summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
narpra01 [Tue, 2 Oct 2018 13:35:53 +0000 (14:35 +0100)]
IVGCVSW-1920 Unittests for NHWC Normalization Workloads and Layer
Change-Id: Iea941c1747454f5a4342351e4e82b10ffb9ccbbd
James Conroy [Mon, 1 Oct 2018 08:15:19 +0000 (09:15 +0100)]
IVGCVSW-1931: Add data layout param for ResizeBilinear
* Added data layout parameter to ResizeBilinear
descriptor, in order to support NHWC.
Change-Id: Ifdbc4529127b7329a056d0a68e2e42b175aeea4a
arovir01 [Mon, 1 Oct 2018 16:08:59 +0000 (17:08 +0100)]
IVGCVSW-1913: Fix for ValidationTest.concat_float_3_relaxed
* Added RefPermuteFloat16Workload to serve as a fallback when CL
does not support the required permute configuration for FP16
* Move Half.hpp to armnnUtils as the utils library should not be
including private header files from the armnn library
Change-Id: Ibf0f698451e8406f7ed7cce470dab60b6d16361d
Matthew Bentham [Mon, 1 Oct 2018 16:22:32 +0000 (17:22 +0100)]
IVGCVSW-1951 Remove type templating from ClSplitterWorkload
Change-Id: Ib289ed5885e0e22ce6d73d7b3b3b103691187a34
David Beck [Fri, 28 Sep 2018 15:19:47 +0000 (16:19 +0100)]
IVGCVSW-1935 : inline code documentation for pluggable backends
Change-Id: I2a46e4dfffcc103bf228aa161bb2fba8f259076a
Derek Lamberti [Mon, 1 Oct 2018 16:10:05 +0000 (17:10 +0100)]
IVGCVSW-1823 Fix uninitialised member
Change-Id: I9b3aa1dd44318e25d520eaa35b99a07ac6b95335
Matthew Bentham [Mon, 1 Oct 2018 15:39:28 +0000 (16:39 +0100)]
IVGCVSW-1951 Remove type templating from ClSubtractionWorkload
Change-Id: Ieda5146d23227e9764f3f867003a13c630494093
Sadik Armagan [Mon, 1 Oct 2018 10:51:37 +0000 (11:51 +0100)]
IVGCVSW-1787 Add Support for Concatenation on TfLite parser
* Concatenation Parser function added to the TfLite Parser
Change-Id: I42a42cd765ea09a30841c66b1942b9e09a876b10
Derek Lamberti [Mon, 1 Oct 2018 08:28:57 +0000 (09:28 +0100)]
IVGCVSW-1824 Fix slow profiling of neon. (~50% reduced end-to-end time)
Change-Id: I58295c298934317a2b365887bd9f9f6705cd0a21
Matteo Martincigh [Mon, 1 Oct 2018 12:11:00 +0000 (13:11 +0100)]
IVGCVSW-1900 Removed unused file src/backends/ClWorkloads.hpp
Change-Id: I43d2c91fb53f6eb76ee37da73625ed79f602f04e
Matthew Bentham [Mon, 1 Oct 2018 10:32:48 +0000 (11:32 +0100)]
IVGCVSW-1207 - Remove typing from ClMultiplicationWorkload
Don't need this now as it uses the compute library validation function,
and all of the code for the supported types is identical.
Adds Uint8 support to Cl backend, and unit test cases.
