dlk: Deep Learning Kit
Project description
A Deep Learning Kit
This project is WIP, only provide some basic method, and is not been tested.
WARNING: The documents in `read the docs` is not complete( I do not know what I miss.). You can build the docs locally by read the docs/readme.md.
Install
pip install dlk
or
git clong this repo and do
python setup.py install
What is this project do?
- Provide a templete for deep learning (especially for nlp) training and deploy.
- Provide parameters search.
- Provide basic architecture search.
- Provide some basic modules and models.
- Provide basic deploy method.
More Feature is Comming
-
Add more documents.
-
Distill structure.
-
Adv training.
-
Add disable tokenizer post process.
-
Predict
- Complete the main predict code.
- Test.
- Convert to TorchScript.
- Convert to ONNXRT
- When the high config change the _name of the basic config, the base config should be coverd.
-
One
optimizer
different para groups use differentscheduler
s.- Ref diff_schedule
- Add loss schedule, get best checkpoint by loss.
-
Support LightGBM, it's maybe not necessary? Will split to another package.
- Make CRF and more modules which uses the op like
for
andif
to be scriptable(for
,if
) - Validating LSTM module is scriptable or not?
- Add UnitTest
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
dlk-0.0.7.tar.gz
(161.1 kB
view hashes)
Built Distribution
dlk-0.0.7-py3-none-any.whl
(252.9 kB
view hashes)