I'm trying to use the benchmarking tool in the developer cloud (https://stm32ai-cs.st.com).
I have a model exported in several formats (fp32, fp16, full 8 bit, etc.)
The input shape is (1, 128, 384, 3).
The model works fine and other tools like Edge Impulse are able to benchmark my model without issues on a variety of hardware.
However, when I upload the model and try to run in developer cloud, I get the following output:
>>> stm32ai validate --model EfficientNet_v1_full_integer_quant_with_int16_act.tflite --allocate-inputs --allocate-outputs --relocatable --compression none --optimization balanced --name network --workspace workspace --output output
Neural Network Tools for STM32AI v1.7.0 (STM.ai v8.0.1-19409)
INTERNAL ERROR: index 1 is out of bounds for axis 0 with size 1
I'm not sure why the tool is over-indexing in the first dimension.
Can someone point me in the right direction?