So far, tfimm allows to create and initialize keras models using default timm weights as follows:
tfimm.create_model(TIMM_MODEL_NAME, pretrained="timm")
It is also useful to be able to load a fine-tuned timm model. This is what I implemented and would like to see in the future releases. I also added a Jupyter notebook to demonstrate usage.
So far,
tfimm
allows to create and initialize keras models using defaulttimm
weights as follows:tfimm.create_model(TIMM_MODEL_NAME, pretrained="timm")
It is also useful to be able to load a fine-tuned
timm
model. This is what I implemented and would like to see in the future releases. I also added a Jupyter notebook to demonstrate usage.