Haven-AI, a framework for minimizing efforts in large-scale machine learning experiments


In end-to-end learning systems, defining experiments, creating the training-validation loop, launching the experiments, tracking the progress of the training and visualizing the results are considered as the main components. However, the large amount of efforts needed in modifying the system for such end-to-end machine learning projects is an obstacle in the efficiency. For minimizing the efforts in such end-to-end machine learning projects, the Haven AI framework is developed to help people strongly boost their productivity for building great products, win machine learning competitions, and get research papers published. Compared with other libraries, the Haven-AI framework focuses on end-to-end learning systems, while others require subscription or servers. Haven-AI allows us to build end-to-end projects succinctly and without bells and whistles. It sets the main foundation to minimize the amount of bugs and the amount of time required to add and compare between new datasets and models and to reproduce and generate results. The library also allows developers to integrate other libraries such as ignite and pytorch lightning.

GitHub Link: GitHub - haven-ai/haven-ai

2 Likes

Thanks for sharing Zheyu!

1 Like

Really good job guys, congrats ! Just to append more resources, I found interesting the list of paper implementations you provide:

1 Like

Thank you for sharing the papers implementations, and there are also some benchmarks we implemented for the Haven-AI framework.

2 Likes