marcus-arcadius / engines

A free and open-source (FOSS) curation of LMs (language models) and APIs, their capabilities and relationships, and logic to select the appropriate engine for a prompt based on that prompt's requirements.

Engines

This is a free and open-source (FOSS) curation of LMs (language models) and APIs, their capabilities and relationships, and logic to select the appropriate engine for a prompt based on that prompt’s requirements.

License	Discord server invite
GPL-3	https://discord.gg/JwKGbAdNHR

See the prompts repo for detailed documentation.

http://github.com/semiosis/prompts

Specifying an engine in a .prompt means that you don’t have to specify the parameters contained by that engine

Typically, the engine specifies the model and the lm-command.

An engine may also load another engine as a template. This is done by supplying the include key with the title of another engine.

Terminology

foundation

This means that a model is generalised and capable of many tasks via prompting.

./glossary.txt

• downstream-engines: Usually means finetuned (extra training).

layers vs specialties vs requirements

Philosophy

• When defining attributes of engines it’s important to not value preciseness over pragmatism. For example, I have specified “foundation, nl, code” for codex as if it had an independent training step for NL on top of the foundation model, which it did not.

Resources

Foundation models

A cautionary note

• https://arxiv.org/abs/2108.07258
• https://crfm.stanford.edu/2021/10/18/reflections.html

Foundation models are neither “foundational”
nor the foundations of AI.

We deliberately chose “foundation” rather than
“foundational”, because we found that
“foundational” implied that these models
provide fundamental principles in a way that
“foundation” does not.

For example, we can readily say “shaky
foundations”, whereas “shaky and foundational”
is unidiomatic.

While many people currently find “foundational
model” more natural to say, “foundation model”
is grammatically a well- formed noun compound
(incidentally parallel to “LM”).

Further, “foundation” describes the (role of)
model and not AI; we neither claim nor believe
that foundation models alone are the
foundation of AI, but instead note they are
“only one component (though an increasingly
important component) of an AI system” (see
Section 1.2; pg. 7-9).