pycaret_ts_sktime_darts.ipynb

pycaret_ts_sktime_darts.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "pycaret_ts_sktime_darts.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/ngupta23/d876b55fdfc47e5573b209d3aa5be222/pycaret_ts_sktime_darts.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GWDARV7nIEJ9",
        "outputId": "a7e3f6e0-2054-48d8-fb64-81fda1bb09a6"
      },
      "source": [
        "try:\n",
        "  import darts\n",
        "except:\n",
        "  !pip install darts"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/distributed/config.py:20: YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.\n",
            "  defaults = yaml.load(f)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iiHG42TGJf6K"
      },
      "source": [
        "try:\n",
        "  import pycaret\n",
        "except:\n",
        "  !pip install pycaret-ts-alpha"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SKPONey9IF3b"
      },
      "source": [
        "import sys\n",
        "import time\n",
        "import os\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "from numpy.random import RandomState\n",
        "import matplotlib.pyplot as plt\n",
        "from datetime import datetime\n",
        "from functools import reduce\n",
        "from typing import Union, List, Optional, Dict\n",
        "\n",
        "from darts import TimeSeries\n",
        "from darts.models import NaiveSeasonal\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "import logging\n",
        "logging.disable(logging.CRITICAL)"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WKAYfuvAUuqp"
      },
      "source": [
        "## DARTS Adapter"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b28MXBD0JAH4"
      },
      "source": [
        "from sktime.forecasting.base import BaseForecaster\n",
        "\n",
        "class _DartsAdapter(BaseForecaster):\n",
        "    \"\"\"Base class for interfacing statsmodels forecasting algorithms.\"\"\"\n",
        "    # https://github.com/alan-turing-institute/sktime/blob/v0.8.0/extension_templates/forecasting.py\n",
        "    _tags = {\n",
        "        \"scitype:y\": \"univariate\",  # which y are fine? univariate/multivariate/both\n",
        "        \"univariate-only\": True,  # does estimator use the exogeneous X?\n",
        "        \"handles-missing-data\": False,  # can estimator handle missing data?\n",
        "        \"y_inner_mtype\": \"pd.DataFrame\",  # which types do _fit, _predict, assume for y?\n",
        "        \"X_inner_mtype\": \"pd.DataFrame\",  # which types do _fit, _predict, assume for X?\n",
        "        \"requires-fh-in-fit\": False,  # is forecasting horizon already required in fit?\n",
        "        \"X-y-must-have-same-index\": True,  # can estimator handle different X/y index?\n",
        "        \"enforce-index-type\": None,  # index type that needs to be enforced in X/y\n",
        "        \"capability:pred_int\": False,\n",
        "    }\n",
        "\n",
        "    _fitted_param_names = ()\n",
        "    \n",
        "    def __init__(self):\n",
        "        self._forecaster = None\n",
        "        self._fitted_forecaster = None\n",
        "        super(_DartsAdapter, self).__init__()\n",
        "       \n",
        "\n",
        "    def _fit(self, y, X=None, fh=None):\n",
        "        \"\"\"Fit to training data.\n",
        "        Parameters\n",
        "        ----------\n",
        "        y : pd.Series\n",
        "            Target time series to which to fit the forecaster.\n",
        "        fh : int, list or np.array, optional (default=None)\n",
        "            The forecasters horizon with the steps ahead to to predict.\n",
        "        X : pd.DataFrame, optional (default=None)\n",
        "            Exogenous variables are ignored\n",
        "        Returns\n",
        "        -------\n",
        "        self : returns an instance of self.\n",
        "        \"\"\"\n",
        "        self._forecaster = self._instantiate_model()\n",
        "\n",
        "        from darts import TimeSeries\n",
