MindsDB VS tiny-cuda-nn

Check us out on GitHub.

If your data lacks uniform time intervals between consecutive entries, QuestDB offers a solution by allowing you to sample your data. After that, MindsDB facilitates creating, training, and deploying your time-series models.

MindsDB is an open-source AI platform for developers that connects AI/ML models with real-time data. It provides tools and automation to easily build and maintain personalized AI solutions.

MindsDB combines both AI and SQL functions in one; users can create, train, optimize, and deploy ML models without the need for external tools. Data analysts can create and visualize forecasts without having to navigate the complexities of ML pipelines.MindsDB is open-source and works with well-known databases like MySQL, Postgres, Redit, Snowflakes, etc.

Mindsdb is a good example. It abstracts everything related to an AI workflow as "virtual tables". For example, you can import OpenAI API as a "virtual table":

Repo : https://github.com/mindsdb/mindsdb

MindsDB is an AI Automation platform for building AI/ML powered features and applications. It works by connecting any data source with any AI/ML model or framework and automating how real-time data flows between them. MindsDB is integrated with LlamaIndex, which makes use of its data framework for connecting custom data sources to large language models. LlamaIndex data ingestion allows you to connect to data sources like PDF’s, webpages, etc., provides data indexing and a query interface that takes input prompts from your data and provides knowledge-augmented responses, thus making it easy to Q&A over documents and webpages.

Now, imagine if you can deploy these highly trained models in your database to get insights, make predictions, understand your users, auto-generate content, and more. MindsDB makes this possible! MindsDB is an open-source AI database middleware that allows you to supercharge your databases by integrating various machine learning (ML) engines.

In the opposite direction from your question is a very interesting project, TinyNN all implemented as close to the metal as possible and very fast: https://github.com/NVlabs/tiny-cuda-nn

This repo only implemented the rendering part of the NGP but is more simple and has a lesser amount of code compared to the original (Instant-NGP and tiny-cuda-nn).

...a C++ library with a CUDA backend. But these high-performance building blocks might only be saturating the GPU fully if the data is large enough.

I haven't looked at implementing these things, but I imagine uf you have smaller networks and thus less data, the large building blocks may not be optimal. You may for example want to fuse some operations to reduce memory latency from repeated memory access.

In PyTorch world, there are approaches for small networks as well, there is https://github.com/NVlabs/tiny-cuda-nn - as far as I understand from the first link in the README, it makes clever use of the CUDA shared memory, which can hold all the weights of a tiny network (but not larger ones).