Transfer-Learning-for-NLP-with-TensorFlow-Hub
Transfer learning is a research problem in machine learning that focuses on storing knowledge gained while solving one problem and applying it to a different but related problem. For example, knowledge gained while learning to recognize cars could apply when trying to recognize trucks.
TensorFlow Hub is a repository of pre-trained TensorFlow models.
Objectives
In this project, we will use pre-trained models from TensorFlow Hub with tf.keras for text classification. we will Use pre-trained NLP text embedding models from TensorFlow Hub Perform transfer learning to fine-tune models on real-world text data, build and evaluate multiple models for text classification with TensorFlow, and visualize model performance metrics with Tensorboard.
Prerequisites: Python programming language, be familiar with deep learning for Natural Language Processing (NLP), and have trained models with TensorFlow or and its Keras API.
Dataset
A downloadable copy of the Quora Insincere Questions Classification data can be found https://archive.org/download/fine-tune-bert-tensorflow-train.csv/train.csv.zip. Decompress and read the data into a pandas DataFrame.
Project Structure
The hands on project on Transfer Learning for NLP with TensorFlow Hub is divided into following tasks:
- Task
0️⃣ 1️⃣ Introduction to the Project - Task
0️⃣ 2️⃣ Setup your TensorFlow and Colab Runtime - Task
0️⃣ 3️⃣ Load the Quora Insincere Questions Dataset - Task
0️⃣ 4️⃣ TensorFlow Hub for Natural Language Processing - Task
0️⃣ 5️⃣ &0️⃣ 6️⃣ Define Function to Build and Compile Models - Task
0️⃣ 1️⃣ Train Various Text Classification Models - Task
0️⃣ 7️⃣ Compare Accuracy and Loss Curves - Task
0️⃣ 8️⃣ Fine-tune Model from TF Hub - Task
0️⃣ 9️⃣ Train Bigger Models and Visualize Metrics with TensorBoard
Visualize model performance metrics with TensorBoard
Accuracy Curves for Models
Loss Curves for Models
Visualization of Metrics with TensorBoard
epoch accuracy
epoch loss
Connect with me:
