introduction_cn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "83973edc",
   "metadata": {},
   "source": [
    "## Overview\n",
    "\n",
    "This is a very small distilled version of the bert-base-multilingual-cased model for Russian and English (45 MB, 12M parameters). There is also an **updated version of this model**, rubert-tiny2, with a larger vocabulary and better quality on practically all Russian NLU tasks.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59944441",
   "metadata": {},
   "source": [
    "This model is useful if you want to fine-tune it for a relatively simple Russian task (e.g. NER or sentiment classification), and you care more about speed and size than about accuracy. It is approximately x10 smaller and faster than a base-sized BERT. Its `[CLS]` embeddings can be used as a sentence representation aligned between Russian and English.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c0e2918f",
   "metadata": {},
   "source": [
    "It was trained on the [Yandex Translate corpus](https://translate.yandex.ru/corpus), [OPUS-100](https://huggingface.co/datasets/opus100) and Tatoeba, using MLM loss distilled from bert-base-multilingual-cased, translation ranking loss, and `[CLS]` embeddings distilled from LaBSE, rubert-base-cased-sentence, Laser and USE.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b0c0158e",
   "metadata": {},
   "source": [
    "There is a more detailed [description in Russian](https://habr.com/ru/post/562064/).\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "28ce4026",
   "metadata": {},
   "source": [
    "Sentence embeddings can be produced as follows:\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d521437a",
   "metadata": {},
   "source": [
    "## How to use"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "da5acdb0",
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install --upgrade paddlenlp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "df2d3cc6",
   "metadata": {},
   "outputs": [],
   "source": [
    "import paddle\n",
    "from paddlenlp.transformers import AutoModel\n",
    "\n",
    "model = AutoModel.from_pretrained(\"cointegrated/rubert-tiny\")\n",
    "input_ids = paddle.randint(100, 200, shape=[1, 20])\n",
    "print(model(input_ids))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "065bda47",
   "metadata": {},
   "source": [
    "> 此模型介绍及权重来源于[https://huggingface.co/cointegrated/rubert-tiny](https://huggingface.co/cointegrated/rubert-tiny)，并转换为飞桨模型格式。\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}