        "        # Darts needs a datetime index (originally passed as PeriodIndex)\n",
        "        y.index = y.index.astype('datetime64[ns]') \n",
        "        y = pd.DataFrame(y)\n",
        "        y_ts = TimeSeries.from_dataframe(y)\n",
        "        self._forecaster.fit(y_ts)\n",
        "        \n",
        "        # this should happen last\n",
        "        self._is_fitted = True\n",
        "\n",
        "        return self\n",
        "\n",
        "    def _predict(self, fh=None, X=None, return_pred_int=False, alpha=0.05):\n",
        "        self.check_is_fitted()\n",
        "        \n",
        "        # Temporary hack assuming continuous values without any gap.\n",
        "        # Will need more work\n",
        "        h = len(list(fh))\n",
        "        # print(h)\n",
        "        y = self._forecaster.predict(h).pd_dataframe()\n",
        "        return y"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wiO42sPqLCry"
      },
      "source": [
        "class Naive(_DartsAdapter):\n",
        "    def __init__(\n",
        "        self,\n",
        "        K=1,\n",
        "    ):\n",
        "      self.K = K\n",
        "      \n",
        "      super(Naive, self).__init__()\n",
        "\n",
        "    def _instantiate_model(self):\n",
        "        # import inside method to avoid hard dependency\n",
        "        from darts.models import NaiveSeasonal as _NaiveSeasonal\n",
        "        return _NaiveSeasonal(K = self.K)\n"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fufH9Lca8oTy"
      },
      "source": [
        "from darts.models import NBEATSModel\n",
        "NBEATSModel?"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IfY7s7Xxz-Fo"
      },
      "source": [
        "# Get Parameters from here:\n",
        "# https://github.com/unit8co/darts/blob/dec5f8c56af92254fdb1d1079f3e9d95b8c91f2b/darts/models/forecasting/nbeats.py#L411\n",
        "\n",
        "class NBEATSModel(_DartsAdapter):\n",
        "    import torch\n",
        "    import torch.nn as nn\n",
        "    from darts.utils.likelihood_models import Likelihood\n",
        "\n",
        "    def __init__(\n",
        "        self,\n",
        "        input_chunk_length: int,\n",
        "        output_chunk_length: int,\n",
        "        generic_architecture: bool = True,\n",
        "        num_stacks: int = 30,\n",
        "        num_blocks: int = 1,\n",
        "        num_layers: int = 4,\n",
        "        layer_widths: Union[int, List[int]] = 256,\n",
        "\n",
        "        # Add to model ----\n",
        "        expansion_coefficient_dim: int = 5,\n",
        "        trend_polynomial_degree: int = 2,\n",
        "        likelihood: Optional[Likelihood] = None,\n",
        "        random_state: Optional[Union[int, RandomState]] = None,\n",
        "\n",
        "        batch_size: int = 32,\n",
        "        n_epochs: int = 100,\n",
        "        optimizer_cls: torch.optim.Optimizer = torch.optim.Adam,\n",
        "        optimizer_kwargs: Optional[Dict] = None,\n",
        "        lr_scheduler_cls: torch.optim.lr_scheduler._LRScheduler = None,\n",
        "        lr_scheduler_kwargs: Optional[Dict] = None,\n",
        "        loss_fn: nn.modules.loss._Loss = nn.MSELoss(),\n",
        "        model_name: str = None,\n",
        "        work_dir: str = os.getcwd(),\n",
        "        log_tensorboard: bool = False,\n",
        "        nr_epochs_val_period: int = 10,\n",
        "        torch_device_str: Optional[str] = None,\n",
        "        force_reset=False,\n",
        "        save_checkpoints=False\n",
        "    ):\n",
        "      self.input_chunk_length=input_chunk_length\n",
        "      self.output_chunk_length=output_chunk_length\n",
        "      self.generic_architecture=generic_architecture\n",
        "      self.num_stacks=num_stacks\n",
        "      self.num_blocks=num_blocks\n",
        "      self.num_layers=num_layers\n",
        "      self.layer_widths=layer_widths\n",
        "\n",
