onnx-llama2_xs_460M_experimental_evol_instruct

INT8 ONNX version of Felladrin/llama2_xs_460M_experimental_evol_instruct to use with Transformers.js.

Example usage

Pipeline API

import { pipeline } from '@xenova/transformers';

const generator = await pipeline('text-generation', 'Felladrin/onnx-int8-llama2_xs_460M_experimental_evol_instruct');
const output =  await generator('Once upon a time,', { add_special_tokens: true, max_new_tokens: 60, repetition_penalty: 1.2});
console.log(output);

Auto Classes

import { AutoModelForCausalLM, AutoTokenizer } from '@xenova/transformers';

const model_path = 'Felladrin/onnx-int8-llama2_xs_460M_experimental_evol_instruct';
const model = await AutoModelForCausalLM.from_pretrained(model_path);
const tokenizer = await AutoTokenizer.from_pretrained(model_path);

const prompt = 'Once upon a time,';
const { input_ids } = tokenizer(prompt);
const tokens = await model.generate(input_ids, { max_new_tokens: 60, repetition_penalty: 1.2});
console.log(tokenizer.decode(tokens[0], { skip_special_tokens: true }));