Jun 01, 2021 Article blog
This article was reproduced to Know ID: Charles (Bai Lu) knows his personal column
Download the W3Cschool Mobile App, 0 Foundation Anytime, Anywhere Learning Programming >> Poke this to learn
T_T people who don't have a scientific research dream come to the water in the middle of the night
Let Python learn to write songs and create songs
It's pure entertainment
Adapted from a tutorial on PyTorch's official website, but I wrote it in TF, and then the generated English became the generated Chinese
Let's Go~~~
Baidu web download link:
https://pan.baidu.com/s/1VUEFR82CqyDj2w9SS4xnFA
Password: 3dmx
Python version: 3.6.4
Related modules:
tensorflow-gpu module;
numpy module;
and some Python's own modules.
The TensorFlow-GPU version is:
1.6.0
Install Python and add it to the environment variable, and pip installs the relevant modules that are required.
Introduction to the principle
For an introduction to RNN, please refer to my previous article:
Python implements a simple machine translation model (click on the blue word to enter)
Here we train using the Char RNN model.
That is, the special case input and output in the Seq2Seq model is equally long.
That's what it looks like in the following image (source network):
The implementation process is detailed in the source code in the relevant file.
Run the Char_RNN.py file in the cmd window and enter the relevant information as prompted.
The training-related parameters should be adjusted in the config.py file:
The data path used for the training and the warm-up data used to generate the new text are adjusted in the Char_RNN.py file:
Training demo video:
Because it took so long, I only recorded the first part of the training.
The final training result is shown in the following image:
Because time and resources are limited.
Only all of Jay Chou's lyrics are shown here as samples for post-training tests.
Of course, you can find other text as training material to generate different styles of "articles", "lyrics" and so on
The results are as follows:
(
Warm up the lyrics as: it's the eaves that have escaped the rain with you
)
Well, I don't know what it's talking about T_T
A case of failure T_T
Probably the model is too low. A nd the training data was downloaded directly from the Internet, and there was no pre-processed T_T. In the middle of the sudden emergence of a bracket I am also a T_T
Play around with the T_T
Interested students can try to change the more complex model
Then the training of the language is best to deal with the . .
Haven't written TF for a long time, the code may be a little bad T_T
That's it