        "      self.expansion_coefficient_dim = expansion_coefficient_dim\n",
        "      self.trend_polynomial_degree = trend_polynomial_degree\n",
        "      self.likelihood = likelihood\n",
        "      self.random_state = random_state\n",
        "\n",
        "      self.batch_size=batch_size\n",
        "      self.n_epochs=n_epochs\n",
        "      self.optimizer_cls=optimizer_cls\n",
        "      self.optimizer_kwargs=optimizer_kwargs\n",
        "      self.lr_scheduler_cls=lr_scheduler_cls\n",
        "      self.lr_scheduler_kwargs=lr_scheduler_kwargs\n",
        "      self.loss_fn=loss_fn\n",
        "      self.model_name=model_name\n",
        "      self.work_dir=work_dir\n",
        "      self.log_tensorboard=log_tensorboard\n",
        "      self.nr_epochs_val_period=nr_epochs_val_period\n",
        "      self.torch_device_str=torch_device_str\n",
        "      self.force_reset=force_reset\n",
        "      self.save_checkpoints=save_checkpoints\n",
        "      \n",
        "      super(NBEATSModel, self).__init__()\n",
        "\n",
        "    def _instantiate_model(self):\n",
        "        # import inside method to avoid hard dependency\n",
        "        from darts.models import NBEATSModel as _NBEATSModel\n",
        "        \n",
        "        return _NBEATSModel(\n",
        "            input_chunk_length=self.input_chunk_length,\n",
        "            output_chunk_length=self.output_chunk_length,\n",
        "            generic_architecture=self.generic_architecture,\n",
        "            num_stacks=self.num_stacks,\n",
        "            num_blocks=self.num_blocks,\n",
        "            num_layers=self.num_layers,\n",
        "            layer_widths=self.layer_widths,\n",
        "            n_epochs=self.n_epochs,\n",
        "            nr_epochs_val_period=self.nr_epochs_val_period,\n",
        "            batch_size=self.batch_size,\n",
        "            model_name=self.model_name\n",
        "        )\n",
        "\n"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "L15ogZhnUqkI"
      },
      "source": [
        "## Get Dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 384
        },
        "id": "upb8xabYLvwr",
        "outputId": "794c07f3-bbe0-4000-9cf1-4836dc079735"
      },
      "source": [
        "from pycaret.datasets import get_data\n",
        "y = get_data(\"airline\")\n",
        "train = y[:-36]\n",
        "test = y[-36:]\n",
        "train.index, test.index"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Period\n",
              "1949-01    112.0\n",
              "1949-02    118.0\n",
              "1949-03    132.0\n",
              "1949-04    129.0\n",
              "1949-05    121.0\n",
              "Freq: M, Name: Number of airline passengers, dtype: float64"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(PeriodIndex(['1949-01', '1949-02', '1949-03', '1949-04', '1949-05', '1949-06',\n",
              "              '1949-07', '1949-08', '1949-09', '1949-10',\n",
              "              ...\n",
              "              '1957-03', '1957-04', '1957-05', '1957-06', '1957-07', '1957-08',\n",
              "              '1957-09', '1957-10', '1957-11', '1957-12'],\n",
              "             dtype='period[M]', name='Period', length=108),\n",
              " PeriodIndex(['1958-01', '1958-02', '1958-03', '1958-04', '1958-05', '1958-06',\n",
              "              '1958-07', '1958-08', '1958-09', '1958-10', '1958-11', '1958-12',\n",
              "              '1959-01', '1959-02', '1959-03', '1959-04', '1959-05', '1959-06',\n",
              "              '1959-07', '1959-08', '1959-09', '1959-10', '1959-11', '1959-12',\n",
              "              '1960-01', '1960-02', '1960-03', '1960-04', '1960-05', '1960-06',\n",
              "              '1960-07', '1960-08', '1960-09', '1960-10', '1960-11', '1960-12'],\n",
              "             dtype='period[M]', name='Period'))"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1zJtsxLnUzM4"
      },
      "source": [
