0.7.0

[jonnor]

What's Changed

• Update to MicroPython 1.24 by @jonnor in Update to MicroPython 1.24 #21
• cnn: Fix hardcoded outputs size leading to corrupted predictions by @jonnor in cnn: Fix hardcoded outputs size leading to corrupted predictions #23
• CNN: Fix wrong output values due to missing dequant by @jonnor in CNN: Fix wrong output values due to missing dequant #25
• trees: Switch to output prediction probabilities by @jonnor in trees: Switch to output prediction probabilities #26
• Example for Human Activity Recognition improved. Including data recording functionality and training custom models https://github.com/emlearn/emlearn-micropython/tree/master/examples/har_trees

Full Changelog: 0.6.0...0.7.0

---

This discussion was created from the release 0.7.0.

[Godswill-IoT]

Does this version work on micropython 1.24.1? I have trained a functional autoencoder neural network and split it into sequential encoder and decoder submodels and converted them using latest emlearn module to deploy on esp32 wroom. However, when I load the models using emlearn_cnn.new( ) I get a tm_load error. this is the loading code:

    # Loading the encoder model
    with open(encoder_model_path, 'rb') as f:
        model_bytes = f.read()
    model_data_array = array.array('B', model_bytes)
    encoder_model = emlearn_cnn.new(model_data_array)
    
     # Loading the decoder model
    with open(decoder_model_path, 'rb') as f:
        model_bytes = f.read()
    model_data_array = array.array('B', model_bytes)
    decoder_model = emlearn_cnn.new(model_data_array)

[jonnor]

@Godswill-IoT no known issues with 1.24.1. However note that I have now released 0.8.0 - https://github.com/emlearn/emlearn-micropython/releases/tag/0.8.0 - which is meant for MicroPython 1.25.

The CNN modules are now provided both for int8 and fp32 versions - and it is important to have the appropriate module for the model files used. If you have problems with this in the latest version, please file an issue with details on how to reproduce. Having an CNN autoencoder example would also be great for the docs.

[Godswill-IoT]

noted, thank you Mr Nordby

…

On Sat, May 31, 2025 at 6:27 PM Jon Nordby ***@***.***> wrote: @Godswill-IoT <https://github.com/Godswill-IoT> no known issues with 1.24.1. However note that I have now released 0.8.0 - https://github.com/emlearn/emlearn-micropython/releases/tag/0.8.0 - which is meant for MicroPython 1.25. The CNN modules are now provided both for int8 and fp32 versions - and it is important to have the appropriate module for the model files used. If you have problems with this in the latest version, please file an issue with details on how to reproduce. Having an CNN autoencoder example would also be great for the docs. — Reply to this email directly, view it on GitHub <#28 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BQTUX2ZRH37HCJBGCDMBZZL3BHJ6TAVCNFSM6AAAAABUO65EJ2VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGMZSHE4TCMQ> . You are receiving this because you were mentioned.Message ID: ***@***.*** com>