q-prashant

In my experience, using `dynamo=True` has had huge loss in accuracy in downstream formats (onnx and trt) for detection models like DFINE.

For me, I get ```console ... (1.7863) loss_ce_enc_unscaled: 0.4922 (0.5725) loss_bbox_enc_unscaled: 0.0028 (0.0043) loss_giou_enc_unscaled: 0.1738 (0.2304) cardinality_error_enc_unscaled: 0.7500 (0.7780) Accumulating evaluation results... Traceback (most recent call last): File "/home/quidich/Documents/train_rf.py", line...

This was too easy to setup and training is ongoing. Fingers crossed for the model to come out as expected. Thanks @SebastianJanampa

yup, working perfectly now. It was a good decision to switch from official repo to yours!