STM32N6570-DK missing from the Edge AI Developer Cloud since November 7th
Hello,
I am a student writing a research paper on the STM32N6.
I want to find the model size from which the Neural-ART NPU runs a network faster than the Cortex-M55 alone, and how the layer type and operators that fall back to the CPU shift this point.
For this, I benchmark synthetic int8 models (dense, convolutional, and dense with one fallback layer) on the STM32N6570-DK in the Edge AI Developer Cloud. I use the REST API through the stm32ai_dc client from the STM32 model zoo services, with ST Edge AI Core 4.1.0. Each model is built once for the Cortex-M55 alone and once for Neural-ART, and each build runs in 3 separate benchmark jobs of 10 inferences each. I run at most 4 jobs in parallel.
On November 7th, 258 jobs ran successfully. At about 17:06 CET (16:06 UTC), 4 jobs I had just submitted stayed "in_queue" for more than 25 minutes. Since 17:33 CET, the STM32N6570-DK has been missing from the board list, while the 24 other board types are still listed.
It was still missing on November 8th at 09:45 CET. I have 3 questions:
1. Why is the STM32N6570-DK not available at the moment? Is it maintenance, a hardware problem, or a limit on my account?
2. When will it be available again?
3. Is there a limit on how many benchmarks an account can run, per day, per month or in total? I need 192 more jobs to finish my measurements.
Is the board missing for anyone else too?
If my jobs put too much load on the STM32N6570-DK boards, please let me know. I can run fewer jobs in parallel or run them at other times.
Thank you for providing the Developer Cloud. It makes this study possible.
Best regards