i was told i could use these bindings with VICUNA model

[shubham8550]

what i pass to constructior i have quantized
file with me

  const gpt4all = new GPT4All('gpt4all-lora-unfiltered-quantized', true);

[shubham8550]

So i installed VICUNA with this repo

https://github.com/mps256/autovicuna

They put everything in C:/VICUNA
(only 13B Vicuna CPU )

they use
llama.zip

and
ggml-vicuna-13b-4bit.bin

you can get these from

https://huggingface.co/eachadea/ggml-vicuna-13b-4bit/blob/main/ggml-vicuna-13b-4bit-rev1.bin

" ...rev1.bin" have some fixes but its not included in mps256/autovicuna ig

you can get 7B from and it can run on same llama files
https://huggingface.co/eachadea/ggml-vicuna-7b-4bit/tree/main

Heres Some SS running 13B VICUNA

and here 7B (its faster and gives answers better than LLAMA/ alpaca in my pov)

output sometimes end with

maybe because of these
Reverse prompt: '### Human:'
Reverse prompt: '### Instruction:

[shubham8550]

C:\VICUNA\main.exe -i --interactive-first -r "### Human:" -t 8 --temp 0 -c 2048 -n -1 --ignore-eos --repeat_penalty 1.2 --instruct -m C:\VICUNA\ggml-vicuna-7b-4bit-rev1.bin

sample command

[jellydn]

Refer to the source code, it's NOT supported yet.

gpt4all-ts/src/gpt4all.ts

Lines 15 to 33 in 429559f

	constructor(model: string = 'gpt4all-lora-quantized', forceDownload: boolean = false, decoderConfig: Record<string, any> = {}) {
	this.model = model;
	this.decoderConfig = decoderConfig;
	/*
	allowed models:
	M1 Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-m1
	Linux: cd chat;./gpt4all-lora-quantized-linux-x86
	Windows (PowerShell): cd chat;./gpt4all-lora-quantized-win64.exe
	Intel Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-intel
	*/
	if (
	'gpt4all-lora-quantized' !== model &&
	'gpt4all-lora-unfiltered-quantized' !== model
	) {
	throw new Error(`Model ${model} is not supported. Current models supported are:
	gpt4all-lora-quantized
	gpt4all-lora-unfiltered-quantized`
	);
	}

[shubham8550]

Refer to the source code, it's NOT supported yet.

gpt4all-ts/src/gpt4all.ts

Lines 15 to 33 in 429559f

	constructor(model: string = 'gpt4all-lora-quantized', forceDownload: boolean = false, decoderConfig: Record<string, any> = {}) {
	this.model = model;
	this.decoderConfig = decoderConfig;
	/*
	allowed models:
	M1 Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-m1
	Linux: cd chat;./gpt4all-lora-quantized-linux-x86
	Windows (PowerShell): cd chat;./gpt4all-lora-quantized-win64.exe
	Intel Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-intel
	*/
	if (
	'gpt4all-lora-quantized' !== model &&
	'gpt4all-lora-unfiltered-quantized' !== model
	) {
	throw new Error(`Model ${model} is not supported. Current models supported are:
	gpt4all-lora-quantized
	gpt4all-lora-unfiltered-quantized`
	);
	}

yes that's why i made one

https://www.npmjs.com/package/vicuna-ts

[jellydn]

Refer to the source code, it's NOT supported yet.

gpt4all-ts/src/gpt4all.ts

Lines 15 to 33 in 429559f

	constructor(model: string = 'gpt4all-lora-quantized', forceDownload: boolean = false, decoderConfig: Record<string, any> = {}) {
	this.model = model;
	this.decoderConfig = decoderConfig;
	/*
	allowed models:
	M1 Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-m1
	Linux: cd chat;./gpt4all-lora-quantized-linux-x86
	Windows (PowerShell): cd chat;./gpt4all-lora-quantized-win64.exe
	Intel Mac/OSX: cd chat;./gpt4all-lora-quantized-OSX-intel
	*/
	if (
	'gpt4all-lora-quantized' !== model &&
	'gpt4all-lora-unfiltered-quantized' !== model
	) {
	throw new Error(`Model ${model} is not supported. Current models supported are:
	gpt4all-lora-quantized
	gpt4all-lora-unfiltered-quantized`
	);
	}

yes that's why i made one

https://www.npmjs.com/package/vicuna-ts

Nice 👍 Do you think that's good for community if your package is part of gpt4all-ts? Have you considered to send PR to gpt4all-ts? Thanks.

[shubham8550]

originally thats what i wanted but gpt4all-ts s code would have become mess if i stuffed vicuna