wavenet
所属分类:音频处理
开发工具:Python
文件大小:0KB
下载次数:0
上传日期:2022-07-05 13:02:01
上 传 者:
sh-1993
说明: 小波网,,
(wavenet,,)
文件列表:
audio_fidelity_test.py (541, 2022-07-05)
data/ (0, 2022-07-05)
data/feather_dot.wav (7735340, 2022-07-05)
dataset.py (1032, 2022-07-05)
generate.py (3180, 2022-07-05)
params.py (666, 2022-07-05)
preprocess.py (4740, 2022-07-05)
test_util.py (1479, 2022-07-05)
train.py (5516, 2022-07-05)
util.py (705, 2022-07-05)
wavenet/ (0, 2022-07-05)
wavenet/__init__.py (0, 2022-07-05)
wavenet/model.py (4255, 2022-07-05)
wavenet/model_test.py (4052, 2022-07-05)
# WaveNet for Music Generation.
## Preprocesing data.
preprocess.py generates tfrecord from audio files.
First create a directory and put audio files to generate training data from.
Then pass that directory as the first argument to the script. Supported audio
formats are "wav" and "mp3".
```
# Run python3 preprocess.py --help to see all options.
# This creates tfrecord files in .
python3 preprocess.py [options]
```
## Training the model.
train.py trains WaveNet using tfrecord data created by preprocess.py and saves
the trained weights to disk.
```
# Run python3 train.py --help to see all the options.
python3 train.py
```
## Generating audio.
generate.py generates audio iteratively by making the model predict the value of
the next timestamp. It starts with an arbitrary timestamp 1 and make the model
predict the next timestamp. Then it appends the output of the model to the input
makind the model generate one timestamp at a time.
By default the script generates an audio clip of 5 seconds.
```
# Run python3 train.py --help to see all the options.
python3 generate.py
```
## Running unit tests.
Make sure these tests pass after modifying model.py or util.py.
```
pytest test_util.py
python3 wavenet/model_test.py
```
近期下载者:
相关文件:
收藏者: