aboutsummaryrefslogtreecommitdiff
path: root/delegate/src
AgeCommit message (Collapse)Author
2023-03-28IVGCVSW-7555 Restructure DelegateTeresa Charlin
* New folders created: * common is for common code where TfLite API is not used * classic is for existing delegate implementations * opaque is for new opaque delegate implementation, * tests is for shared between existing Delegate and Opaque Delegate which have test utils to work which delegate to use. * Existing delegate is built to libarmnnDelegate.so and opaque delegate is built as libarmnnOpaqueDelegate.so * Opaque structure is introduced but no API is added yet. * CmakeList.txt and delegate/CMakeList.txt have been modified and 2 new CmakeList.txt added * Rename BUILD_ARMNN_TFLITE_DELEGATE as BUILD_CLASSIC_DELEGATE * Rename BUILD_ARMNN_TFLITE_OPAQUE_DELEGATE as BUILD_OPAQUE_DELEGATE Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: Ib682b9ad0ac8d8acdc4ec6d9099bb0008a9fe8ed
2023-03-22IVGCVSW-7526 Upgrade ArmNN to Tensorflow 2.12Ryan OShea
When creating a flatbuffers model, we need to provide an empty buffer 0 that is reserved by tensorflow. When creating empty buffers for inputs and outputs we can not pass in an empty vector, or tflite will assume that we know how many bytes to allocate in advance. Instead we need to only pass in the builder. * Update libraries in FindTfLite.cmake * Add nullptr to delegate struct for OpaqueDelegateBuilder * Fix issue in unit tests where Flatbuffers model was not being parsed by tflite * Tensorflow 2.12 now includes C++ 17 features. Update our cmake build to require a compiler to support these features. * Change minimum cmake in Arm NN to 3.7 as that's the minimum for the delegate build. Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Signed-off-by: Narumol Prangnawarat <narumol.prangnawarat@arm.com> Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: I7d15b196b8c59b1914f8fc1c4c2f8960630c069c
2023-03-14IVGCVSW-3808 Add ElementwiseBinaryLayerMike Kelly
!android-nn-driver:9329 * Added ElementwiseBinaryLayer that can represent all ElementwiseBinary operations including Add, Div, Sub, Maximum, Mul and Minimum. * Updated Delegate to use ElementwiseBinaryLayer instead of the Add, Div, Sub, Maximum, Mul and Minimum layers. * Updated Deserializer to use ElementwiseBinaryLayer instead of the Add, Div, Sub, Maximum, Mul and Minimum layers. * Updated OnnxParser to use ElementwiseBinaryLayer instead of the Add layer. * Updated TfLiteParser to use ElementwiseBinaryLayer instead of the Add, Div, Sub, Maximum, Mul and Minimum layers. * Updated CL and Neon tests to use ElementwiseBinaryLayer. * Updated CL and Neon Backend Specific Optimizations to accept ElementBinaryLayers as well as Add, Div, Mul, Sub, Maximum and Minimum layers. Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Signed-off-by: Mike Kelly <mike.kelly@arm.com> Change-Id: I7cbb96b60eb01f0e2b57b0541016d48a08b86c75
2023-02-28Add constant version of IConnectableLayer::GetConstantTensorsByRefMatthew Bentham
This makes it easier to use, particularly in backends where it is common to pass around const pointers to IConnectableLayer. The non-constant version is rewritten to use the constant version. Signed-off-by: Matthew Bentham <matthew.bentham@arm.com> Change-Id: Id3a8384447e93c213299a85ade9a667df5960534
2023-02-21MLCE-753 Expand Tensorshape for relevent layers before verifying supportRyan OShea
Previously we were adding a reshape layer to "broadcast" tensors for elementwise operations. This broadcast was happening too late and was really just an expand dims. This was breaking the constant attributes of tensors and layer support of certain backends. * Remove addition of reshape layer when expanding dimensions * Replace broadcast function with expand dims to equal rank function * Fix some error status checks in various layers * Add new TensorUtil function that expands dims to a defined rank * Add unit tests to new TensorUtil function Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: I31aca47c98075fef4f86864a15470f5faa55ab8d
2023-02-06IVGCVSW-7476 Remove ProfilingOptions from DelegateOptions.Colm Donelan