Change-Id: I35d4edacc1aca241e95d1b19ae525a23d9513c99
David Beck [Fri, 28 Sep 2018 15:51:01 +0000 (16:51 +0100)]
IVGCVSW-1906 : make Android make and CMake files more similar
Change-Id: I15e689a6288b8476db7e67d7153899814c3cdf9e
David Beck [Wed, 26 Sep 2018 16:41:13 +0000 (17:41 +0100)]
IVGCVSW-1900 : CL backend folder structure
* moving backends/ClWorkloads to backends/cl
* and moving pure Cl workload related code to
backends/cl/workloads
Change-Id: I019a3c6b4da5e7a23074bf03fb057e63199ad129
Matteo Martincigh [Fri, 28 Sep 2018 13:14:12 +0000 (14:14 +0100)]
IVGCVSW-1863 Support NHWC for L2Normalization
* Added L2NormalizationDescriptor struct with m_DataLyaout member
* Updated all IsL2NormalizationSupported calls to take a descriptor
as an argument
* Updated L2NormalizationLayer to take a descriptor as an argument
!android-nn-driver:150116
Change-Id: I0459352d19cfd269bc864a70cf73910bf44fdc01
jimfly01 [Fri, 21 Sep 2018 11:09:51 +0000 (12:09 +0100)]
IVGCVSW-1882: Added ClPadWorkload
* added ClPadWorkload class
* Updated ClWorkloadFactory
* added IsPadSupportedCl function
* Updated build files
Change-Id: Iadccc340a0eae1ff30fab6441854acac7f8edc92
narpra01 [Fri, 28 Sep 2018 10:07:51 +0000 (11:07 +0100)]
IVGCVSW-1812 Adding Ref implementation and tests of MeanWorkloads
Change-Id: I6fb15c407024e3b91d5abf4513f8090be5821760
narpra01 [Thu, 27 Sep 2018 15:46:14 +0000 (16:46 +0100)]
IVGCVSW-1919 - data layout parameter for Normalization
Change-Id: I33dce72bb0f1e25425dc058d6213a7cdf56eecd2
David Beck [Mon, 24 Sep 2018 14:59:27 +0000 (15:59 +0100)]
IVGCVSW-1899 : Neon backend folder structure
armnn:149855
Change-Id: I26e8cf83422a65049386a5ebdb6d0001627aefaa
David Beck [Mon, 24 Sep 2018 12:18:27 +0000 (13:18 +0100)]
IVGCVSW-1898 : Ref backend folder structure
* Reference backend is renamed to backends/reference as per
https://confluence.arm.com/display/MLENG/Pluggable+backends
Change-Id: I27a13c274eb60995dfb459e3c49c0e2f60bcd32c
Nina Drozd [Thu, 27 Sep 2018 10:53:34 +0000 (11:53 +0100)]
IVGCVSW-1821 - taking out scale factor again as it's polluting the WallClockTimer interface
Change-Id: Ia90b709ddfff321dbc218add4ab19737f68a44bf
Nina Drozd [Fri, 21 Sep 2018 17:42:09 +0000 (18:42 +0100)]
IVGCVSW-1821 - update NEON workload utils to use timers in correct order, updated units used in NeonTimer
Change-Id: I593af42bd2930dd9d147354b706087e3ac260fe9
kevmay01 [Wed, 26 Sep 2018 09:21:55 +0000 (10:21 +0100)]
IVGCVSW-1927 Add Neon 8-bit FullyConnected support
Change-Id: Idf4cc7a9a7d3261b9eceb653b999257506cdae76
jimfly01 [Mon, 24 Sep 2018 11:47:14 +0000 (12:47 +0100)]
IVGCVSW-1924: Fixing clframework version update compile err
* Changed run_workloads function invocation to run_workloads_tagged
* Also ClMultiplication now supports QASYMM8
!builder:149820
!android-nn-driver:149821
Change-Id: Id76446aeccf5c0eaae2fd62da34c54a374cc4daa
Nikhil Raj [Tue, 25 Sep 2018 15:16:13 +0000 (16:16 +0100)]
IVGCVSW-1910 Add data layout parameter for DepthwiseConvolution
Change-Id: Ia14c9d8c0a38b669a7589e63d74424e398790e54
David Beck [Mon, 24 Sep 2018 09:46:38 +0000 (10:46 +0100)]
IVGCVSW-1921: move common Acl code to a separate folder
Change-Id: I400be8e7c0cc5a31eb9d2a7396da145d50d51b6e
Mohamed Nour Abouelseoud [Mon, 24 Sep 2018 12:30:09 +0000 (13:30 +0100)]
IVGCVSW-1881 Add PadLayer to ArmNN
Change-Id: Ib70883a20fa7b82bfb090aa98cc51a6d645b5722
Francis Murtagh [Mon, 24 Sep 2018 14:01:18 +0000 (15:01 +0100)]
IVGCVSW-1888 Plumb data layout parameter for Convolution2D
* Added the DataLayout parameter to the Convolution2dDescriptor
* Added the DataLayout parameter the Convolution2dQueueDescriptor
* Set the DataLayout on the Descriptor in CreateWorkload()
* Added overloaded factory methods for CreateTensorHandle()
* Updated BuildArmComputeTensorInfo() to take DataLayout parameter.
