| aec086a4a628f1f186323f7db8f4574515d74f24 |
|
03-May-2018 |
Michael Butler <butlermichael@google.com> |
Free wrapper resource before assignment
Any resources owned by "this" (lefthand side of assignment) are never
released (as the destructor would do), but are overwritten by the
move-assignment process and are therefore leaked.
This CL frees the resource before assignment. Additionally, it adds
missing ownership documentation for Compilation.
Bug: 79205067
Test: mma
Test: NeuralNetworksTest_static
Change-Id: Iacd561adee5205cb039fc7dea4fd1a816ca8d641
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 247996038e3595d638dad6a54aeae69a634d8727 |
|
22-Mar-2018 |
Michael Butler <butlermichael@google.com> |
Explicitly set dimension to be nullptr when empty vector
An empty std::vector is allowed to have any value for the
std::vector::data (pointer) field. When a std::vector is
transferred into ANeuralNetworksOperandType::dimensionCount
and ANeuralNetworksOperandType::dimensions,
ANeuralNetworksOperandType::dimensions must be explicit set
to nullptr if the std::vector is empty.
This CL additionally has come code cleanup, such as removing the
now-unused setFromIntList utility function.
Bug: 76122637
Test: mma
Test: NeuralNetworksTest
Change-Id: I097464df78ee037fd87bc8a970f696d5d3bb6e3e
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| b74d2837ab1687c1a4f913aa5f90a9838efe0add |
|
21-Feb-2018 |
Miao Wang <miaowang@google.com> |
Add relaxed execution tests
- Extend the test generator and test wrapper to allow relaxed
execution.
- Make the accpeted error range to be 5 ULP of FP16.
- Copy and modify existing V1_0 tests to be V1_1 relaxed tests.
Bug: 63911257
Test: mm
Test: Relaxed tests pass on devices with FP16 support
Change-Id: I8954aeeebcf7312d7d94580992708f1b42e2b32e
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| a7e5f8d601737eebef06b7b4e3866975cc4645df |
|
22-Nov-2017 |
David Gross <dgross@google.com> |
Improve error handling for Model::finish().
- finish() fails if Model is in invalid state.
- If finish() fails put Model in invalid state.
Test: nn/runtime/test
Change-Id: I9bf16a332505136d67d371ba5d55f56b7c3cce53
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 46ada93580e5036924e969b56799e2c62bdb8699 |
|
01-Nov-2017 |
David Gross <dgross@google.com> |
Add missing Result values to NeuralNetworksWrapper.h.
Bug: 68728827
Test: nn/runtime/test
Change-Id: I00b4a09b5d0dcf3d1d6776fb4b57c9003820f16f
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 66d5cb6e3a90aefc8d545f6369080ab88de9d667 |
|
05-Oct-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Rename setInputsAndOutputs
To avoid confusion with other set* functions, rename setInputsAndOutputs
to identifyInputsAndOutputs.
Also added a few new generated tests when I reran generate_test.sh.
Bug: 63905942
Test: local & VTS tests
Change-Id: Ie947044bee1a97426c847b9d66e60f0c25395bbc
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| e3178825b8686f3300a895572691a2e8c1f0676b |
|
05-Oct-2017 |
David Gross <dgross@google.com> |
Asynchronus partitioned execution, plus minor code cleanup.
Partioned execution is now asynchronous end-to-end, although
the individual execution steps are serialized with respect to
one another.
Other changes:
- Minor code cleanup per comments from previous CL
https://googleplex-android-review.git.corp.google.com/#/c/platform/frameworks/ml/+/3000624/
including comments on memory management for
inter-partition temporaries.
- Remove "simulation" form of partitioning, and renumber
values of debug.nn.partition property accordingly.
- Fix android:nn:wrapper::Event -- add missing contructor
and fix comment.
Bug: 63905942
Test: mma (user)
mma (userdebug)
ml/nn/runtime/tests (userdebug)
(and spot check that debug.nn.partition unspecified,
0, and 1 exhibits the correct control over
partitioning behavior)
Change-Id: I5767d2144c4a42258444a6445866529919d4fe8d
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 45bf79e5b9fee354fde7c1f64417d9ca4a1da7da |
|
25-Sep-2017 |
Miao Wang <miaowang@google.com> |
Clarify the expectation of scale and zeroPoint for affected ops.
- Remove the OperandType constructor taking f_min and f_max as
quantization parameters as it causes confusions for ops like
LOGISTIC and SOFTMAX.
- Update the documenation to clearly state the expected scale and
zeroPoint for LOGISTIC, SOFTMAX and CONCATENATION.
- Update the tests to directly input scale and zeroPoint.
