bacalhau VS onnxruntime

This is really interesting - we’ve tried really hard to solve some of these with Bacalhau[1] - a much simpler distributed compute platform. Would love your feedback!

[1] https://github.com/bacalhau-project/bacalhau

Disclosure: I confounded Bacalhau

It's an interesting question!

The problem is that the data is spread everywhere - no choice about that. So with that in mind, how do you query that data? Today, the idea is that you HAVE to put it into a central location. With tools like Bacalhau[1] and DuckDB [2], you no longer have to - a single query can be sharded amongst all your data - EFFECTIVELY giving you a lot of what you want from a data lake.

It's not a replacement, but if you can do a few of these items WITHOUT moving the data, you will be able to see really significant cost and time savings.

[1] https://github.com/bacalhau-project/bacalhau

[2] https://github.com/duckdb/duckdb

Very selfishly - distributed compute. Not decentralized, distributed.

Compute and data are being created and run everywhere, we need platforms that understand how to use it and get insights without (or before) moving it.

Our contribution: https://github.com/bacalhau-project/bacalhau (think Kubernetes but built for the distributed world).

Disclosure: I co-founded the Bacalhau Project

Thank you so much for the feedback. Yeah, we REALLY do want to figure out a better naming/reference scheme. Do you have anything you've seen you really like?

Disclosure: I work on Bacalhau[1]

https://github.com/bacalhau-project/bacalhau

ONNX Runtime: ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator

Embedding generation uses the ONNX runtime under the hood. This is a cross-platform inferencing library that supports multiple execution providers from CPU to specialized GPUs.

tfjs is dead, looking at the commit history. The standard now is to convert PyTorch to onnx, then use onnxruntime (https://github.com/microsoft/onnxruntime/tree/main/js/web) to run the model on the browsdr.

- slang[2] has the potential, but the meta programming part is not as strong as C++, existing libraries cannot be used.

The above conclusion is drawn from my work https://github.com/microsoft/onnxruntime/tree/dev/opencl, purely nightmare to work with thoes drivers and jit compilers. Hopefully Vcc can take compute shader more seriously.

[1]: https://www.circle-lang.org/

I did. It depends what you want, for an overview of how ONNX Runtime works then Microsoft have a bunch of things on https://onnxruntime.ai, but the Java content is a bit lacking on there as I've not had time to write much. Eventually I'll probably write something similar to the C# SD tutorial they have on there but for the Java API.

For writing ONNX models from Java we added an ONNX export system to Tribuo in 2022 which can be used by anything on the JVM to export ONNX models in an easier way than writing a protobuf directly. Tribuo doesn't have full coverage of the ONNX spec, but we're happy to accept PRs to expand it, otherwise it'll fill out as we need it.

What about models besides GPT? Most of the popular vector encoding models aren't using this architecture.

If you really didn't want PyTorch/Transformers, you could consider exporting your models to ONNX (https://github.com/microsoft/onnxruntime).