Aioway is a work in progress, builds on top of the original koila (moved to a branch). The torch team built FakeTensor which overlaps a lot with koila's functionality, so it's no longer maintained. See the rationale in the koila branch.
Conceptually, aioway works in a similar way, but instead of Tensor ops, aioway focuses on a higher level, on algorithm building. See below for the promised features:
- Simple and declarative, yet reproducible.
- Detects the tasks at hand, resource available, and select the best algorithms and models.
- The models built from aioway would be white box (explanable).
- Allows you to scale up the model size, and to different machines.
- Extensible with custom pytorch.
For the pre-release version (v0.0.*), see project for more details.
In the recent years, machine learning's entry barrier has gotten higher, rather than lower. With the increasing number of algorithms and libraries and models, it's no wonder qualified data scientists are rare because you would need years of training to keep up to the status quo.
We designed Aioway in a way such that instead of thinking about how to do ML, you specify what to do. Instead of focusing on what algorithms and models to use, Aioway allows you to focus on the use cases by taking into account the context of the problem, and perform compliation according to the data to ensure good performance. Automatically.
Contributing is of course welcome. Please see the contributing guide and follow the code of conduct.