Bug: 63905942
Test: mm
Test: NeuralNetworksTest pass
Change-Id: Ia450d6ce9509205d22e6383bd7e454afa0568cbb
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 39dcb0f2cc87b0eba06b314a91d55e50efea5abe |
|
14-Sep-2017 |
David Gross <dgross@google.com> |
Merge changes I5922f13a,I7194e277 into oc-mr1-dev
* changes:
Bring back Event, and wait on Event rather than on Execution.
Make Compilation API synchronous.
|
| 425b2594c76e934dfdbc93209253e3c189571149 |
|
14-Sep-2017 |
David Gross <dgross@google.com> |
Bring back Event, and wait on Event rather than on Execution.
This is at the recommendation of the API council.
This undoes change I50b8e155d9589210e14920237878a4f295e7755f
Bug: 63905942
Test: nn/runtime/test
Change-Id: I5922f13ae7070f7ec0af035557a9921c88ca9221
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 65aa556323f4a054f80a75b6c4c721b2a7ed3298 |
|
14-Sep-2017 |
David Gross <dgross@google.com> |
Make Compilation API synchronous.
Per advice from the API council, the Compilation API no longer exposes
asynchronous behavior to the user -- rather than start() and wait()
APIs, we now have a finish() API.
Also:
- Track whether or not Compilation::finish() has been invoked, for
the sake of API usage checking.
- Changed Model::finish() to return Result not int, for consistency.
- Made Model and Compilation documentation more similar to one another.
- Fixed typos in documentation.
NOTE: It's possible to implement compilation asynchronously behind the
synchronous API -- e.g., Compilation_finish() starts compilation
asynchronously, and we implicitly wait for it to complete at the point of
Request_startCompute(). However, this would defeat one of the purposes of
an explicit Compilation API -- giving the application control over when
to spend the time compiling, to keep it off a critical timing path, such
as repeatedly applying a model to a sequence of frames in real time.
Bug: 63905942
Test: nn/runtime/tests
Change-Id: I7194e277aa6211af66794d917dcc5254c7d81af4
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| a5c16e4e73a34e6776a8f071ab884444c28b6893 |
|
14-Sep-2017 |
Miao Wang <miaowang@google.com> |
Rename ANeuralNetworksOperandType.offset to zeroPoint
Bug: 63905942
Test: NeuralNetworksTest
Change-Id: I15464657dc4eed9adbfc444df306cd0062993be3
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 7612f29b31f97f3b15769264131566b36dea9a25 |
|
12-Sep-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Remove the initialization and shutdown APIs.
Initialize the list of drivers as needed rather than
requiring an API call. This initialization will be
triggered when compiling the first model.
Bug: 63905942
Test: Compiled and ran the unit tests.
Change-Id: I1893a9f9d96d9d241916de3a7dd3cb7a54e33b06
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| d2d0c031c43e8e5aafc75e8a652d79bcc2aaca99 |
|
12-Sep-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Update documentation. Remove IntList.
As NDK council request, updated some documentation and removed
IntList. Other issues will be addressed in other CLs.
Bug: 63905942
Test: Compiled and ran the unit tests.
Change-Id: Iff6898737696daa6bbf41643b07d0a4508af4c0d
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 66d56404cdfab9ab8aa79d4bda83be3832a3aff9 |
|
14-Sep-2017 |
Miao Wang <miaowang@google.com> |
Make the enums pass as int32_t, and make offset and length size_t.
- Also removes fp16 perf and bootup time info.
Bug: 63905942
Test: NeuralNetworksTest
Change-Id: I1f722fed06a4227eb5f204eb51f7728fe749f5ec
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 01df5047f2661fd8428d5fc966bc3b7984ec1857 |
|
14-Sep-2017 |
David Gross <dgross@google.com> |
Merge "Rename Request to Execution at the API level." into oc-mr1-dev
|
| 5e7827e7bdc7ee6b6fc9b1989070c98fa10e9797 |
|
13-Sep-2017 |
David Gross <dgross@google.com> |
Move OEM codes out of NeuralNetworks.h to new file NeuralNetworksOEM.h.
Also remove FAKE_QUANT operation.
Bug: 63905942
Test: nn/runtime/tests, vts
Change-Id: Ia9262ec727f93dfe1b0ccade4ed76347c55ba163
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 3ced3cfd5b8f22b632c35f24e585c4847383b195 |
|
13-Sep-2017 |
David Gross <dgross@google.com> |
Rename Request to Execution at the API level.
There's a separate notion of Request at the HIDL level;
this is not renamed.