ProfilingOptions is not used in DelegateOptions. Instead the parameters are passed in through the RuntimeOptions. This is done in ExecuteNetwork and TfliteExecutor. Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: Iaab3d4ef277c47e1ff82a51ba2648f5f51ec3e2c
2023-02-03IVGCVSW-7501 Allow constant tensors as inputs for input data in the delegateRyan OShea
In the TLCT tests we were failing many tests because they used constant tensors as data input for the layers. We had the functionality in place but we didnt have it spread across the visit functions. * Check if inputs are constant tensors and attempt to assign them to input slot of layers. * Add missing checks to some functions that return a kTfLiteStatus so we can see if they fail * Clean up CreateConstTensor function Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: I8610b770aea56932a98f91c961d59b3de47c2ab5
2023-01-30IVGCVSW-7475 Fix uncaught warnings treated as errors in delegate release buildRyan OShea
* Add initialization of activation type in delegate layers with fused activations * Remove unused variable activationOutputInfo on layers with fused activations Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: Ieb2ba00516a159639871c391069faf162cbb666d
2023-01-27IVGCVSW-7450 Fix delegate fallback when fused activation is unsupportedRyan OShea
In layers that support fused activations, we check for activation layer support after we already create the base layer. This breaks the fallback as we already added the base layer to the graph. * Creates ValidateFusedActivation shared function * Moves Activation validation higher in the VisitFunction Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: I239af360923f695fc374ddeaeefa24c062eaf9e8
2023-01-27IVGCVSW-7441 Checking for constant input tensors before populating.Colm Donelan
* When the tfLiteExecutor attempts to populate the input tensors it did not check whether the tensor was constant. This was causing segmentation faults. Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: I80a4cc788de4ffe08afb2df9185d04fcb8b27c3a
2023-01-24IVGCVSW-7277 Fixed issues with FullyConnected on certain TFLite modelsMike Kelly
* TFLite Parser: * Fixed issue in ParseReshape where the targetShape wasn't always calculated correctly * Fixed issue in ParseFullyConnected where the wrong name was used for the ReshapeLayer * Added an ExpandDims to the FullyConnected to ensure that we reshape the output correctly * TFLite Delegate: * Added an ExpandDims to the FullyConnected to ensure that we reshape the output correctly Signed-off-by: Mike Kelly <mike.kelly@arm.com> Change-Id: I129dfcb8543f8a3a297c0589c841be20ef3b6407
2023-01-24IVGCVSW-7455 Workaround to allow CLBatchMatMul to parse some 4D modelsMike Kelly
* Added ability to reduce dimension sizes when calling BuildArmComputeTensorInfo or BuildArmComputeTensorShapes, this will attempt to remove leading 1s in order to squeeze the number of dimensions but retain the size. * Changed ClBatchMatMulWorkload to attempt to squeeze the number of dimensions to 3 as the CL Gemm Kernel can only support up to 3 dimensions. Signed-off-by: Mike Kelly <mike.kelly@arm.com> Change-Id: I6b3d0886c5b97fdb686838fc3dc292833ddc4643
2023-01-24IVGCVSW-7297 When creating multiple Executors only the lastMike Kelly
one works fine * All ArmNNExecutors now share a single IRuntime. * All armnn_delegates now share a single IRuntime. * Increased delegate major version. Signed-off-by: Mike Kelly <mike.kelly@arm.com> Change-Id: I95cbdc32655ec0beb476dbb2d60f1a0209df8f04
2023-01-13Update license from commit:Ryan OShea
Fix delegate fallback during VisitNode 05b6a3e5946a3f58b6f5b0caface9153a1c2b364 Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: Iba2c70fd49ce8d7c49fb6deb7183b42db2cc68c7
2023-01-12Fix delegate fallback during VisitNodeRyan OShea
During VisitNode throwing an ArmNN exception incorrectly terminates the process instead of handing over to tflite * Catches ArmNN exceptions during VisitNode Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: I6c71be11e9b73694747b27fe9febab8d9669b4d4
2023-01-09IVGCVSW-6493 Add GpuAcc Batch MatMul workload Fp32Teresa Charlin
