Error when optimising v11/v26 nano models in Jetson Nano Orin

Discussion
,
1

I optimise models following the work flow ultralytics API → onnx → .engine (trtexec command). For v8n, no problems were obtained.

For v11n/v26n, things wen’t no so well. I obtained for both the same error: on the Attention Layer I did not have enough workspace (4098MiB were given) for optimising the model (this happened when using the trtexec command).

My Jetson Nano has near 8000MiB. I tried with 6000MiB aprox. Still same issue. Does this usually happen with these models and this hardware specs or is just me?

2

Are you using latest Ultralytics? You should try latest Ultralytics and export directly with Ultralytics to TensorRT.

Also can you post the output of yolo checks