Bug: 63905942
Test: nn/runtime/tests
Change-Id: I671e3b387d735a9091dc5a2d14afbea6b45f444c
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 05292dc440641a38fab6a07c1372ce6780816c1c |
|
13-Sep-2017 |
TreeHugger Robot <treehugger-gerrit@google.com> |
Merge "Update ANeuralNetworksMemory related API" into oc-mr1-dev
|
| 5f916fc9a7ae95f172492bdfe5344c37beff3a6f |
|
12-Sep-2017 |
Michael Butler <butlermichael@google.com> |
Neuralnetworks HAL cleanup -- frameworks/ml/nn
Does three primary cleanups:
1) Removes unused OperandTypes, creates OEM types
2) Creates explicit ErrorStatus; functions return status
3) IDevice::initialize renamed to getCapabilities,
IDevice::getSupportedSubgraph renamed to getSupportedOperations
It makes the corresponding changes to the runtime.
Bug: 63905942
Test: mm, frameworks/ml/nn/runtime/test
Change-Id: I1545373dfa30780aaf9b61b3938be536cf122e74
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 42dc6a6cd68877cd85e3bc475b41bda0fd946c41 |
|
13-Sep-2017 |
Miao Wang <miaowang@google.com> |
Update ANeuralNetworksMemory related API
- Remove ANeuralNetworksMemory_createShared.
- Remove ANeuralNetworksMemory_getPointer.
- Allow a offset specified for
ANeuralNetworksMemory_createFromFd.
- Update the tests.
Bug: 63905942
Test: mm
Test: updated NeuralNetworksTests pass.
Change-Id: I1fd7362ec379f5f855c71c6758e0b647be9aaa76
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 4ffc65b9cb9c9d04c694f0948117554526a46b17 |
|
12-Sep-2017 |
David Gross <dgross@google.com> |
Remove Event and wait on Request instead.
Bug: 63905942
Test: nn/runtime/tests
Change-Id: I50b8e155d9589210e14920237878a4f295e7755f
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 18c58d289c2346d750301392866229630960b392 |
|
12-Sep-2017 |
David Gross <dgross@google.com> |
Merge "Fix bug in the way an Event handles a bound thread." into oc-mr1-dev
|
| 402baa3fde7795bcea6e2ebeffab97c70b228dd4 |
|
12-Sep-2017 |
David Gross <dgross@google.com> |
Fix bug in the way an Event handles a bound thread.
Previously, we were calling std::thread::join() from the
Event destructor. However, if the Event's reference count
drops to zero due to action in the bound thread, then the
bound thread calls the Event's destructor, and tries to
join itself; this throws an exception, and produces a
message like
terminating with uncaught exception of type std::__1::system_error:
thread::join failed: Resource deadlock would occur
However, we do not seem to have had any test cases that actully
call ANeuralNetworksEvent_free(), so every Event would leak, and
we never hit this problem.
Now also change Request::compute() to stop leaking its event -- doing
this without changing bound thread handling would produce the exception
shown above.
Bug: 63905942
Test: nn/runtime/tests
Change-Id: I0d9370d187c10be00a4a69db8eb5f79854c9be12
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 544739620cd7f37d40524d2407c92042e485c73f |
|
11-Sep-2017 |
David Gross <dgross@google.com> |
More discipline for models and requests.
- Must explicitly call new finish() API on model before compiling
or freeing.
- Must not modify a request once start() has been called on it.
Bug: 63905942
Test: nn/runtime/tests, nn/common/operations tests
Change-Id: Ifc6e614bda647d729e8702023a02613e629ca6a0
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 83e24dc4706a5b7089881a55daf05b3924fab3b7 |
|
10-Sep-2017 |
David Gross <dgross@google.com> |
Add compilation APIs.
For now, compilation is implemented as a synchronous no-op.
Bug: 63905942
Test: nn/runtime/tests, nn/common/operations tests
Change-Id: I1cb96a7e800671d0c70c9cc1b7eab6f8244232cb
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 0afe5897f4034528b027294efbe45c836924643c |
|
09-Sep-2017 |
Michael Butler <butlermichael@google.com> |
Merge "Disallow copy semantics for NeuralNetworkWrapper objects." into oc-mr1-dev
|
| 105807d963d969197fe78185ed588bfad3dc0ea5 |
|
05-Sep-2017 |
Miao Wang <miaowang@google.com> |
Add support for mmaped file descriptor
- create new API to create ANeuralNetworks_Memory from a mmaped fd.
- add the corresponding Memory constructor to NeuralNetworksWrapper.h
- Add unit test TestFd for the new API.
Bug: 63905942
Test: mm
Test: TestFd pass with CPU path.
