Illegal instruction (core dumped) using tflite on pynq z1

Support
1

Hello all,
I am trying to run tflite on pynq z1 but I am getting an error "Illegal instruction (core dumped) when I am calling the function interpreter.invoke. I have built the compatible tflite runtime from scratch following the instructions from the official website Build TensorFlow Lite Python Wheel Package
but could not fix this issue.
Anyone got a similar problem?
Thank you in advance

2

Hi,I have the same problem.
Did you already solve the problem?

3

Hi @Arif_PENS, not yet, I am wondering if the pynq z1 cpu supports this.

1 Like
4

@Lotfi

Along my basic tutorial develop, I recall the ZYNQ itself cannot handle that much of computation.
As the DDR3 itself only got 1GB. So what would you expect is quite expectable.
So I will highly suggest train on host and inference on ZYNQ.
Meantime, if you are lazy as me you can use the weight load function only via the exported weight file.

it is been a while, so any new ARM-based low tensorlite might make me wrong.

As always, enjoy~ =]

1 Like
5

Thank you for your answer @briansune I will try this out.

1 Like
6

Is this still an open issue? I recently encountered the same error. The issue arises because the PYNQ-Z1 uses an ARM Cortex-A9 processor, which is based on the ARMv7 architecture and supports the VFPv3 floating-point unit. However, the default TensorFlow Lite (TFLite) build is compiled for VFPv4.

As a result, using the default TFLite build causes an illegal instruction error.

Solution: Build TensorFlow Lite specifically for the ARM Cortex-A9 architecture.

I’ve written a detailed blog post explaining how to fix this issue. Please give it a read—I’m confident it will resolve your problem. If not, feel free to reach out. I’ve included my contact details at the end of the blog.

Blog Link: Cross-Compiling TensorFlow Lite for ZYNQ 7000 (ARMv7-A + VFPv3) | by Harshith N | May, 2025 | Medium

1 Like