* Updated handles to take DataLayout parameter
* Updated (Cl/Neon)Convolution2dWorkloadValidate
* Updated (Cl/Neon)Convolution2dFloatWorkload
* Updated (Cl/Neon)Convolution2dUint8Workload
Change-Id: I8410668b3d727ca587bee66755cc4c4c78422f1f
David Beck [Fri, 21 Sep 2018 14:48:23 +0000 (15:48 +0100)]
IVGCVSW-1897 : Android support for pluggable backends
Change-Id: I9143485f542d58307a604248200a4ecf839ff39a
Matthew Bentham [Fri, 21 Sep 2018 14:16:53 +0000 (15:16 +0100)]
IVGCVSW-949 Further simplify InitializeArmComputeClTensorData
Remove the overload that takes a raw pointer and always pass a ConstTensor reference
Change-Id: I8fbff8dd5c260e7cdf4a3c51eb54cc5f3f6e874c
Matthew Bentham [Fri, 21 Sep 2018 09:29:58 +0000 (10:29 +0100)]
IVGCVSW-949 Simplify use of IntialiseArmComputeClTensorData
Change-Id: I556881e34f26e8152feaaba06d99828394872f58
David Beck [Wed, 19 Sep 2018 11:03:20 +0000 (12:03 +0100)]
IVGCVSW-1897 : build infrastructure for the src/backends folder
Change-Id: I7ebafb675ccc77ad54d1deb01412a8379a5356bb
Matthew Bentham [Thu, 20 Sep 2018 14:53:04 +0000 (15:53 +0100)]
IVGCVSW-949 Remove unnecessary includes
Change-Id: I923fd5d03e53187c18ac05e2cde90d43604f3821
Matthew Bentham [Thu, 20 Sep 2018 14:35:30 +0000 (15:35 +0100)]
IVGCVSW-949 Refactor - clean up includes of ClWorkloadUtils.hpp
Move ClWorkloadUtils.hpp into ClWorkloads and reduce the number of places that include it
Change-Id: Iac661be34c185d6208ca75509155b049c24293ca
Matthew Bentham [Mon, 17 Sep 2018 10:17:41 +0000 (11:17 +0100)]
IVGCVSW-949 Add 8-bit fully connected support
Change-Id: I0953bb8dbc4b76001f207e37c8c2742a6ebd888b
James Conroy [Tue, 18 Sep 2018 16:06:44 +0000 (17:06 +0100)]
IVGCVSW-1822: Fix CaffeYolo-Armnn error reporting
* Changed message to 'Prediction for test case <N>
is incorrect: Expected (<X>) but predicted (<Y>)'
Change-Id: Ie54b70683bf5bb26de2f2a6bcf4a162c99ba5474
Sadik [Wed, 19 Sep 2018 14:30:00 +0000 (15:30 +0100)]
IVGCVSW-1650 Add Support for Reshape layer on TF Lite parser
* Added Reshape operator support for the TfLite Parser.
Change-Id: I64a5650dac089905a402be4a9cb6032aa0d81f00
James Conroy [Wed, 19 Sep 2018 10:58:44 +0000 (11:58 +0100)]
IVGCVSW-1843: Edit comment to ref correct function
* Changed comment to reference the function
CreateArithmeticWorkloadTest, as the
original CreateSubtractionWorkloadTest
no longer exists.
Change-Id: Ief60e6f3684bd5539d8fb21dd09fc7ef8e8a92f5
jimfly01 [Wed, 19 Sep 2018 09:59:49 +0000 (10:59 +0100)]
IVGCVSW-1880: Add no-op factory implementations of CreatePad
* Added PadDescriptor and PadQueueDescriptor
* Factory method CreatePad on all backends returning null workloads
Change-Id: Ifeb50f9249ea31edfc3586d49b008ea1b5c59ae2
Derek Lamberti [Tue, 18 Sep 2018 15:02:25 +0000 (16:02 +0100)]
IVGCVSW-1883 Add support for different memory layouts
Change-Id: I6e9973bf25acad980fb4e96af8080ac829db0d28
Sadik Armagan [Mon, 17 Sep 2018 13:14:39 +0000 (14:14 +0100)]
IVGCVSW-1651 Add Support for Relu on TF Lite parser
* Added Relu and Relu6 Support for the TfLite Parser.
