original_logic.ipynb

original_logic.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "display_name": "Hugging",
      "language": "python",
      "name": "huggingface"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.8.8"
    },
    "colab": {
      "name": "original_logic.ipynb",
      "provenance": [],
      "include_colab_link": true
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/Proteusiq/4a9d93f29c7042b757f6e2a502ce6322/original_logic.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dRVfaljk7CIH"
      },
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "from transformers import AutoTokenizer\n",
        "from transformers import AutoModelForSequenceClassification\n",
        "\n",
        "\n",
        "MODEL_NAME = \"microsoft/deberta-xlarge-mnli\"\n",
        "\n",
        "\n",
        "class Model:\n",
        "    _shared_model = {\n",
        "        \"softmax\": nn.Softmax(dim=1),\n",
        "        \"tokenizer\": AutoTokenizer.from_pretrained(MODEL_NAME),\n",
        "        \"model\": AutoModelForSequenceClassification.from_pretrained(MODEL_NAME),\n",
        "        'sentences': [],\n",
        "    }\n",
        "\n",
        "    def __init__(self):\n",
        "        self.__dict__ = self._shared_model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i0HKWkfc7CIM"
      },
      "source": [
        "class Logic(Model):\n",
        "    \"\"\"Sentence Logic\n",
        "    Compare two sentence to see if they entail or contradict each other\n",
        "\n",
        "    Usage:\n",
        "\n",
        "    >>> driving = Logic(\"\"I am driving a car.\") # add a sentence\n",
        "    >>> in_car = Logic(\"I am in not a car.\"\") # add another sentence\n",
        "    >>> driving.entails(in_car) # check for entailment\n",
        "    >>> driving > in_car # check for entailment\n",
        "    >>> driving != in_car # check for contradition\n",
        "    >>> driving.contradicts(in_car, verbose=True) # check for contraidiction with extra info\n",
        "    \"\"\"\n",
        "\n",
        "    def __init__(self, sentence, tokenizer=None, model=None):\n",
        "        super().__init__()\n",
        "        self.sentences.append(sentence)\n",
        "        self.sentences = self.sentences[-2:] # keeps only two sentences\n",
        "        if model:\n",
        "            self.tokenizer = tokenizer\n",
        "            self.model = model\n",
        "        else:\n",
        "            # initiate the first instance with default model\n",
        "            if not hasattr(self, \"model\"):\n",
        "                raise RuntimeError(\"no model to  perform operations!\")\n",
        "\n",
        "    def __repr__(self):\n",
        "        return f\"{self.__class__.__name__}(model={self.model_name})\"\n",
        "    \n",
        "          \n",
        "        \n",
        "    def entails(self, other, threashold=0.7, verbose=False):\n",
        "        sentence, other = self.sentences # overiding other\n",
        "        scores = self._predict(sentence, other)\n",
        "        return self._post_predict(scores, \"entails\", threashold, verbose)\n",
        "\n",
        "    def contradicts(self, other, threashold=0.7, verbose=False):\n",
        "        sentence, other = self.sentences # overiding other\n",
        "        scores = self._predict(sentence, other)\n",
        "        return self._post_predict(scores, \"contradicts\", threashold, verbose)\n",
        "\n",
        "    def _predict(self, sentence, other_sentence):\n",
        "\n",
        "        inputs = self.tokenizer(\n",
        "            [f\"{sentence} [SEP] {other_sentence}\"], return_tensors=\"pt\", is_split_into_words = True\n",
        "        )\n",
        "        labels = torch.tensor([1] * 3).unsqueeze(0) \n",
        "        outputs = self.model(**inputs, labels=labels)\n",
        "        predictions = self.softmax(outputs.logits)\n",
        "        return {\n",
        "            label: score\n",
        "            for label, score in zip(\n",
        "                [\"contradicts\", \"neutral\", \"entails\"], predictions.detach().numpy()[0]\n",
        "            )\n",
        "        }\n",
        "\n",
        "    @staticmethod\n",
        "    def _post_predict(scores, task, threashold, verbose):\n",
        "\n",
        "        results = scores[task] > threashold\n",
        "        if verbose:\n",
        "            return {\"results\": results, \"explanations\": scores}\n",
        "        return results\n",
        "    \n",
        "    # some other methods we can use\n",
        "    __ne__ = contradicts\n",
        "    __lt__ = entails\n",
        "    __gt__ = entails\n",
        "    __le__ = entails\n",
        "    \n",
        "    \n",
        "    "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AWkpvfxK7CIO",
        "outputId": "79519910-a00d-4574-c24f-8cddcafe2801"
      },
      "source": [
        "in_car = Logic(\"I am driving a Ford C-Max.\")\n",
        "driving = Logic(\"I am in a car.\")\n",
        "\n",
        "in_car.entails(driving)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mP8lkzm_7CIO",
        "outputId": "7c69c679-a323-40e9-8360-952b3c76facf"
      },
      "source": [
        "in_car > driving # entails > and <= means the same"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hNRoAc0F7CIP",
        "outputId": "a81aa8f5-fe34-4171-a9f6-44aca9953d59"
      },
      "source": [
        "in_car.contradicts(driving, verbose=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'results': False,\n",
              " 'explanations': {'contradicts': 0.0013153073,\n",
              "  'neutral': 0.02768614,\n",
              "  'entails': 0.9709986}}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aB6Zie1E7CIP",
        "outputId": "7b262a3d-9b9b-4747-e990-e2a4dd1e3aa2"
      },
      "source": [
        "black_cat = Logic(\"That cat is black.\")\n",
        "white_cat = Logic(\"The cat's color is white.\")\n",
        "\n",
        "black_cat.entails(white_cat, verbose=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'results': False,\n",
              " 'explanations': {'contradicts': 0.99652416,\n",
              "  'neutral': 0.001198517,\n",
              "  'entails': 0.0022773156}}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9Zvl7Qgr7CIQ",
        "outputId": "db7ca20a-6391-4a4b-b5ba-70af0c07bb1b"
      },
      "source": [
        "black_cat != white_cat"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AbqgHvrW7CIQ"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}