Age | Commit message (Collapse) | Author |
|
* Changed ClL2Normalisation from using CLNormalizationLayer
to use CLL2NormalizeLayer to normalise along the channel axis
in either NCHW or NHWC format.
Change-Id: I399cbee408a277d1ef8c6c85ebcbd86d6c3e407b
|
|
Change-Id: Ibe826db4412c54aba42c45c155e7e5329861bdb1
|
|
Change-Id: Ie2f134fd51de919c50fa9d8d82bbb4ed4ee46469
|
|
Change-Id: I0b9818c3bc09101aa6563ea8e4a886483d3b09d1
|
|
* Ported Google's LSTM implementation to RefLstmFloat32Workload
* Fixed the code throughout because of an error in the docs around the
scratch buffer size
* Updated IsLstmSupported
* Added the unit tests
!android-nn-driver:127
Change-Id: I5577b7e39ca52df1a7f102a9b437df6aa99520b6
|
|
Change-Id: I812d09a869fe7c45ce71108fb675f3e2d974e2d9
|
|
Change-Id: I7cbfd5a6ab0746e0351b64a45617ed089992c1a9
|
|
stable test (memory-wise)
This prevents false-positives when running the tests on Master
Change-Id: I4348378f3689cb3ec98686a93315d2b88bb7bcc7
|
|
implementations
Change-Id: I906dcb8b4c9b491bf3a661f208c09d0ea1c0fa35
|
|
Change-Id: Idc0f94590566ac362f7e1d1999361d025cc2f67a
|
|
Change-Id: Iccc1272e59bd71c59f810a54d6d3742859789212
|
|
Change-Id: I4d02a6cca8e99d70a389ca3b0ce47fb7f15ed21f
|
|
* Removed the convertFromTfToArmnnFormat flag from the
CreateConstTensor and CreateConstTensorImpl functions
* Removed the SwizzleInDeswizzleOut function and its invocations
in the ParseConv2D, ParseDepthwiseConv2D and ParsePool functions.
Change-Id: Id935e28ab6eff4731f838af9490d016e108f6df6
|
|
* adding missing include <functional> into BackendRegistry.hpp
Change-Id: I7a4ba867a856c90a3099a782ff41b8a5fb9a9efb
|
|
BatchToSpaceNdLayer.
* added Unit test
Change-Id: I80f55e8c7afb39d96006c8dd027fc9683ea8182e
|
|
* L2Normalisation Tests now take the layout as an argument.
Change-Id: Ib393889eedb3a81c9a500da91593e90a18549c33
|
|
Change-Id: I93223c8678165cbc3d39f461c36bb8610dc81c05
|
|
* Fixed error in InferOutputShape implementation
* Added better error checking to the BatchToSpace implementation.
* Added defaults to the batchToSpace descriptors.
* Changed crops to be a vector of pairs to align with the SpaceToBatch implementation
Change-Id: Ib1c16d871f0898a1caeb6629c1fee6380a773e14
|
|
Change-Id: I56da4780f8f5fcef7ff01d232d5d61bf299364bf
|
|
Change-Id: I13c6728dbb60643d0e086d171225c5d802987f92
|
|
ACL fix the convolution issue.
Change-Id: If72f6f4e32567bcfa4af13836e411a3199ede010
|
|
Change-Id: Ieec34224b433e1d2f3bbe66632cd6016cac5498c
|
|
Change-Id: I27ffebdece6e68460931a44c15b9b029f9fce638
|
|
* add IBackendContext interface
* add ClBackendContext implementation
Change-Id: I13e4d12b73d4c7775069587675276f7cee7d630b
|
|
* Added quick workaround for the memory management errors in the
CL/Neon unit tests
Change-Id: I56250db462cdbdc1acc0a5824807d288fb0c1d11
|
|
also in IVGCVSW-2019 + IVGCVSW-2005 + IVGCVSW-1987 IVGCVSW-1642
Change-Id: I33809fd2f30aac248ec45f070f717b8dcf5de3cf
|
|
Change-Id: I48242425c6a6856e13ebcee1b140cbd2af94a3aa
|
|
This reverts commit d4dfa684941a21314b70593d01b0fc2167eebad4.
