prismadic / magnet Goto Github PK
View Code? Open in Web Editor NEWthe small distributed language model toolkit; fine-tune state-of-the-art LLMs anywhere, rapidly
Home Page: https://prismadic.github.io/magnet/
License: MIT License
the small distributed language model toolkit; fine-tune state-of-the-art LLMs anywhere, rapidly
Home Page: https://prismadic.github.io/magnet/
License: MIT License
title
NATS has integrated a lot more connection parameters and likely some in other classes besides the client.
Should implement all of them.
ggml or something, there are great libraries we may depend on if they are small enough
All data entries going into NATS should be hashed.
title
spit stream volume equal lengths, create new NATS stream for jobs with pull_subscribe consumer model using index as job criteria (and other criteria)
title
NATS has improved the object meta data for the consumer/producers of data, should absolutely be used for basic bandwidth transparency
can query on the fly on the node with the magnet.ron
llm but cannot query remotely to a distributed llm node
put this in as _f argument and add levels for info warn fatal.
should have a default writable log path for unix machines minimum with timestamps prior to what otherwise would have just been printed to stdout
it's just the data, it should be meta like origin filename, etc. (need to use JSON from NATS documentation: they require a schema)
basically there should be built in callbacks for the NATS magnet.ic.field.Resonator
other than print() and pymongo's insert_document
could be first
create mega class hierarchy
need to add params for instruction prepend, whatever else I find.
Make it possible to do accelerated embedding on Macs.
Seems n->β duplicates are observed after indexing and using retrieval later.
Since these are processed directly from NATS using a durable consumer session, there must be an issue with the distributed messaging protocol or perhaps the inserts.
Another possibility is duplicates are created as a part of the processing part of the workflow but since the number of them seems to increase without relation to the order in the stream, this seems unlikely.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
π Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. πππ
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google β€οΈ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.