代码拉取完成,页面将自动刷新
Automatic data generation with CVAEs -- Internship by Stéphane
Requirements: Python3.6, pip
virtualenv venv
. venv/bin/activate
pip install -e .
You might need to download some NLTK resources:
>>> import nltk
>>> nltk.download('punkt')
The abstract class automatic-data-generation.data.base_dataset.py
provides the interface for representing a training dataset. To implement a new dataset format, write a class inheriting from Dataset and implement its abstract methods.
Use the script automatic-data-generation.train_and_eval_cvae.py
to train a model, generate sentences, and evaluate their quality.
python automatic_data_generation/train_and_eval_cvae.py --dataset-size 200 --n-generated 1000 --n-epochs 5 --none-size 100 --none-type subtitles --restrict-to-intent GetWeather PlayMusic
--dataset-size
: number of sentences in the training dataset--none-size
: number of None sentences to be added to the training dataset--none-type
: type of None sentences--restrict-to-intent
: list of intents to filter on for training--n-epochs
: number of epochs for training--n-generated
: number of generated sentencesAn folder will be created with the following elements:
model
: a folder with a model.pth
file and its associated config.json
tensorboard
: a folder with the checkpoints for tensorboardrun.pkl
: a dictionnary with every runtime parameterstrain_*.csv
: the training datasettrain_*_augmented.csv
: the training dataset augmented with generated sentencesvalidate_*.csv
: the validation dataset此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。