Change-Id: Id61ce69215505c3cf5d30ec2a7ec9127fb2554fc
|
|
DataLayout (3*3 Optimization)
Change-Id: Icfc137201c52e2c856076b7795572cc4ba75cc95
|
|
Change-Id: I663a0a0fccb43ee960ec070121a59df9db0bb04e
|
|
implementations
Change-Id: Ibd457f3a2d4342c4d6335bd3c471282a14ab6b14
|
|
* Modifies ResizeBilinear unit tests to use 2-Channel
tensor shapes for input and output data, to improve
test coverage when exercising NHWC data layout.
* Refactors unit tests to permute input and output
data when exercising NHWC data layout.
Change-Id: Ib7fb438cac23e78ff0104c895c3b7596bf7c3aa7
|
|
Change-Id: Ibaa0b909680a6f2a155e4ca6fa24e1144d6e6e73
|
|
Change-Id: Ib0568024b4f43ae17e6c75ac35d31cf103ce3482
|
|
I had forgotten that as a workaround for the testing using a different
number of test cases for different hardware backends, we repeat
the expected test results an arbitrary number of times
Change-Id: I4d1ea8daa1cbfd63ebcd7ac93ac30acb5806fafe
|
|
* Fixes bug when calcuating indexes for NHWC in
Pooling2D CpuRef implementation, it now uses
TensorBufferArrayView.
* Adds 2-Channel unit tests for Pooling2d on CpuRef,
Cl and Neon. The single channel tests were not
properly exercising Pooling2d using NHWC data layout.
* Refactors Pooling2D NHWC tests so that the input and
output data are permuted to NHWC when necessary,
instead of hard coding the data in NHWC format.
Change-Id: I5b9d41ed425ff283ea8c8ef6b1266ae0bc80f43b
|
|
Change-Id: If2befd3bd75da70c57f924cbe18fa0a47bf7a27f
|
|
Change-Id: Iaf03755b0d30598f5d09ca36af8eeeb369191950
|
|
Change-Id: Ib7ea0a0b809bdfe1d705b31a08f42e1b7815b0d2
|
|
Remove incorrect early out from image resize
Change-Id: I58014680f5cbe43dfc6b4af66638db67c070a146
|
|
When looking for the top probability, use the 'first' result not the 'second'.
This avoids an issue where for batched tests the classification index was reported wrongly.
Still doesn't correctly handle multiple results with the exact probabibility, or batched testing,
but it's slightly more correct than before.
Change-Id: I57d33552754667613e222d9d2037e12c87a96854
|
|
Change-Id: I1bc8d2f270d4abfaa62e4843b8d96bcf462c6b09
|
|
* In ImagePreprocessor, typecasted the number of image elements
(unsigned int) to a const_iterator::difference_type (int)
Change-Id: Idb8611185f6471751678a0a77d8225bceb204a34
|
|
Change-Id: I99498a633d75a2b4bdcd059627b07beaa9984890
|
|
shape in ValidateTensorShapesFromInputs
* Convolution2dLayer
* DepthwiseConvolution2dLayer
* Pooling2dLayer
* ResizeBilinearLayer
* Unittests for ValidateTensorShapesFromInputs
Change-Id: I057caf8a90d822175a7dd7271f960b65c6154bb4
|
|
Change-Id: Iae9a76b10d84d7ba80136b54355f8d37d2df475d
|
|
* Calculate index for NHWC to compute Convolution
* add more unit test for NHWC
Change-Id: I800d649b9b42be2758c445e3b3e76142888c1377
|
|
Change-Id: Ib642b0c983fed21cf38be8233bd13d86f4bf413d
|
|
implement test for TfLite VGG16 quantized
Change-Id: I2061f62f62684b963fa0f090718f1dcffe5c93ce
|
|
Change-Id: Ice19d3f763298bc14585267df389e99df846320d
|