TFLite Quantization Aware Training

[Valerian G.]

Hello,

I successfully used a post-training quantized TFLite Model in X-CUBE-AI. Nonetheless, I can't make it work with a quantization aware training model, because I get the following error when I try to analyze or generate code : " TOOL ERROR: Unknown layer type FAKE_QUANT, stopping ".

Is X-CUBE-AI compatible with quantization aware training model ? If so, how am I supposed to export the model from TFLite so that it doesn't contain FAKE_QUANT layers ?

Sincerely