* GpuAcc only supports up to 3D, so no 4D test have been added Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: Ie926cd45c350be624cbdc6cb27c89d2d3f60884b
2023-01-05Add backend name to delegate test names that did not have it.Teresa Charlin
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: If9531f74d0b503c36f1b72ef51eee7e9f65559f2
2022-12-20IVGCVSW-7409 GPU backend options not being passed from the delegate.Colm Donelan
Two problems here: * First the Delegate was using the parameter options after the execution of std::move on it. * In ExecuteNetworkParams 3 GPU backend options were instead being set as optimizer options. Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: I61c7fad8a5819a0a4aec0243899019a342c5cc5f
2022-12-12Optimize the calling of IsLayerSupported().Cathal Corbett
* Done as part of 22.11/23.02 innovation days. * IsLayerSupported() is called in model prepare (delegate, android-nn-driver and shim/support_library) and again in ArmNN once model otimization is performed. * From calling IsLayerSupported() the first time, we should know that the layers are supported and what backend they are supported on. * Solution is to set the BackendId of the IConnectableLayer when IsLayerSupported() is called the first time, * In the Optimize() function we then check if the backend is set. If so, we do not call IsLayerSupported() again. * In the case a layer that is supported gets optimized, then the BackendId of that layer get set to "Unknown" for the new optimized layer and IsLayerSupported() will get called on the newly optimized layer. * Includes bug fix IVGCVSW-7213 for Android Mean FP16 CpuAcc tests. Also related to bug IVGCVSW-7211. Signed-off-by: Cathal Corbett <cathal.corbett@arm.com> Change-Id: I7a7820d0cdb079ffb5a3a2e0c44e252f652df53b
2022-11-22IVGCVSW-6980 Delegate support for slice operatorCathal Corbett
Signed-off-by: Cathal Corbett <cathal.corbett@arm.com> Change-Id: I90d800160b070e25d999b5102a7ce6d3e0ed6a81
2022-11-16IVGCVSW-7214 Disable BF16-Turbo-Mode and remove conversion layersRyan OShea
- Remove Bf16ToFp32 Conversion Layer - Remove Fp32ToBf16 Conversion Layer - Remove B16 Conversion tests * Throw exception if m_ReduceFp32ToBf16 optimzer option is set to true * Provide comments to enable fast math in order to use bf16 * Update docs to inform users to enable fast math for bf16 Execute Network Changes * Require bf16_turbo_mode to also have fast_math_enabled set to true - Remove setting m_ReduceFp32ToBf16 optimizer option Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: Ibaa6da9d29c96a1ce32ff5196b0847fde9f04a1c
2022-11-15Minor error formatting fixes.Colm Donelan
Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: I17823fb8b6bbabc4da327187167ce9582ee29b32
2022-10-28IVGCVSW-7296 REDUCE_PROD tests fail when using Tf 2.10Teresa Charlin
* In TF what ArmNN calls quantized data types can be non-quantized as well. * This patch creates 2 models: * ArmNN: model where int8 and uint8 will always be quantized, but scale can be 1 and offset 0 * TFLite: model where int8 and uint8 can be quantized and non-quantized Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: Id960f2f30988f2bbec88cb4e0c52c189ac957bae
2022-10-28IVGCVSW-6494 Add CpuAcc Batch MatMul Workload Fp32Teresa Charlin
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: I2def6995f81d33e68f1ea45d8d19a1e6294049b1
2022-10-23IVGCVSW-7126 Modify odrer of Tf 2.10 librariesColm Donelan
* Modify the order of Tf absl libraries in delegate build. * Disable failing delegate unit tests. * Add -Wno-comment to delegate ExecuteNetwork build. Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: I9e4696e032a5e26df8ba578b980456cbac4d0ab0
2022-10-21IVGCVSW-7126 Update Arm NN to Tensorflow 2.10Colm Donelan
* Modify get_tensorflow.sh to point to the 2.10 Tensorflow tag. * Modify FindTfLite.cmake to handle the break up of libruy.a * Modify armnn_delegate.cpp to add registration_external field. * Prevent the serializer/deserializer from using a system installed version of flatc. Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: I688f33f387924ba29bf400f60e56f73b2329fdc1
2022-09-22IVGCVSW-6498 Add Support for Batch MatMul to TfLite DelegateRyan OShea
