How do I use a custom trained model to then train on again, using a different dataset?

TensorFlow General Discussion
,
1

I have a ResNet50 model that was used for transfer learning on a custom dataset. That new, custom model was saved as a SavedModel format. I want to use that model to then use transfer learning again, on a different dataset. The previous dataset and the new one I want to use are very similar, so I think the model will train better. How do I load the custom model again so I can train?

This is how the ResNet50 model was loaded in (using TF 1.15, but I’m using TF 2.10 right now)

net = ResNet50(include_top = False, weights = 'imagenet', input_tensor = None,
              input_shape = (IMAGE_SIZE[0], IMAGE_SIZE[1], 3)),
3

Hi @GB1

Welcome to the TensorFlow Forum!

Could you please share the standalone code of ResNet50 model and the new Custom model to replicate and understand the issue better? Thank you.

5

Thanks for the response! But I ended up going in a different direction when I found out that the custom model was saved with a ‘serve’ tag and I couldn’t load it in to train off of, without having to rewrite and retrain to save that custom model in a ‘train’ tag. Thanks again!

1 Like