GRU layer support on ST Edge AI

[Dresult]

 

Hello everyone,

I have a GRU-based model that I trained using PyTorch and then exported to ONNX. I'm trying to run it on various ST boards.

According to this documentation: ST Neural-ART Operator Support
it seems that the Neural-ART accelerator does not support the GRU layer. However, from what I understand, other ST Edge AI boards should support it—with some limitations described here: ONNX Supported Ops

That said, when I try to compile the model using the ST Edge AI Developer Cloud, I get the following error:

TOOL ERROR: operands could not be broadcast together with shapes (512,) (256,)

This is the same issue I previously encountered here: Community Thread - GRU/LSTM Model Issue

At this point, I'm not sure whether GRU layers are simply not supported at all, or if there's a way to make them work under specific conditions.

I'm attaching the ONNX model in case it helps.

Thanks in advance for any support or suggestions!

[Julian E.]

Hello @Dresult,

 

Sorry for all these issues. GRU and LSTM are still badly supported.

Our team is working on it but also on a lot of other subject.

I opened an internal ticket with your model. 

 

I will try to keep you informed but it may take a lot of time before being worked one.

 

Have a good day,

Julian

[Dresult]

Thanks Julian, I really appreciate that.

Don't worry, I asked mainly to understand if it was a problem of my model or not...

[Julian E.]

Hello @Dresult,

 

This issue was fixed internally and the fix should be part of the next st edge ai core update (v2.2)

 

Have a good day,

Julian