Hi DeGirum Team,
I’m using the new Cloud Compiler (Early Access) to convert an Ultralytics YOLOv11m model for my AI Hat (Hailo-8L).
I’ve successfully compiled a YOLOv11n model using the same parameters. However, when I try to compile my YOLOv11m model (best_epoch_014_001746 (1).pt, trained with Ultralytics 8.3.160), it consistently fails with an “Unknown exception.”
Here’s the error I receive:
{
“error”: “Unknown exception. Please review your inputs, and contact DeGirum support.”,
“input parameters”: “model_file: best_epoch_014_001746 (1).pt —— model_name: Test3 —— model_version: 1 —— image_height: 640 —— image_width: 640 —— cloud_zoo_url: amil_biju/Yolo_Drone_Model —— runtime_device_quant: [hailort-hailo8l-quant]”,
“result”: “FAILED”
}
Key details:
Model Type: Ultralytics YOLOv11m (best_epoch_014_001746.pt, ~115 MB)
Input Resolution: 640x640
Target: HailoRT, Hailo-8L, Quantized
Success Case: YOLOv11n compiles without issues using the same settings.
Given your announcement states YOLOv11 (n, s, m, l, x variants) are supported, I suspect this might be an edge case bug with the ‘m’ variant’s compilation or a backend resource limitation, just a theory.
Could you please investigate this? Any guidance or a timeline for a fix would be greatly appreciated.
Thanks,
AB