Tested at Yolov8x and Yolo11

Hello again,
Thank you very much for your answer. Today I tried yolov8x again and this time I managed to run it. I think I made a mistake yesterday. Is yolov8x the biggest model that can run? Yolov8's FLOP value is 344. Have you tried a larger model than this? 

Thank you, I will follow the developments.

That’s awesome news ​@oguz, great work!  What did you need to do differently to get it running this time?

Actually I didn't change anything. 😅 To explain the error more accurately, I deployed again and this time it worked. I guess I did something wrong the first time without realizing it, or a reboot fixed the problem.

​@oguz , 
You can see the full list of models we actively support here https://github.com/axelera-ai-hub/voyager-sdk/blob/release/v1.2.5/docs/reference/model_zoo.md#object-detection. 
As you figured out by testing yolov8x, we support many more but do not include all of them in our list of reference models. Model size (no. params) and needed operations (OPS) are two different things. We do not have these numbers in our model-zoo docs but for most models you can find this as public information. I would say yolov8x is indeed already a very demanding model (in terms of OPS). 
Finally, note that we often refer to ‘OPS’ instead of ‘FLOPS’ since metis is mostly performing INT8 operations. 

By the way, I forgot to mention this. The models I tried are Yolo's segmentation models.

Could you please provide more information how you managed to run yolov8x?

I copied the Yolov8l yaml file and changed everything related to 'L' to 'X'. I added yolov8x to the test_yaml_mc_card_name.py file. I created the folder with the weights as in yolovl and put them in it. Then I ran it the same way.
 

By the way, don't forget to look at the “make” command.

Nice! Thanks for helping out with this request ​@oguz!

Did that work for you ​@Berrisius? What's your project?

Hi, I haven’t been able to test it out yet since it was weekend, will get back to you with the details during the week.