Test: TestFd pass on sample HIDL driver, with ag/2588539
Change-Id: I1fcb0f0bd01e58bd592e18eff11c27034ee4a4c3
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 7b87fec6bb919d16a8cc2820d470733a2776e8fa |
|
07-Sep-2017 |
Michael Butler <butlermichael@google.com> |
Disallow copy semantics for NeuralNetworkWrapper objects.
Before this CL, the NeuralNetworks runtime would double-free a
NeuralNetworks runtime object if its corresponding wrapper object was
copied as in the following scenario:
{
Model base_model = Model(...); // ANeuralNetworksModel_create x1
Model copied_model = base_model; // Copies runtime object's pointer
} // ANeuralNetworksModel_free x2, double-free
This CL removes copy semantics on Wrapper objects. Additionally, it
explicitly handles move semantics to ensure the runtime object is only
freed once.
Bug: 63905942
Test: mm -j40
Change-Id: Ibd0e8e67eb851a97c18482683c8a37edd665b162
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 2150f1d186b2854fb5aa609594be12a667f845f0 |
|
01-Sep-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Improvements to the Memory code.
Addresses the comments of a previous CL that arrived after
submission. Change_Id of that CL was I997a10f6102dcae7c6876c96a4f6f913eaf54c74.
- Improved documentation.
- getPointer is now getPointerAndSize. It returns a status code, as
not all Memory will be able to return a pointer.
Test: Compiled and ran the unit test.
Bug: 63905942
Change-Id: I8aade863212dd7e95856d7c3d583c1b5f00eda48
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 910c9f04913e3bee1a0b6406b6e146457d19c5e7 |
|
31-Aug-2017 |
Miao Wang <miaowang@google.com> |
Fix the OperandType constructor in NN wrapper.
- Use constructor delegation in OperandType constructors for the
quantized variants.
Bug: 63905942
Test: mm
Test: Created a quantized AVERAGE_POOL tests, and it pass
Change-Id: I1c0a4c26fc056873ff837ba2f60ec85923985fe2
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| b8810af67d4efbd966d6788c56cd01c5aa06a71e |
|
30-Aug-2017 |
TreeHugger Robot <treehugger-gerrit@google.com> |
Merge "Add full support for Memory pools." into oc-mr1-dev
|
| 8b99bb1d98a42b67ba1c00e12c7abb3708cf7c05 |
|
21-Aug-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Add full support for Memory pools.
Weights can now be passed via memory pools.
Test: Compiled and ran the unit test.
Bug: 63905942
Change-Id: I997a10f6102dcae7c6876c96a4f6f913eaf54c74
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| ca0ae89800ff957f76345c94eeb06d8c5643974b |
|
29-Aug-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Remove baseline models
Also remove submodels, which were to be used with baseline models.
Bug: 63905942
Test: Compiled and ran the tests on Marlin.
Change-Id: I54495e4e0dfbd6dabf049c8bdc2997600451b047
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 27e9be3904b034e422ee9b6ab70b35ea994d2b39 |
|
03-Aug-2017 |
Miao Wang <miaowang@google.com> |
Initial implementation of the following quantized ops.
- CONV_QUANT8
- DEPTHWISE_CONV_QUANT8
- AVERAGE_POOL_QUANT8
- MAX_POOL_QUANT8
- LOGISTIC_QUANT8
Additionally, added functions to plumb through quantization
parameters.
Bug: 63905942
Test: mm
Test: end-to-end MobileNet quantized test pass
Change-Id: Ib2753c68bf2c51467ae1c158b45541bcfdf10789
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 707dbd2d55f5dacf78ffb3ad7c8b3f37c2e9d758 |
|
25-Jul-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Pass the model as an HIDL component.
Major rework where instead of serializing the model and
passing it in a shared memory, we pass it via HIDL.
The runtime/test code does two run, one goes through the
default CPU path, the other through any available driver.
Note: The code in hardware/../sample was written by Michael Buttler.
Bug: 63905942
Test: Runs both the CPU path and through the sample driver.
Change-Id: Ie3ee975d33056ba299895b13193f4698a690dd04
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|
| 96775128e3bcfdc5be51b62edc50309c83861fe8 |
|
12-Jul-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
First implementation of the Neural Networks API.
This first version can run a simple query on the CPU either
via the fallback path or through a simulated driver.
This code has many deficiencies: single threaded, not all
validation are done, not going through HIDL, and not
enough unit tests. Expect more changes!
Test: Compiled and ran the unit tests
Change-Id: I9f6a485a2e7207aeb5f91a2904dcb4b7fd8a6f65
/frameworks/ml/nn/runtime/include/NeuralNetworksWrapper.h
|