typescript: "Do you have runtime libraries installed?", null llmodel

[michalusio]

Problem with: gpt4all backend/bindings

Typescript version: 5.1.1
Package version: [email protected] (happens on 2.1.1-alpha and 2.1.0-alpha too)

Hello,

I have a problem getting the embedding models working using TypeScript bindings for gpt4all.

As in the title, when trying to load the model ggml-all-MiniLM-L6-v2-f16 it throws an error:

Should I point the library path to somewhere? I thought those would be taken from the node_modules.
I am not sure what I am doing wrong, so any help would be appreciated

[iimez]

Tried reproducing using #1390

// node spec/embed.mjs
import { loadModel, createEmbedding, LLModel, DEFAULT_DIRECTORY, DEFAULT_LIBRARIES_DIRECTORY } from '../src/gpt4all.js'
const embedder = await loadModel("all-MiniLM-L6-v2-f16.gguf", { verbose: true, type: 'embedding'})
console.log(createEmbedding(embedder, "Accept your current situation"))
const llm = new LLModel({
  model_name: 'all-MiniLM-L6-v2-f16.gguf',
  model_path: DEFAULT_DIRECTORY,
  library_path: DEFAULT_LIBRARIES_DIRECTORY,
  device: 'cpu'
})
console.log(llm.embed("Embrace the present"))

Both calls behaving as expected.
So this was most likely caused by model format changes and can be closed once the update is merged.

[jacoobes]

stale