I wish to use the TFBertTokenizer as a layer in my custom model where there is:
- Input Layer - Receiving String
- Tokenizer to process string tensor
- Bert model
- Other layers of my custom layer
The model is for classification problem and the intentions are to have a model that has both preprocessing as part of the model. I’m thinking this is possible since TFBertTokenizer is an in-graph tokenizer.
Please help.