* Creates delegate/src/BatchMatMul.hpp * Add VisitBatchMatMul function * Add BatchMatMul to switch in armnn_delegate * Creates delegate/src/test/BatchMatMulTest.cpp * Creates delegate/src/test/BatchMatMulTestHelper.hpp * Add Int8 and Fp32 unit tests on ref backend * Add BatchMatMul to delegate supported ops Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: I50e61314cf063f986c8a0f7d508847a96953735e
2022-08-29IVGCVSW-6603 'Add a no fallback mode to the TfLite Delegate'Sadik Armagan
* Added disable-tflite-runtime-fallback option to armnn_delegate * Updated armnn_delegate version Signed-off-by: Sadik Armagan <sadik.armagan@arm.com> Change-Id: I449b16404d3ffe98e6dac52a43e7c25225addd73
2022-08-29IVGCVSW-7154 'Constant Tensors As Inputs'Sadik Armagan
* Fixed the issues about converting Conv2D and DepthwiseConv2d input issues * Read 1D input tensors that do not have shape specified Signed-off-by: Sadik Armagan <sadik.armagan@arm.com> Change-Id: I12f3d1c57a2afedac42c6e7b31e4b1fc689abeca
2022-07-27IVGCVSW-7094 Add LOG and SIN support to tflite delegateTeresa Charlin
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: I355298f365b82cad1e3f46cfebf1c1375716cf92
2022-06-27IVGCVSW-6981 Remove deprecated code 22.05 [Post Release]Nikhil Raj
Signed-off-by: Nikhil Raj <nikhil.raj@arm.com> Change-Id: I9ccaefbe28ea572e9e2b4a2168574804667f7460
2022-06-22Revert "Revert "IVGCVSW-6873 Import inputs but don't export outputs fails.""Francis Murtagh
This reverts commit a0f8b15d4ddb5075f380003ff31b271d389d3b66. Reason for revert: <Test ClDmaBufInternalTests review > Change-Id: Ibc4a77fa008643849da7330391942e4c87b941e2
2022-06-21Revert "IVGCVSW-6873 Import inputs but don't export outputs fails."James Conroy
This reverts commit 03bf98a8bc51ad20eef4b9ca5fbf6ce15e063721. Reason for revert: Caused failures in tests located in internal repo. Change-Id: If35cb0ede349b270e4e7827324382e09455d8cfa
2022-06-20IVGCVSW-6873 Import inputs but don't export outputs fails.Colm Donelan
Only one bool is used to indicate whether inputs should be imported. However, its possible for the user to want to import inputs but not export outputs. In addition it's possible for a user to enabled import during optimize but then pass a memory source that does not require import. * Add m_ExportEnabled to INetwork.hpp. * Modify Network::dNetwork to consider both m_ImportEnabled and m_ExportEnabled. * Add ValidateSourcesMatchOptimizedNetwork to LoadedNetwork to validate import options between optimize and network load. * Update the TfLite delegate consider exportEnabled flag in the optimizer. !armnn-internal-tests:425350 Signed-off-by: Colm Donelan <colm.donelan@arm.com> Change-Id: I776eab81595898e43f91ab40306962eae61329f4
2022-06-15IVGCVSW-6946 Add Pool3D to tflite delegateRyan OShea
* Add new test and test helper for Pool3d * Add new custom operator to switch in armnn_delegate.cpp * Add new pool3d function to pooling.hpp * Update doxygen Signed-off-by: Ryan OShea <ryan.oshea3@arm.com> Change-Id: I77a541bf423b337c749e70c564cdd727efe2fd05
2022-05-18IVGCVSW-6929 Support for models with implicit expandedMike Kelly
dimensions * Added allow-expanded-dims to TFLite parser and ArmNN delegate * If true ArmNN will disregard dimensions with a size of 1 when validating tensor shapes. Tensor sizes must still match. * This allows us to support models where tensors have expanded dimensions (i.e. extra dimensions with a size of 1). * Fixed bug in Network where it assumed that only the first option could be ShapeInferenceMethod. * Fixed bug where m_ShapeInferenceMethod was lost when copying or moving Graphs. * Changed Delegate to pass "infer-output-shape", "allow-expanded-dims" and other BackendOptions through to the Network during construction. Signed-off-by: Mike Kelly <mike.kelly@arm.com> Change-Id: Ibe7c5ae6597796fc9164cb07bd372bd7f8f8cacf
2022-05-16IVGCVSW-6124 ConstTensorsAsInput: Conv2d - FrontEndKeith Davis
* Update Front-end and Tools. * Updated Serializer, Deserializer and unit tests to reflect this. * Updated TfLiteDelegate, TfLiteParser and OnnxParser. * Updated Ref. * Fixed resulting Neon / CL tests * Unified optimizers for conv2d ops * Optimizer Fix - Fp32ToBf16 * Partial implementation for ACL backends to fix VTS failures !android-nn-driver:7477 Signed-off-by: Keith Davis <keith.davis@arm.com> Change-Id: I5fb18877f7ee32643e15a9818945356274bb401b
