Dockerise existing Bert model training task

Job ID: 33494156

Budget: $80 – $240 HKD

I've got an existing training Bert model task, it is a binary classification, but I want the code to look friendly and reusable. And ultimately I want to dockerizse everything that I can run on the GPU cloud for the training process.

The task to-do list:
Add custom token and weight to the given Bert model
Restructure the code to be friendly and reusable
Dockerise everything

PS:
Existing code is written in python
Deep Learning framework: Tensorflow 2.x
the current model: https://tfhub.dev/tensorflow/bert_multi_cased_L-12_H-768_A-12/4

and its preprocess
https://tfhub.dev/tensorflow/bert_multi_cased_preprocess/3
Related categories: Python Tensorflow Docker Compose NLP