        "## Naive Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZIKQdO-oLxWj",
        "outputId": "6b18aff7-6322-4ccd-f0fe-e0b428ee30a4"
      },
      "source": [
        "model1 = Naive()\n",
        "model1.fit(train)\n",
        "\n",
        "# #### Preiction works with underlying DARTS model ----\n",
        "# model1._forecaster.predict(36).values().T\n",
        "\n",
        "#### Prediction works with sktime adapter now ----\n",
        "predictions = model1.predict(fh=np.arange(1, 37))\n",
        "type(predictions), predictions.shape"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(pandas.core.series.Series, (36,))"
            ]
          },
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ziVKlX2N0wCZ"
      },
      "source": [
        "###NBests Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8C2z9PX5xhIg",
        "outputId": "5512f858-cc78-421d-8f92-4b906484bcc7"
      },
      "source": [
        "model2a = NBEATSModel(\n",
        "    input_chunk_length=30,\n",
        "    output_chunk_length=7,\n",
        "    generic_architecture=True,\n",
        "    num_stacks=10,\n",
        "    num_blocks=1,\n",
        "    num_layers=4,\n",
        "    layer_widths=512,\n",
        "    n_epochs=100,\n",
        "    nr_epochs_val_period=1,\n",
        "    batch_size=800,\n",
        "    model_name='nbeats_run'\n",
        ")\n",
        "model2a.fit(train)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "NBEATSModel(batch_size=800, input_chunk_length=30, layer_widths=512,\n",
              "            model_name='nbeats_run', nr_epochs_val_period=1, num_stacks=10,\n",
              "            output_chunk_length=7)"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RT5uRc85KIpY",
        "outputId": "cbe22763-f6aa-42ff-f0f2-174d71d7e9c8"
      },
      "source": [
        "predictions = model2a.predict(fh=np.arange(1, 37))\n",
        "type(predictions), predictions.shape"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(pandas.core.series.Series, (36,))"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jEdQkXqLKsHI",
        "outputId": "f1faa607-47f9-4d1b-edcc-0da9824432a0"
      },
      "source": [
        "model2b = NBEATSModel(\n",
        "    input_chunk_length=30,\n",
        "    output_chunk_length=7,\n",
        "    generic_architecture=False,\n",
        "    num_blocks=3,\n",
        "    num_layers=4,\n",
        "    layer_widths=512,\n",
        "    n_epochs=100,\n",
        "    nr_epochs_val_period=1,\n",
        "    batch_size=800,\n",
        "    model_name='nbeats_interpretable_run'\n",
        ")\n",
        "model2b.fit(train)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "NBEATSModel(batch_size=800, generic_architecture=False, input_chunk_length=30,\n",
              "            layer_widths=512, model_name='nbeats_interpretable_run',\n",
              "            nr_epochs_val_period=1, num_blocks=3, output_chunk_length=7)"
            ]
          },
          "metadata": {},
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bpaD93CcQcMG",
        "outputId": "c6e14acd-64c1-4300-9a9d-64bdb7d0a9fd",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "predictions = model2b.predict(fh=np.arange(1, 37))\n",
        "type(predictions), predictions.shape"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(pandas.core.series.Series, (36,))"
            ]
          },
          "metadata": {},
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vOjJ9dTUUQoI"
      },
      "source": [
        "## Backup"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LSgxwiE-QREc",
        "outputId": "0ac99de1-57ac-42e0-8c5a-8e9bb634e487",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 117
        }
      },
      "source": [
        "sys.exit()"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "error",
          "ename": "SystemExit",
          "evalue": "ignored",
          "traceback": [
            "An exception has occurred, use %tb to see the full traceback.\n",
            "\u001b[0;31mSystemExit\u001b[0m\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zwH88xAX66r-"
