Age | Commit message (Collapse) | Author |
|
Signed-off-by: John Mcloughlin <john.mcloughlin@arm.com>
Change-Id: Id4bdc31e3e6f18ccaef232c29a2d2825c915b21c
|
|
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
|
|
- 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
|
|
* Add functionality to print output tensors to file in tempdir
* UnitTests
Signed-off-by: Keith Davis <keith.davis@arm.com>
Change-Id: Idfb4c186544187db1fecdfca11c662540f645439
|
|
* dot file to be generated when -v is given. It was only being generated when using the delegate as executor
* output name read from m_Params.m_OutputNames instead of m_TfLiteInterpreter
* typo: "delage" instead of "delegate"
* QAsymmS8 templated as int8, instead of uint8
Change-Id: Ie13ae0f7e6395c0ebcb5ecda32e72082dee8aa6c
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>
Change-Id: Iac97a23927ba42290ebeb3446bbd36da15045e07
|
|
This reverts commit 1a7f033768acb27da11503bd29abb468d2e77f9e.
List of fixes to be able to add this code again:
* "emplacing_back" the vector inputTensors into the vector m_InputTensorsVec outside the for loop
* GetIOInfo() uses IOptimizedNetwork instead of INetwork, where the infered shapes are not saved
* Add missing data type Signed32 to SetupInputsAndOutputs()
* PrintOutputTensors() prints the actual output without dequantizing
* Add profilingDetailsMethod as input in networkProperties in ArmNNExecutor constructor
* Fix typos
Change-Id: I91de166f87228282db3efa27431fe91458834442
Signed-off-by: Teresa Charlin <teresa.charlinreyes@arm.com>
Change-Id: Ic6634d48892d11e5f146cdf285e1e333e93e9937
Signed-off-by: Francis Murtagh <francis.murtagh@arm.com>
|
|
This reverts commit 615e06f54a4c4139e81e289991ba4084aa2f69d3.
Reason for revert: <Breaking nightlies and tests>
Change-Id: I06a4a0119463188a653bb749033f78514645bd0c
|
|
* Remove InferenceModel
* Add automatic IO type, shape and name configuration
* Depreciate various redundant options
* Add internal output comparison
Signed-off-by: Finn Williams <finn.williams@arm.com>
Change-Id: I2eca248bc91e1655a99ed94990efb8059f541fa9
|
|
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
|
|
* Pass through the value of m_EnableProfiling from Executenetwork to
DelegateOptions.
* If internal profiling is enabled print it out from inside the delegate.
* Remove an unnecessary ProfilerImpl instance from WorkingMemhandle.hpp
* Remove an unnecessary parameter from TfLiteDelegateMainImpl in
ExecuteNetwork.
Signed-off-by: Colm Donelan <colm.donelan@arm.com>
Change-Id: Ia1d4b1eb3a05ca5b4d80cc39e138c7fac182d948
|
|
* Introduce a mechanism to construct a DelegateOptions from an
ExecuteNetworkParams.
* Modify ExecuteNetwork to use this constructed DelegateOptions.
Signed-off-by: Colm Donelan <Colm.Donelan@arm.com>
Change-Id: Ied663a1e00ac3eece42244ed313ddafd6d2ce078
|
|
* Allows to supply different input data for every execution of a
model when using the 'iterations' option in ExecuteNetwork
* Removes the option 'simultaneous-iterations' because it's functionallity
is now covered by 'iterations'
* Adds a deprecation warning message to notify users
* Little refactor of warning messages
Signed-off-by: Jan Eilers <jan.eilers@arm.com>
Change-Id: Ib3ab0d6533f6952bfee20d098a890b653c34cc12
|
|
* Allow the user to use create a tread pool with a single thread
* This is in keeping with how the android-nn-driver was implemented
* Add it to ExecuteNetwork thread pool creation
Signed-off-by: Kevin May <kevin.may@arm.com>
Change-Id: I05b8048a9e0e45ae11d2b585080af28d9d008d81
|
|
* Remove concurrent flag from ExecuteNetwork as it is possible
to deduce if SimultaneousIterations > 1
* Add void RunAsync()
* Refactor some unit tests
Change-Id: I7021d4821b0e460470908294cbd9462850e8b361
Signed-off-by: Keith Davis <keith.davis@arm.com>
Signed-off-by: Kevin May <kevin.may@arm.com>
|
|
* Remove from .md files and Doxygen
* Remove from armnn/docker build
* Remove Tensorflow model format from ExecuteNetworkParams
* Remove Tensorflow model format from ImageTensorGenerator
Signed-off-by: Kevin May <kevin.may@arm.com>
Change-Id: Id6ed4a7d90366c396e8e0395d0ce43a3bcddcee6
|
|
* Enabled async mode with '-n, concurrent' and 'simultaneous-iterations'
in ExecuteNetwork
* Number of input files provided should be equal to number of input files
provided multiply by number of simultaneous iterations divided by comma
!armnn:5443
Signed-off-by: Sadik Armagan <sadik.armagan@arm.com>
Change-Id: Ibeb318010430bf4ae61a02b18b1bf88f3657774c
|
|
Signed-off-by: Nikhil Raj <nikhil.raj@arm.com>
Change-Id: Ib00be204f549efa9aa5971ecf65c2dec4a10b10f
|
|
* Added package manger to turn internal calls to find_package into a no-op
* Changed delegate cmake so it can now be built within armnn
Change-Id: I2a7ecb9a3c1ca05474cd1dccd91498f6f6c0b32e
Signed-off-by: Finn Williams <Finn.Williams@arm.com>
Signed-off-by: Sadik Armagan <sadik.armagan@arm.com>
|
|
* When invalid backend specified an ARMNNLOG should be invoked
to fail more gracefully
Signed-off-by: Keith Davis <keith.davis@arm.com>
Change-Id: Iec34fbc03dbeeb66836b0d5e1dd381d021a379b1
|
|
* Stops execute network rejecting backendId based on BackendRegistry
* Dynamically loaded backends arent visible yet as runtime isn't initialized
Change-Id: I87adfd137b2225ab07f8c3e996db9565caf276eb
Signed-off-by: Francis Murtagh <francis.murtagh@arm.com>
|
|
* Removed boost program options and replaced it with cxxopts
* Unified adding, parsing and validation of program options
into the struct ProgramOptions
* Program options are now parsed directly into ExecuteNetworkParams
which can be passed directly to MainImpl
* Split NetworkExecutionUtils into header and source
* Removed RunTest
* Removed RunCsvTest
* Removed RunClTuning
* Moved MainImpl back to ExecuteNetwork.cpp
* Added additional util functions
The functionality of ExecuteNetwork remains the same. Only
cl tuning runs need to be started separately
and there is no short option for fp16-turbo-mode because -h is
reserved in cxxopts to print help messages
Signed-off-by: Jan Eilers <jan.eilers@arm.com>
Change-Id: Ib9689375c81e1a184c17bb3ea66c3550430bbe09
|