~nnstreamer/tensorflow/+git/trunk:sandbox/semun/tensorflow-lite_1.12

Last commit made on 2019-02-22
Get this branch:
git clone -b sandbox/semun/tensorflow-lite_1.12 https://git.launchpad.net/~nnstreamer/tensorflow/+git/trunk

Branch merges

Branch information

Name:
sandbox/semun/tensorflow-lite_1.12
Repository:
lp:~nnstreamer/tensorflow/+git/trunk

Recent commits

3fbef84... by Semun Lee <email address hidden>

Fix build breaks for Tizen

Change-Id: I1647da1539e9060585a567ab030b42b54fede6bc
Signed-off-by: Semun Lee <email address hidden>

abc5c76... by Chunseok Lee <email address hidden>

Remove __ANDROID__ ifdef from main shim header

- remove __ANDOIRD__ checker

Change-Id: I285b154b2b6f782e92e47acc16d840f310ff0f34
Signed-off-by: Chunseok Lee <email address hidden>

352e4e8... by Chunseok Lee <email address hidden>

Remove andoird verison check from nnapi delegate

- On tizen, there is no way to avoid this routine check.

Change-Id: If63b6187073ba82578a64cf537796d12d5688419
Signed-off-by: Chunseok Lee <email address hidden>

11ca05b... by Chunseok Lee <email address hidden>

[cherrypick]add 4 ops into nnapi_delegate

- TRANSPOSE, L2_NORM, SQUEEZE, HASHTAABLE_LOOKUP

upstream commit:
1. HASHTAABLE_LOOKUP
1a22e9a608b43cbdf7d990d09c7d317ee3c57d8c
Author: A. Unique TensorFlower <email address hidden>
AuthorDate: Tue Aug 7 10:18:12 2018 -0700
Commit: TensorFlower Gardener <email address hidden>
CommitDate: Tue Aug 7 10:23:42 2018 -0700

Parent: 36a606e Merge pull request #20928 from kbsriram:tf-fill-grad
Merged: master r1.9 v1.7.0
Follows: v1.9.0-rc2 (1853)

Add NNAPI delegation support for MEAN, RNN, SVDF, LSTM and lookup functions.

PiperOrigin-RevId: 207736388

2. L2Normalization
32fe0302b0cf02d6cc3ae6cf67b233ad65c74bfe
Author: A. Unique TensorFlower <email address hidden>
AuthorDate: Mon Jul 23 08:03:04 2018 -0700
Commit: TensorFlower Gardener <email address hidden>
CommitDate: Mon Jul 23 08:06:09 2018 -0700

Parent: 8647db8 Add C API for kernel info
Merged: master r1.9 v1.7.0
Follows: v1.9.0-rc2 (1102)

Delegate L2Norm to nnapi.

PiperOrigin-RevId: 205661557

3. TRANSPOSE
bc2674d09efbd87ae81ae41b81f1d152f37fac2a
Author: Jared Duke <email address hidden>
AuthorDate: Mon Jul 9 15:51:18 2018 -0700
Commit: TensorFlower Gardener <email address hidden>
CommitDate: Mon Jul 9 15:54:45 2018 -0700

Parent: 3eb314b Fix NVVM libdevice detection in open source build with CUDA 8.
Merged: master v1.7.0
Follows: v1.9.0-rc2 (255)
Precedes: v1.10.0-rc0 (350)

Add NNAPI transpose op support

PiperOrigin-RevId: 203845358

4. SQUEEZE
80bc59b99bca7f9bc167975bab1c295bc4793c9a
Author: Jared Duke <email address hidden>
AuthorDate: Wed Jun 27 09:40:26 2018 -0700
Commit: TensorFlower Gardener <email address hidden>
CommitDate: Wed Jun 27 09:44:48 2018 -0700

Parent: 67cd3f7 Add baseline arg to EarlyStopping
Follows: 0.5.0 (23767)
Precedes: v1.9.0-rc2 (12079)

Add NNAPI squeeze op support

PiperOrigin-RevId: 202323072

Change-Id: I5d80fb951026bc879edb920a349cc6d326e4ba03
Signed-off-by: Chunseok Lee <email address hidden>

b3fe381... by Sangjung Woo

deb: Enable tensorflow for Ubuntu

This patch enables the tensorflow debian package for Ubuntu.

Change-Id: Iaed77cfceb9fdb5e7b7fb6d0af9467ca958f92e9
Signed-off-by: jijoong.moon <email address hidden>
Signed-off-by: Sangjung woo <email address hidden>

fb1544c... by Sangjung Woo

Bugfix for ARM 32bit environment

* Add "-mno-unaligned-access" option to fix the sigbus because of
  Alignment problem
* Temporary disabled the 'DCHECK_LT' to avoid sigabrt when
  device.frequency() returns 0

Change-Id: Iad639d3ea92a2449ede646ef0fdc2836fa89e222
Signed-off-by: Sangjung woo <email address hidden>

6fcecf4... by Chunseok Lee <email address hidden>

Support tensorflow-lite single build

Full build takes too much time on developer's local machine.
This commit allows user to build tensorflow-lite package w/o tensorflow build

Change-Id: I5818ddb64a920027c4057615c6167f47cb6f4daf
Signed-off-by: Chunseok Lee <email address hidden>

a66d7d3... by Sangjung Woo

spec: Disable tensorflow package for i586 architecture

In order not to support the tensorflow package for i586 architecture,
this patch add a conditional block for %arm, aarch64 and x86_64.

Change-Id: I2b1a7d59393d3dd9aa0e587db08de6cfcb55d82f
Signed-off-by: Sangjung woo <email address hidden>

4ec7aeb... by Semun Lee <email address hidden>

Update to v1.9.0

Signed-off-by: Semun Lee <email address hidden>

39714ce... by Sangjung Woo

spec: Remove unnecessary BuildRequires section

In order to build this package on tizen.org infra, this patch removes
the unnecessary BuildRequires sections, which are added for internal
test.

Change-Id: Ic8b219c9f8e6910a1876b787b6abe9c0c4247357
Signed-off-by: Sangjung woo <email address hidden>