Describe the solution you'd like
Try DETR training with small dataset, if verified, add as a new detection module
also there might be several Vortex limitation for training DETR.
Additional parameters not covered by existing model parameters in configuration files, may need to add additional_args to model config
Forwarding params group to optimizer. Current condition only allow model.parameters() directly forwarded to optimizer. May need additional dedicated model_components key for params gorup so that in the trainer can check whether it exist or not and forward params group to the optimizer if exist rather than whole model.parameters(). Important to be noted that params group modification must be done under create_model_components -> in the model's object side
validator (BoundingBoxValidator) arguments passed to model doesn't have the same behavior as vortex runtime for predictor model (not generalized).
Describe the solution you'd like Try DETR training with small dataset, if verified, add as a new detection module
also there might be several Vortex limitation for training DETR.
additional_argstomodelconfigBoundingBoxValidator) arguments passed to model doesn't have the same behavior as vortex runtime for predictor model (not generalized).