STM32 Cube AI Studio FP32 Model Compilation Issue

[RanjithRemasan]

Hi

I have FP32 TFLite model. I can able to compile quantised model with STM32 Cube AI studio. i am getting some error while compiling FP32 model with STM32 Cube AI studio. Please let me know the steps to build FP32 model and make the C files correspond to the model.

 

[Julian E.]

Hi @RanjithRemasan,

 

What is the error that you get?

 

Have a good day,

Julian

[RanjithRemasan]

HI @Julian E. 

 

Please see the error log. i have tried to compile Resnet8 

 

[Julian E.]

Hi @RanjithRemasan, 

 

Could you share your model (in a .zip)?

Based on the error message it seems that the st edge ai core fails to replace your float32 input to a uint8 one (in the .sh, in the generate command you have "--input-data-type uint8")

 

I tried it with our resnet8 and got no issue.

stm32ai-modelzoo/image_classification/resnet/ST_pretrainedmodel_public_dataset/cifar10/resnet8_32_tfs at main · STMicroelectronics/stm32ai-modelzoo · GitHub

 

This model has a uint8 input, unlike your model. Maybe it causes the issue.

But it would be strange that it fails to replace the float32 input as the "--input-data-type uint8" is made for that...

 

Have a good day,

Julian