microbit-foundation / pxt-ml-extension-poc

Experimental Makecode Extension to test machine learning integration with the micro:bit

This extension is experimental and is being used for testing purposes.

Building locally

Ensure you have the required toolchain to build for V1 and V2 (arm-none-eabi-gcc, python, yotta, cmake, ninja, srec_cat) or docker.

git clone https://github.com/microbit-foundation/pxt-ml-extension-poc
cd pxt-ml-extension-poc
npm install pxt --no-save
npx pxt target microbit --no-save
npx pxt install
PXT_FORCE_LOCAL=1 PXT_NODOCKER=1 npx pxt

For the V1 build Yotta can hit the GitHub rate limits quite easily if the project is built from a clean state more than once. A V2-only build can be performed with the PXT_COMPILE_SWITCHES=csv---mbcodal environmental variable.

PXT_FORCE_LOCAL=1 PXT_NODOCKER=1 PXT_COMPILE_SWITCHES=csv---mbcodal npx pxt

Build flags

Model predictions per second

By default the model will run every 250 ms, to change this value the ML_INFERENCE_PERIOD_MS config can be modified.

{
    "yotta": {
        "config": {
            "ML_INFERENCE_PERIOD_MS": 250
        }
    }
}

Model events

By default this extension configures the Model prediction events to not be queued for the same event. So if an event raised when its handler is still running it will be dropped.

{
    "yotta": {
        "config": {
            "ML_EVENT_LISTENER_DEFAULT_FLAGS": 32
        }
    }
}

The values are defined in the codal-core/inc/core/CodalListener.h file:

#define MESSAGE_BUS_LISTENER_REENTRANT              0x0008
#define MESSAGE_BUS_LISTENER_QUEUE_IF_BUSY          0x0010
#define MESSAGE_BUS_LISTENER_DROP_IF_BUSY           0x0020
#define MESSAGE_BUS_LISTENER_NONBLOCKING            0x0040
#define MESSAGE_BUS_LISTENER_URGENT                 0x0080

Built-in ML model

The MLRUNNER_USE_EXAMPLE_MODEL flag can be configured as described in: https://github.com/microbit-foundation/pxt-ml-runner-poc#built-in-ml-model

Debug messages

To enable debug print from this extension, add the following into your pxt.json file:

{
    "yotta": {
        "config": {
            "ML_DEBUG_PRINT": 1
        }
    }
}

License

This software made available under the MIT open source license.

SPDX-License-Identifier: MIT

Code of Conduct

Trust, partnership, simplicity and passion are our core values we live and breathe in our daily work life and within our projects. Our open-source projects are no exception. We have an active community which spans the globe and we welcome and encourage participation and contributions to our projects by everyone. We work to foster a positive, open, inclusive and supportive environment and trust that our community respects the micro:bit code of conduct. Please see our code of conduct which outlines our expectations for all those that participate in our community and details on how to report any concerns and what would happen should breaches occur.

Metadata (used for search, rendering)

• for PXT/

<script src="https://makecode.com/gh-pages-embed.js"></script><script>makeCodeRender("{{ site.makecode.home_url }}", "{{ site.github.owner_name }}/{{ site.github.repository_name }}");</script>