Proposal
Incorporate the ability to load GPT2 models (and others in future) from the GGUF format as defined here:
https://github.com/ggerganov/ggml/blob/master/docs/gguf.md
There is a fairly approachable library written in TypeScript that could be a good basis:
https://github.com/ahoylabs/gguf.js
Benefits
- Allows loading of models from a single file.
- Easy extensibility for new model support such as Llama2 models in the future.
Proposal
Incorporate the ability to load GPT2 models (and others in future) from the GGUF format as defined here:
https://github.com/ggerganov/ggml/blob/master/docs/gguf.md
There is a fairly approachable library written in TypeScript that could be a good basis:
https://github.com/ahoylabs/gguf.js
Benefits