TensorFlow2.0-from-zero-to-advanced

This repo is the excises of course TensorFlow2.0-from-zero-to-advanced. Thanks for the tutor of this course.

The outline of this course and excises lists:

Chapter01 Tensorflow Introduction and Environment Setup

1. What is Tensorflow?
2. Tensorflow version changes and tf1.0 architecture
3. Tensorflow2.0 architecture
4. Tensorflow VS pytorch
5. Tensorflow environment configuration
6. Google_cloud without GPU environment
7. Google_cloud_Remote jupyter_notebook configuration
8. Google_cloud_gpu_tensorflow configuration
9. Google_cloud_gpu_tensorflow mirror configuration
10. AWS cloud platform environment configuration

Chapter02 Tensorflow keras Excises

1. tfkeras brief introduction
2. Classification regression and objective function
3. Data reading and display of actual classification model
4. Model Construction of Classification Model
5. Data normalization of classification model
6. Callbacks
7. Regression model
8. Neural netorks
9. Deep neural network
10. Batch normalization, activation function, dropout
11. wide_deep model
12. Function API to implement wide & deep model
13. Subclass API to implement wide & deep model
14. Multi-input and multi-output excises of wide & deep model
15. Hyperparameter search
16. Manual implementation of hyperparameter search
17. sklearn package keras model
18. sklearn hyperparameter search

Chapter03 Tensorflow Basic API Usage

1. tf.constant
2. tf.strings and ragged tensor
3. sparse tensor and tf.Variable
4. self define loss function and DenseLayer
5. Use subclasses and lambdas to define levels separately
6. tf.function function conversion
7. @tf.function function conversion
8. Function signature and graph structure
9. Approximate derivative
10. tf.GradientTape Basic Usage
11. tf.GradientTape与tf.keras结合使用

Chapter04 Tensorflow dataset Usage

1. tf_data basic API usage
2. Generate csv file
3. tf.io.decode_csv usage
4. tf.data reads a csv file and uses it with tf.keras
5. tfrecord basic API usage
6. Generate tfrecords file
7. tf.data reads tfrecord file and uses it with tf.keras

Chapter05 Tensorflow Estimator Usage and tf1.0

1. Titanic problem analysis
2. feature_column usage
3. keras_to_estimator
4. Predefined estimator usage
5. Cross feature excises
6. TF1.0 computational graph construction
7. TF1.0 model training
8. TF1_dataset usage
9. TF1 self defined estimator

Chapter06 Convolutional Neural Network

1. Problems solved by convolution
2. Calculation of convolution
3. Pooling operation
4. CNN excises
5. Deep separable convolutional network
6. Deep separable convolutional network excises
7. 10monkeys dataset
8. Keras generator reads data
9. 10monkeys basic model building and training
10. 10monkeys model fine-tuning
11. Keras generator reads cifar10 dataset
12. Model training and prediction

Chapter07 Recurrent Neural Network

1. RNN and embedding
2. Data set loading and construction of vocabulary index
3. Data padding, model construction and training
4. Sequential problems and recurrent neural networks
5. Text Classification by RNN
6. Data processing for text generation
7. Model construction for text generation
8. Sample text for text generation
9. LSTM
10. Text classification and text generation by LSTM
11. Dataset loading and tokenizer for subword text classification
12. Dataset transformation and model training for subword text classification

Chapter08 Tensorflow Distribution

1. GPU settings
2. GPU default settings
3. Memory growth and virtual device excises
4. GPU manual settings excises
5. Distribution strategy
6. Keras distribution excises
7. Estimator distribution excises
8. Self define process excises
9. Distributed self define process excises

Chapter09 Tensorflow Model Saving and Deployment

1. TFLite_x264
2. Save model structure plus parameters and save parameters in practice
3. Keras model convert to SavedModel
4. Signature function convert to SavedModel
5. Signature function, SavedModel and Keras model convert to concrete function
6. tflite preservation and interpretation and quantification
7. tensorflowjs convert model
8. tensorflowjs build server and load model
9. Android deployment model

Chapter10 Machine Translation

1. seq2seq+attention model
2. Data preprocessing and reading
3. Convert string to id and dataset generation
4. Build Encoder
5. Build attention
6. Build decoder
7. Loss function and one step trainning
8. Model training
9. Model prediction
10. Transformer
11. Encoder-Decoder with Zoom click attention
12. Multi head attention with position encoding
13. Data preprocessing and dataset generation
14. Position encoding
15. Build mask
16. Build Zoom click attention
17. Build Multi head attention
18. Feedforward layer
19. Encoder layer
20. Decoder layer
21. Encoder model
22. Decoder model
23. Transformer
24. Self define learning rate
25. Model training and evaluation