Applications: Add Artificial Intelligence examples #34
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
It is possible for the FVP to output characters that cannot be decoded as UTF-8. This change ensures that when such characters are encountered they are stripped out returning the string without them. Signed-off-by: Hugues Kamba-Mpiana <[email protected]>
The source code will be the same for all apps. Signed-off-by: Hugues Kamba-Mpiana <[email protected]>
The configurations are added for Corstone 300 and 310 to be used for configuring the Ethos NPU by ML applications to be added. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
The updated version of the `arm-corstone-platform-bsp` project provides necessary changes to use IO on non-secure side for Corstone 300 and Corstone 310 platforms. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
Abstract the MPS3 LED native driver so they can be used by applications. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
Some IRQs are specific to the FVP version of corstone3xx platforms. The `CORSTONE300_FVP` and `CORSTONE310_FVP` macros should be defined for the `arm-corstone-platform-bsp` library in order to make these IRQs available. Signed-off-by: Ahmed Ismail <[email protected]>
Some ML applications will require more image size. Signed-off-by: Hugues Kamba-Mpiana <[email protected]>
Increase the clock rate from 60 Hz to 300 Hz mosty to benefit the execution speed of ML applications. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
A new patch is introduced to modify ISRAM0 to be non-secure application storage while the secure data would be placed inside Data tightly coupled memory (DTCM). The reason behind this change is that ML applications need more RAM memory than 1MB (ISRAM1 size) which is the only current RAM storage available. However, with this change we should have 2MB of RAM (ISRAM0 + ISRAM1 sizes) for the application (non-secure side). Signed-off-by: Ahmed Ismail <[email protected]>
This commit is meant to modify Corstone-300 linker files to be able to run ML applications, the changes are: - Use both ISRAM0 and ISRAM1 for non-secure data storage as ML applications would require more RAM space. - Increase the heap size in the Arm GNU Toolchain linker script to match Arm Compiler for Embedded's scatter file allocated heap size. Note that the greater heap size is necessary for Arm Compiler for Embedded builds of the ML apps. - Place the neural network models at the DDR space. - Modify AN552 GNU linker script to match Total Solutions AN552 GNU linker script as they should be matched to get ML applications working on Corstone-300 - VSI needs a buffer to transfer audio data from S to NS space, so in order to provide consistency, VSI audio buffer would be placed at the DDR space for Corstone-300 right after where the activation buffer for SRAM is placed. Signed-off-by: Ahmed Ismail <[email protected]>
Corstone-310 GNU and ArmClang linker scripts are modified to use macros which are more descriptive than magic numbers. Signed-off-by: Ahmed Ismail <[email protected]>
VSI needs a buffer to transfer audio data from S to NS space, so in order to provide consistency, VSI audio buffer would be placed at the DDR space for Corstone-310 right after where the activation buffer for SRAM is placed. Signed-off-by: Ahmed Ismail <[email protected]>
Hooks can now be added by applications to be notified when the OTA is active or not active. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
Provide caller of event init function information about the outcome of the initialisation by adding a return value to the initialisation API. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
This library is added to build Machine Learning (ML) applications. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
hugueskamba
force-pushed
the
feature/add_ml_applications
branch
from
fbdcf9e to
2c558fd
Compare
In order for ML applications to access the API headers for the ML Evaluation Kit libraries, various paths in the ML Embedded Evaluation Kit's source directory need to be made available to them as include directories. Helper libraries have been created with library archives and include directories added to those helper libraries. ML applications will link the helper libraries they need, in order to inherit the library archives for linking and include directories for API headers. Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
When performing quality checks with uncrustify, do not longer override the programming langauge with C. Instead, let the application detect the language on its own. This is because C++ files will be added for ML examples and the C language override for uncrustify introduces error when it encounters C++ files. Signed-off-by: Hugues Kamba-Mpiana <[email protected]>
Applications can now have: * `fail/pass_output.log` files to test the application output. * `fail/pass_ota_output.log` files to the test the application output when OTA is tested. Signed-off-by: Hugues Kamba-Mpiana <[email protected]>
* Add all source and header files to build the application * Update open-iot-sdk-toolchain version to include the `-fno-rtti` C++ flag * Remove no longer needed `--specs=nosys.specs` flag as include already in the updated version of open-iot-sdk-toolchain * Update build and run scripts to support keyword-detection application * Add documentation for the Keyword Detection application Signed-off-by: Hugues Kamba-Mpiana <[email protected]> Signed-off-by: Ahmed Ismail <[email protected]>
Signed-off-by: Hugues Kamba-Mpiana <[email protected]>
Signed-off-by: Hugues Kamba-Mpiana <[email protected]>
Add keyword-detection and speech-recognition testing with the available configuration (ETHOS/SOFTWARE) to the CI. Increase the timeout to accommodate the applications execution time as it takes longer. Signed-off-by: Ahmed Ismail <[email protected]>
hugueskamba
force-pushed
the
feature/add_ml_applications
branch
from
2c558fd to
e514fea
Compare
urutva
approved these changes
Jan 16, 2024
aggarg
approved these changes
Jan 16, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Add two Machine Learning applications:
See individual commits for more information.
Rebase and Mergeinstead ofSquash and Mergeto keep individual commits.Test Steps
Checklist:
Related Issue
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.