An deep leaning util based on tensorflow
Project description
# mautil An deep learning util based on tensroflow. - easy to build model and do experiments - easy to debug - support for colab tpu - no need to transfer data to tf records(non tpu)
# requirements tensorflow >= 1.11
# Install pip install mautil
# Demo ## Text style transfer - There are two text style transfer models. CA is an implementation of the paper [Style Transfer from Non-Parallel Text byCross-Alignment](https://papers.nips.cc/paper/7259-style-transfer-from-non-parallel-text-by-cross-alignment.pdf). CAR model just use reinforcement learning to bypass gumbel softmax - run the text style transfer model in debug mode
python train.py -m CAR -dataset yelp -d
## TPU - see the [notebook](demo/colab/mautil_tfxl_colab_tpu.ipynb) about how to train a transformer-xl model for dataset wikitext103 - the demo is for colab cloud tpu, you need google cloud storage access to run the demo
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for mautil-0.2.1.linux-x86_64.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | fd773c839b81c8875ac138b17c1a27c82b9f0ca968039c2aefca43144b8c6c6b |
|
MD5 | 688c71947adc2a9ff8f22b6a75361c04 |
|
BLAKE2b-256 | ba5c3d8756a191f3b2f88a0f998186dcd391da23480c648732ee636d3ea3153a |