Change-Id: I3cc5e4922910e556f25b633eae6d2d361cea61b5
Matthew Bentham [Fri, 7 Sep 2018 09:40:55 +0000 (10:40 +0100)]
Fix for Github issue #63
Update FindBoost.cmake to version from Kitware/CMake Github cb800eb
Change-Id: I3988350a09d5431868d8776fd20fd4493a8dfb1d
David Beck [Fri, 14 Sep 2018 09:00:09 +0000 (10:00 +0100)]
IVGCVSW-1802 : fix typo
Change-Id: I1d5c558f6575e5d21c45412c8848e53392cae618
surmeh01 [Thu, 13 Sep 2018 09:23:42 +0000 (10:23 +0100)]
Updated the inception_v3 model taken from official Tf
!devenv:148227
Change-Id: I550c885c125f2fe12951e38d4157276562cb49ec
David Beck [Mon, 10 Sep 2018 13:47:28 +0000 (14:47 +0100)]
IVGCVSW-1802 : Neon Sub Workload
Change-Id: I1e914b046cd8fd25669390d477f101098fe0d476
narpra01 [Thu, 13 Sep 2018 10:07:48 +0000 (11:07 +0100)]
IVGCVSW-1813 - Add MeanLayer
* add MeanLayer functionalities
* modify MeanQueueDescriptor to use parameter
* add IsMeanSupported placeholder for all backends
Change-Id: Ic69a34a61df667849977aad9b38f9a01eef565b5
David Beck [Wed, 12 Sep 2018 15:00:08 +0000 (16:00 +0100)]
IVGCVSW-1832: add missing DIV quantized, broadcasting tests
Change-Id: I042b71e77d175163837bf39d5703ada08d25e532
David Beck [Wed, 12 Sep 2018 12:50:03 +0000 (13:50 +0100)]
IVGCVSW-1843 : remove duplicate code for Ref Arithmetic workloads
Change-Id: If94d7b7b06a8c4e2c155b2ab470604a8d20d1027
David Beck [Tue, 11 Sep 2018 14:21:14 +0000 (15:21 +0100)]
IVGCVSW-1843 : refactor ClAdditionWorkload and ClSubtractionWorkload
Change-Id: I0ca9f16217f8e32bb57a49b841611f10dabf021a
David Beck [Wed, 12 Sep 2018 15:00:08 +0000 (16:00 +0100)]
IVGCVSW-1804 : add subtraction layer to the public interface
Change-Id: I2c1b622b5199000c01af387bb994267c2b91f89c
surmeh01 [Wed, 12 Sep 2018 15:00:26 +0000 (16:00 +0100)]
Added and modified the vanilla tensorflow values for inception and mobilenet
Change-Id: Ib0d59015aa104909f3922f9749e3787e15c7c86e
David Beck [Tue, 11 Sep 2018 15:37:14 +0000 (16:37 +0100)]
IVGCVSW-1843 : replacing trivial arithmetic helpers
Change-Id: Iddf637694f1a3a7ef00f006a41b8044a35c7e73c
Nina Drozd [Mon, 10 Sep 2018 13:26:44 +0000 (14:26 +0100)]
IVGCVSW-1733 - set enableGpuProfiling in CreationOptions for Runtime if -e flag is set,
false by default, use this globally same as concurrent flag, removed -e option from RunCsvTest options,
this is passed in from main
Change-Id: I246c2c40b1a113b896be0d41aba528e79ecdba0d
narpra01 [Mon, 10 Sep 2018 16:35:27 +0000 (17:35 +0100)]
IVGCVSW-1831 - Add dimension check to MeanQueueDescriptor::Validate to check if the output dimension is correct from a given input and options.
Change-Id: Ibc15d9ea3151a7ba1935feafeb1843ee035e7f2e
David Beck [Fri, 7 Sep 2018 15:19:24 +0000 (16:19 +0100)]
IVGCVSW-1801 : Cl implementation for SUB
Change-Id: Ia2e1dda8653197454a50679d49020397f5327979
narpra01 [Mon, 10 Sep 2018 08:50:09 +0000 (09:50 +0100)]
IVGCVSW-1810 - Add no-op factory implementations of the CreateMean for all backend factories
Change-Id: Ic1054a32f59659d10f67a6b02527887ccf3c850f
Matthew Bentham [Mon, 10 Sep 2018 10:45:46 +0000 (11:45 +0100)]
IVGCVSW-1520 Re-enable quantised add
Change-Id: I1944242b67e8c4a0e54f1a59377a1d78633edb3f
surmeh01 [Wed, 5 Sep 2018 16:09:22 +0000 (17:09 +0100)]
IVGCVSW-1808 : Updated the network used from Official Tf(download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_224.tgz)
Change-Id: I3a07d7a161c8ac81c6a94cc0b3557ee7fa5e48e5
David Beck [Thu, 6 Sep 2018 15:46:34 +0000 (16:46 +0100)]
IVGCVSW-1803 : add Ref Subtraction layer
Change-Id: I4c019d626f9369245eca6d549bbe7a28e141f198
David Beck [Wed, 5 Sep 2018 14:00:38 +0000 (15:00 +0100)]
IVGCVSW-1804 : Add Subtraction layer types and placeholders
Change-Id: Ib9a477e5ce590df74ba05fece77258b9204f6523
Sadik Armagan [Wed, 5 Sep 2018 15:33:58 +0000 (16:33 +0100)]
MLCE-52 Where has PROFILING gone?
