TSV.ipynb

untitled0.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "TSV.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPXgApcuZuQ31Fqkc3QFGqN",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/artreven/ef532e2a71f3ff4cb404cfeb541f8657/untitled0.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Fht9KB7dygdi"
      },
      "source": [
        "# Training\n",
        "\n",
        "First clone the repo, install the requirements and train the model. If you are using bash then just delete the first symbols `!` or `%` and put `export` instead of `env` in lines 4 and 6."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "96QABfZeqYU9"
      },
      "source": [
        "%cd /content/\n",
        "!git clone https://github.com/semantic-web-company/wic-tsv\n",
        "%cd wic-tsv/\n",
        "%env PYTHONPATH=/content/wic-tsv/\n",
        "!pip install -r requirements.txt\n",
        "%env HYPERBERT_PATH=HyperBert/eval/\n",
        "!python3 HyperBert/HyperBert3.py --dataset_path ./data --model_output_path $HYPERBERT_PATH --model_name bert-base-uncased"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0Klu8gtly6Wo"
      },
      "source": [
        "# Use the model\n",
        "Now you have the WiC-TSV pretrined HyperBert3. So the next step is to use it."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AQndz-28qrbb"
      },
      "source": [
        "import os\n",
        "from pathlib import Path\n",
        "\n",
        "import torch\n",
        "from transformers import AutoTokenizer\n",
        "\n",
        "from HyperBert.HyperBert3 import HyperBert3\n",
        "from model_evaluation.wictsv_dataset import WiCTSVDataset\n",
        "\n",
        "model_path = Path(os.getenv('HYPERBERT_PATH'))\n",
        "model = HyperBert3.from_pretrained(model_path)\n",
        "model.eval()\n",
        "tok = AutoTokenizer.from_pretrained('bert-base-uncased')\n",
        "\n",
        "cxt1 = \"The jaguar's present range extends from Southwestern United States and Mexico in North America, \" \\\n",
        "       \"across much of Central America, and south to Paraguay and northern Argentina in South America.\"\n",
        "target_word_ind1 = 1\n",
        "cxt2 = \"Jaguar's business was founded as the Swallow Sidecar Company in 1922, originally making motorcycle sidecars \" \\\n",
        "       \"before developing bodies for passenger cars.\"\n",
        "target_word_ind2 = 0\n",
        "def_ = 'wild cat'\n",
        "hypernyms = ['animal']\n",
        "ds = WiCTSVDataset(contexts=[cxt1, cxt2],\n",
        "                   target_inds=[target_word_ind1, target_word_ind2],\n",
        "                   hypernyms=[hypernyms, hypernyms],\n",
        "                   definitions=[def_, def_],\n",
        "                   tokenizer=tok,\n",
        "                   focus_token='$')\n",
        "logits = model(**ds[:])[0]\n",
        "probs = torch.sigmoid(logits).squeeze().tolist()\n",
        "print(f'Probabilities: {probs}')\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0LsMHzTFmbuc"
      },
      "source": [
        "## Output:\n",
        "`Probabilities: [0.9429619312286377, 0.052496373653411865]`"
      ]
    }
  ]
}