I’m interested in fine-tuning EfficientFormer for an image classification task. However, I see in the model docs under the
EfficientFormerForImageClassificationWithTeacher class that “This model supports inference-only. Fine-tuning with distillation (i.e. with a teacher) is not yet supported.” Does this mean that fine-tuning is not supported for the teacher model or that fine-tuning is just not yet possible for this model in entirety?
I attempted to fine-tune with the Trainer, but I did get an error about
EfficientFormerForImageClassificationWithTeacher.forward() not taking the param “labels” which makes me think it’s the latter. If this is the case and if no one internally is working on additional training capability, I’d be happy to take a swing at it if I could be pointed to some references/guidance documentation.
Thanks in advance!