The Heart of the Machine editor, participation: The source of thought.

TensorFlow now has an official Chinese version of its Tutorials, and previous English versions of its Tutorials have Chinese translations. For those of you still hanging around TensorFlow, you can now learn about popular models from a Chinese tutorial.

TensorFlow recently offered its official Chinese version of Tutorials and Guides. The tutorials focus on the basic concepts of TensorFlow and the simple implementation of various basic models, which are implemented using easy-to-understand high-level apis such as Keras. The guide provides an in-depth introduction to how TensorFlow works, including high-order API, Estimator, low-order API, and TensorBoard.

The address of the project: tensorflow. Google. Cn/tutorials /?…

Chinese Version Tutorial

TensorFlow is an open source machine learning library for research and production. TensorFlow provides a variety of apis for beginners and experts to develop on desktop computers, mobile devices, or the cloud. The purpose of the Chinese tutorial is to provide beginners with the knowledge to understand and get started with TensorFlow, including implementing basic classification and regression models with Keras, building customized neural networks with Eager Execution, and building large-scale machine learning with Estimator.

The first basic classification describes how to use a fully connected network to process a Fashion MNIST data set. This tutorial provides a complete introduction to the main flow of TensorFlow. The remaining four initial tutorials will also walk you step-by-step through how to use it.

In addition to entry-level Keras, Eager Execution for research and Estimator for large-scale training are also introduced in Chinese. However, they will also offer some Github projects, most of which are still in English, with only the introductory articles in Chinese.



What’s left is a lot of leading-edge models, many of which are provided with Github addresses or Colab tutorial addresses, so most of them are in English, but once you’re in the door, you’ll be able to understand them. In total, these tutorials can be divided into 18 models, including generation model, visual model, sequence model and data representation model.

Finally, the tutorial section provides a follow-up learning plan, including a detailed understanding of TensorFlow and machine learning. TensorFlow provides guidance, tensorflow. js, TFLite and other learning paths, while Machine learning provides CS231n, Machine Learning Crash Course, CS 20 and other course recommendations.

Chinese guide

If the reader already has a good foundation, we may encounter many specific problems in practice, such as calling TPU, using static computation diagrams, or using TensorBoard for visualization. When encountering these specific problems, we can take a look at the TensorFlow guide, which may provide a detailed introduction.

The Chinese guide is mainly divided into the following parts, which are introduced in Chinese, so it is quite easy to read.

High-level API

  • Keras: TensorFlow high-level API for building and training deep learning models.

  • Eager Execution: An API for writing TensorFlow code as commands, just like NumPy.

  • Estimator: A high-level API that provides a fully packaged model ready to perform large-scale training and production.

  • Import Data: A simple input pipe for importing your data into the TensorFlow program.

Estimator

  • Estimator: Learn how to use Estimator for machine learning.

  • Pre-created Estimator: The basics of pre-created Estimator.

  • Checkpoint: Saves training progress and continues training or inferences from where it was saved.

  • Feature columns: Handle various types of input data without making changes to the model.

  • Estimator’s data set: Input data using Tf.data.

  • Create a custom Estimator: Write your own Estimator.

The accelerator

  • Using the GPU: Describes how TensorFlow assigns operations to devices and how to manually change such assignments.

  • Using TPU: Describes how to modify the Estimator program to run on TPU.

Low API

  • Introduction: Introduces the basics of how to use the low-order TensorFlow API in addition to the higher-order API.

  • Tensors: Describes how to create, manipulate, and access tensors (basic objects in TensorFlow).

  • Variables: details how to represent shared persistent state in a program.

  • Data Flow diagram: This is TensorFlow’s representation of computations as dependencies between operations.

  • Sessions: TensorFlow’s mechanism for running data flow diagrams across one or more local or remote devices. Be sure to read and understand this unit if you program using the low-level TensorFlow API. If you program with a high-level TensorFlow API (such as Estimator or Keras), the high-level API creates and manages graphs and sessions for you, but it still helps to understand graphs and sessions.

  • Save and Restore: Describes how to save and restore variables and models.