Tensorflow NotImplementedError error with Camembert

Question

Tensorflow NotImplementedError error with Camembert

Kirire opened this issue 4 years ago · comments

Hi,

I think I have detected a bug when we want to save the model (save_nn_model / load_nn_model) in the melusine.models.train module with the pretrained Camembert hidden layer. I received a Python NotImplementedError error from Tensorflow library (from get_config). I think this is because TFCamembertModel is a Subclass Networks and isn't a Graph Networks. It seems there are answers in this topic.

Best regards

Tiphaine Fabre · Answer 1 · Thu Oct 22 2020 20:54:11 GMT+0800 (China Standard Time)

Hi,

Thank you so much for finding a bug! We'll take a look as quickly as possible and fix it.

Best regards

Tiphaine Fabre · Answer 2 · Tue Nov 17 2020 22:27:57 GMT+0800 (China Standard Time)

Hi @Kirire, we released Melusine 2.2.5 last week and it should fix this issue. The current transformers library does not allow to use the keras save function at its full potential. We implemented a work around. I will close this issue now.