Jetpack 3.2 is not compiling. Else condition in if-then-else block generates a path "%s/../include" which results in "/usr/lib/include" but it must be "/usr/include/aarch64-linux-gnu"... So this fix targets ARM architectures and with this fix, Tensorflow 1.6rc compiles fine with TensorRT support.
|
| | | |
|
|\| | |
|
| | |
| | |
| | |
| | |
| | |
| | | |
corresponding script.
PiperOrigin-RevId: 188324090
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* Fix build issue with KafkaDataset
This fix tries to address the issue raised in 17210 where
error of `NotFoundError: Op type not registered 'KafkaDataset' in binary.`
returned from kafka ops. The issue was that the inclusion of kafka ops
was removed due to the conflict merge from the other PR. This fix
fixes the issue.
This fix fixes 17210.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Change `import readers.Dataset` to `import dataset_ops.Dataset`,
due to the changes in some other places.
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Fix library dependency issues in bazel
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Add dependency to bazel rules
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Add license to lib and pip package
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Remove unneeded changes in bazel
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Address review feedback
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Fix sanity check
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Add zlib dependency and include path
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Add copts in bazel to address the discrepancy in clang and gcc
Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
|
|\| |
| | |
| | |
| | | |
o Fixed simple merge issue in tf/contrib/timeseries/python/timeseries/BUILD
|
| |/
|/|
| |
| |
| | |
Previously, we put python.lib in data attribute of a cc_library and
manually added the link option. That caused the build to be
non-hermetic. This change fixed the problem.
|
| |
| |
| |
| |
| |
| | |
This detects negativo17's CUDA packages for Fedora.
PiperOrigin-RevId: 187923472
|
|\| |
|
| |
| |
| |
| | |
PiperOrigin-RevId: 187643585
|
|\| |
|
| |
| |
| |
| | |
PiperOrigin-RevId: 187191730
|
|\| |
|
| |
| |
| |
| |
| |
| | |
* updated ShuffleIndex because of API change
PiperOrigin-RevId: 186822637
|
| |
| |
| |
| | |
PiperOrigin-RevId: 186674197
|
|\| |
|
| |
| |
| |
| | |
PiperOrigin-RevId: 186592198
|
| |
| |
| |
| | |
PiperOrigin-RevId: 186073337
|
|\| |
|
| |
| |
| |
| |
| |
| | |
in inference output.
PiperOrigin-RevId: 185939805
|
|\| |
|
| |
| |
| |
| |
| |
| | |
* Also de-flake a test.
PiperOrigin-RevId: 185637742
|
|\ \
| | |
| | | |
Introducing TensorRT operator
|
|\ \ \
| | |/
| |/| |
|
| | |
| | |
| | |
| | |
| | |
| | | |
an Orc API change.
PiperOrigin-RevId: 185185088
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The same patch was sent via PR to cub upstream:
https://github.com/NVlabs/cub/pull/125
PiperOrigin-RevId: 184975304
|
|\| | |
|
| | |
| | |
| | |
| | | |
PiperOrigin-RevId: 184897758
|
| | |
| | |
| | |
| | | |
This reverts commit d0c2b84c2d6cd3c235671e67f22e3639b4b2b54c.
|
| | | |
|
| | |\
| |_|/
|/| |
| | | |
tensorrt to fix some of the failed tests.
|
| | |
| | |
| | |
| | | |
(#16580)
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
|\ \ \
| | | |
| | | | |
Branch 184220615
|