Followed a blog to create bi-LSTM model that that takes ELMo embeddings as inputs.
- Trained it on GPU.
- Used Cached tools to fasten the process of getting embeddings.
- Saved_model format does not work with tf hub models using Lambda layer. (Have to check further)
- note-to-self: when using TF hub models, have to disable eager execution and default behavior