* Call the Print() function on the Profiler when profiling is enabled.
Change-Id: I4c06ce9bfa961ac8d81d02a516502337d1fbc968
David Beck [Wed, 5 Sep 2018 11:52:57 +0000 (12:52 +0100)]
IVGCVSW-1807 : change license text in file headers
All changes are the same:
//
// Copyright © 2017 ARM Ltd. All rights reserved.
-// See LICENSE file in the project root for full license information.
+// SPDX-License-Identifier: MIT
//
Change-Id: I37eae011411133663ca9d2b059714d92f8bf8e24
David Beck [Wed, 5 Sep 2018 12:36:46 +0000 (13:36 +0100)]
IVGCVSW-1807 : add license section to the Readme file
Change-Id: I3f6823d67436e75b1f451afc6f982777b6b51705
Les Bell [Mon, 3 Sep 2018 15:24:52 +0000 (16:24 +0100)]
IVGCVSW-1783 refactor add/mul/div layers
Change-Id: Ic043030455b3cf8ad7f92fd0a75175c1827a95f4
arovir01 [Fri, 31 Aug 2018 14:26:35 +0000 (15:26 +0100)]
IVGCVSW-1784: Rename float32 workloads for ACL
Change-Id: I98bdfe9cb12c663d1d5cfa456e2cc967d70ab22b
saoste01 [Fri, 31 Aug 2018 15:23:11 +0000 (16:23 +0100)]
IVGCVSW-1521: Updating TensorFlow README.md to contain realdiv operator
Change-Id: Iad169a279c9fc67bd8a44511a646dded8d4ba1aa
saoste01 [Tue, 28 Aug 2018 14:41:51 +0000 (15:41 +0100)]
IVGCVSW-1779: Updating TF Parser to include the RealDiv operator
Change-Id: I4b193d9119f5aaf41081335df043352953364c87
Francis Murtagh [Thu, 30 Aug 2018 16:18:37 +0000 (17:18 +0100)]
IVGCVSW-1786 - Division by zero
* Added logic for different cases of division by zero for ref
* Added DivisionByZeroTest
* Updated SelectiveComparer to compare NAN and INFINITY
Change-Id: Iec6f89264b17a0b03fad5d7ec4b2dafc31fea5df
Francis Murtagh [Wed, 29 Aug 2018 11:42:10 +0000 (12:42 +0100)]
IVGCVSW-1200 Division layer
*IVGCVSW-1772 Create QueueDescriptors
*IVGCVSW-1773 Add a CL implementation of the DivisionWorkload
*IVGCVSW-1774 Add Neon implementation of the DivisionWorkload
*IVGCVSW-1775 Add a Ref implementation of the DivisionWorkload
*IVGCVSW-1776 Add a Division Layer
* Added simple division unit tests with broadcasting
Change-Id: I05751fb7f868789f6c06f91e8d25e52b4f12ab5e
arovir01 [Tue, 28 Aug 2018 16:40:45 +0000 (17:40 +0100)]
IVGCVSW-1324: Make biases truly optional for Conv2d and DepthwiseConvolution
!android-nn-driver:145961
Change-Id: I039ab0adc61725859514246365b5e5b5fa603eaa
telsoa01 [Fri, 31 Aug 2018 08:22:23 +0000 (09:22 +0100)]
Release 18.08
surmeh01 [Thu, 5 Jul 2018 13:03:08 +0000 (14:03 +0100)]
Release 18.05.02
surmeh01 [Mon, 11 Jun 2018 16:12:45 +0000 (17:12 +0100)]
Release 18.05.01
Documentation fixes
Change-Id: Ice181dee80a8ddecf5acedecce01a063d12e3567
surmeh01 [Fri, 18 May 2018 15:31:43 +0000 (16:31 +0100)]
Release 18.05
surmeh01 [Thu, 29 Mar 2018 15:29:27 +0000 (16:29 +0100)]
Release 18.03
telsoa01 [Fri, 9 Mar 2018 14:13:49 +0000 (14:13 +0000)]
Release 18.02
Change-Id: Id3c11dc5ee94ef664374a988fcc6901e9a232fa6
projects / platform / upstream / armnn.git / log