2022-05-06IVGCVSW-6929 Fix for segfault in tflite delegateMike Kelly
* It's possible that a model may have an input entry for bias tensors but that the index for those is -1. If it's -1 then it's not present. * Fixed logic error in IsOptionalOperandPresent: it returned false if it was present and true if it was missing. Signed-off-by: Mike Kelly <mike.kelly@arm.com> Change-Id: I45ad8d8552122493c529b1a35a5689416ccfbb71
2022-05-05IVGCVSW-6127 ConstTensorsAsInput: DepthwiseConvolution2dCathal Corbett
!android-nn-driver:7418 * Update Front-end and Tools. * Updated Serializer, Deserializer and unit tests to reflect this. * Updated TfLiteDelegate, TfLiteParser and OnnxParser. * Change NNDriver to new API. * Updated Ref. * Neon and Cl backend partially completed (Backend.cpp files). * Added dynamic or constant input EndToEnd tests. * Added ConstantTensorAsInputMemeberVariableRedirect Optimization. Signed-off-by: Cathal Corbett <cathal.corbett@arm.com> Change-Id: Ib18b6c10a093042e165e25237dc04a4c67ba82da
2022-05-05IVGCVSW-6862 Add GATHERNd Neon workloadTeresa Charlin
* Changing the test in the delegate to match one of the unit tests Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: I553ca266116ba8ee173fc951ab1ffd2b6eed1428
2022-05-05IVGCVSW-6806 Add Unidirectional Sequence Lstm support to NeonMike Kelly
* Corrected TensorInfo order for IsUnidirectionalSequenceLstmSupported * outputStateOut TensorInfo is not optional. * cellStateOut TensorInfo is not optional. * TensorInfo Order matches other QLSTM/LSTM layers. * Added missing parameters to UnidirectionalSequenceLstmOperator for delegate. * Added quantized UnidirectionalSequenceLstm support to Neon !android-nn-driver:7457 Signed-off-by: Mike Kelly <mike.kelly@arm.com> Change-Id: I26dde1bb96793dd25eb9081ca5ae5f63752288c4
2022-05-04IVGCVSW-6858 Add GATHERNd Support to the TfLite DelegateTeresa Charlin
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com> Change-Id: I56418875b3bb2ae45b5c69bfeaafa1a6126b8085
2022-03-01IVGCVSW-6704 Change the namespace from armnn::profiling to arm::pipeCathal Corbett
* Updated ABI version to 29 due to being the first ABI break in 22.05 !android-nn-driver:7226 Signed-off-by: Cathal Corbett <cathal.corbett@arm.com> Change-Id: I9c50007dcd5b5e792757e7bd1213606df5ffec36
2022-02-14IVGCVSW-6708 Break Profiling Dependence on IRuntime ExternalProfilingOptionsJim Flynn
Change-Id: I30a46f3368bbbf33019eac4fa1245f6ff69deacd Signed-off-by: Jim Flynn <jim.flynn@arm.com>
2022-02-11MLCE-754 'Improve operator support error/warning from Arm NN Delegate'Sadik Armagan
* Improved error reporting on armnn_delegate Signed-off-by: Sadik Armagan <sadik.armagan@arm.com> Change-Id: I1bd131fb56d64b32b1fafad0465256178720226c
2022-02-03IVGCVSW-6724 Accessing ConstTensors from IConnectableLayerNikhil Raj
Signed-off-by: Nikhil Raj <nikhil.raj@arm.com> Change-Id: I01f42a520d15c6dabd2f77c7715c91b8f7026476
2022-02-02IVGCVSW-6639 Add GetParameters to IConnectableLayerJim Flynn
Change-Id: Id55a460ecb510f5b30235b03f54390f2c8188fc2 Signed-off-by: Jim Flynn <jim.flynn@arm.com>
2022-01-18IVGCVSW-6717 'Inference Passes on Tflite-Parser but fails when using Arm NN ↵Sadik Armagan
Delegate' Signed-off-by: Sadik Armagan <sadik.armagan@arm.com> Change-Id: I40ab6aff4bc076eafcef08c1c20aacdbd52da853
2022-01-18MLCE-736 'Adding support for TfLite Models'Sadik Armagan
* Added constant input supports for Pack/Stack, Concatenation operators * Added Int32 support to Pack/Stack operator on CpuRef * Removed unsupported operator from TfLite Delegate Signed-off-by: Sadik Armagan <sadik.armagan@arm.com> Change-Id: I64203e174300d23eedeb22bddefe07e931c4eff3
2021-12-08Add addition timing logging outputJan Eilers
* Adds ExecuteNetwork when building the delegate only * Adds timings to delegate subgraph creation * Adds executions times Signed-off-by: Jan Eilers <jan.eilers@arm.com> Change-Id: Ieff2f67ea8dbb6c2a708f8810e84a20485b7a631
generated by cgit v1.2.1 (git 2.23.0) at 2024-07-10 01:26:53 +0000