{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5ROgtvTfTdy9"
      },
      "source": [
        "## FlexTab Examples\n",
        "Model page: https://huggingface.co/SAP/FlexTab"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N-54dB6iTdy_"
      },
      "source": [
        "The model you are trying to use is gated. Please make sure you have access to it by visiting the model page.To run inference, either set HF_TOKEN in your environment variables/ Secrets or run the following cell to login. 🤗"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "o7iIwHVHTdzA"
      },
      "outputs": [],
      "source": [
        "from huggingface_hub import login\n",
        "login(new_session=False)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "# Uninstall conflicting packages from the default Colab environment\n",
        "!pip uninstall torch torchvision torchaudio -y`"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "6GqTz9gYTdzB"
      },
      "outputs": [],
      "source": [
        "# Install FlexTab core code\n",
        "!pip install git+https://github.com/SAP-samples/flextab"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "jzA1dexsTdzB"
      },
      "outputs": [],
      "source": [
        "from sklearn.datasets import load_breast_cancer\n",
        "from sklearn.metrics import accuracy_score\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from flextab import FlexTabClassifier\n",
        "\n",
        "# Load sample data\n",
        "X, y = load_breast_cancer(return_X_y=True)\n",
        "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, random_state=42)\n",
        "\n",
        "# Initialize a classifier, 8k context and 8-fold bagging gives best performance, reduce if running out of memory\n",
        "clf = FlexTabClassifier(max_context_size=8192, bagging=8)\n",
        "\n",
        "clf.fit(X_train, y_train)\n",
        "\n",
        "# Predict probabilities\n",
        "prediction_probabilities = clf.predict_proba(X_test)\n",
        "# Predict labels\n",
        "predictions = clf.predict(X_test)\n",
        "print(\"Accuracy\", accuracy_score(y_test, predictions))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ZlG2jXfWTdzC"
      },
      "outputs": [],
      "source": [
        "from sklearn.datasets import fetch_openml\n",
        "from sklearn.metrics import r2_score\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from flextab import FlexTabRegressor\n",
        "\n",
        "# Load sample data\n",
        "df = fetch_openml(data_id=531, as_frame=True)\n",
        "X = df.data\n",
        "y = df.target.astype(float)\n",
        "\n",
        "# Train-test split\n",
        "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, random_state=42)\n",
        "\n",
        "# Initialize the regressor, 8k context and 8-fold bagging gives best performance, reduce if running out of memory\n",
        "regressor = FlexTabRegressor(max_context_size=8192, bagging=8)\n",
        "\n",
        "regressor.fit(X_train, y_train)\n",
        "\n",
        "# Predict on the test set\n",
        "predictions = regressor.predict(X_test)\n",
        "\n",
        "r2 = r2_score(y_test, predictions)\n",
        "print(\"R² Score:\", r2)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "from sklearn.metrics import accuracy_score, roc_auc_score\n",
        "\n",
        "from flextab import FlexTabMatcher\n",
        "from flextab.utils.test_utils import load_febrl4\n",
        "\n",
        "left_train, left_test, right_train, right_test, y_train, y_test = load_febrl4()\n",
        "\n",
        "matcher = FlexTabMatcher(max_context_size=8192, bagging=1)\n",
        "matcher.fit(left_train, right_train, y_train)\n",
        "predictions = matcher.predict(left_test, right_test)\n",
        "print(f'Accuracy {accuracy_score(y_test, predictions):.2%}')\n",
        "# Probabilities (e.g. for thresholding or AUROC):\n",
        "# probas = matcher.predict_proba_matching(left_test, right_test)\n",
        "# print(f'AUROC {roc_auc_score(y_test, probas[:, 1]):.2%}')"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "gpuType": "T4",
      "machine_shape": "hm",
      "provenance": []
    },
    "kaggle": {
      "accelerator": "gpu"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}