Mantra : A Deep Learning Development KitΒΆ
Mantra is a deep learning development kit that manages the various components in an deep learning project, and makes it much easier to do routine tasks like training in the cloud, model monitoring, model benchmarking and more. It works with your favourite deep learning libraries like TensorFlow, PyTorch and Keras.
You might like mantra if:
- You need to structure deep learning projects: versioning, monitoring and storage.
- You need devops tasks like cloud integration and file syncing taken care for you.
- You need to evaluate your model against benchmark tasks, e.g. CIFAR-10 accuracy.
π InstallationΒΆ
Mantra is a Python that you can install via pip:
$ pip install mantraml
It is currently tested on Python 3.5-7.
Additional dependencies you need to install are TensorFlow or PyTorch depending on which framework you want to use. If you want to use the TensorBoard feature of Mantra with PyTorch then you should also install TensorboardX.
π Get StartedΒΆ
Find a directory where you want to create a project and run:
$ mantra launch my_project_name
This will create a my_project_name directory with a folder structure like this:
data/
models/
tasks/
trials/
__init__.py
mantra.yml
README.md
settings.py
- The
data/
folder contains your datasets - The
models/
folder contains your models - The
tasks/
folder contains your tasks - The
trials/
folder contains trial data (data when you train a model) - The
mantra.yml
file contains project metadata - The
settings.py
file contains project settings
Now we have a project! To view the current project through the Mantra UI, execute the following from your project root:
$ mantra ui
Now that you are ready, itβs time to learn how Mantra models and datasets work!
- π€ Get started with Mantra models
- πΎ Get started with Mantra datasets
- π Get started with Mantra tasks
- π Get started with Mantra training