search

tensorflow / tflite-micro

Infrastructure to enable deployment of ML models to low-power resource-constrained embedded targets (including microcontrollers and digital signal processors).
Apache License 2.0
1.74k stars 770 forks source link

The trained model structure is inconsistent #2531

Closed huxiangjs closed 3 months ago

huxiangjs commented 3 months ago

Hello,

According to the instructions of train_micro_speech_model.ipynb, I am training a Micro Speech model, but the trained model cannot run successfully.

When I compared the original micro_speech_quantized.tflite model with the model.tflite model generated by my training, I found their differences.

train_micro_speech_model.ipynb: image

model.tflite(My trained model): image

As you see, DepthwiseConv2D becomes Conv2D.

What did I do wrong?

Please give me some guidance, thank you very much.

huxiangjs commented 3 months ago

I lowered the tensorflow version to 1.15.0 and this issue has been successfully solved. :)