      },
      "source": [
        "from darts import TimeSeries\n",
        "from darts.models import RNNModel\n",
        "from darts.models import NBEATSModel"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uo0vEd9BURfc"
      },
      "source": [
        "from darts.metrics import mape, mase\n",
        "from darts.utils.statistics import check_seasonality, plot_acf, plot_residuals_analysis, plot_hist\n",
        "from darts.datasets import AirPassengersDataset"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wq5dULjXPnt-"
      },
      "source": [
        "series = AirPassengersDataset().load()\n",
        "series.plot()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vqRcfIEwUVXr"
      },
      "source": [
        "train, val = series.split_before(pd.Timestamp('19580101'))\n",
        "train.plot(label='training')\n",
        "val.plot(label='validation')\n",
        "plt.legend()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Px8t6wA0UYFc"
      },
      "source": [
        "naive_model = NaiveSeasonal(K=1)\n",
        "naive_model.fit(train)\n",
        "naive_forecast = naive_model.predict(36)\n",
        "\n",
        "series.plot(label='actual')\n",
        "naive_forecast.plot(label='naive forecast (K=1)')\n",
        "plt.legend();"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kzwtw4mP29Or"
      },
      "source": [
        "NBEATSModel?"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1QfYgjYj25PA"
      },
      "source": [
        "from darts.models import NBEATSModel\n",
        "model_nbeats = NBEATSModel(\n",
        "    input_chunk_length=30,\n",
        "    output_chunk_length=7,\n",
        "    generic_architecture=True,\n",
        "    # num_stacks=10,\n",
        "    num_blocks=1,\n",
        "    num_layers=4,\n",
        "    layer_widths=512,\n",
        "    n_epochs=100,\n",
        "    nr_epochs_val_period=1,\n",
        "    batch_size=800,\n",
        "    model_name='nbeats_run'\n",
        ")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aFdN4Uqf252S"
      },
      "source": [
        "model_nbeats.fit(train, verbose=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lhWg3S8B7W42"
      },
      "source": [
        "model_nbeats.predict(n=36)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lpywP8GBLZRc"
      },
      "source": [
        "model_nbeats = NBEATSModel(\n",
        "    input_chunk_length=30,\n",
        "    output_chunk_length=7,\n",
        "    generic_architecture=False,\n",
        "    num_blocks=3,\n",
        "    num_layers=4,\n",
        "    layer_widths=512,\n",
        "    n_epochs=100,\n",
        "    nr_epochs_val_period=1,\n",
        "    batch_size=800,\n",
        "    model_name='nbeats_interpretable_run'\n",
        ")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pC8At8mKLhTg"
      },
      "source": [
        "model_nbeats.fit(train, verbose=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7WBlkxHgUaqs"
      },
      "source": [
        "class RNNModel(_DartsAdapter):\n",
        "    def __init__(\n",
        "        self,\n",
        "        model='LSTM',\n",
        "        hidden_dim=20,\n",
        "        dropout=0,\n",
        "        batch_size=16,\n",
        "        n_epochs=300,\n",
        "        optimizer_kwargs={'lr': 1e-3}, \n",
        "        model_name='Air_RNN',\n",
        "        log_tensorboard=True,\n",
        "        random_state=42,\n",
        "        training_length=20,\n",
        "        input_chunk_length=14,\n",
        "        force_reset=True,\n",
        "        save_checkpoints=True\n",
        "    ):\n",
        "      self.model=model,\n",
        "      self.hidden_dim=hidden_dim,\n",
        "      self.dropout=dropout,\n",
        "      self.batch_size=batch_size,\n",
        "      self.n_epochs=n_epochs,\n",
        "      self.optimizer_kwargs=optimizer_kwargs, \n",
        "      self.model_name=model_name,\n",
        "      self.log_tensorboard=log_tensorboard,\n",
        "      self.random_state=random_state,\n",
        "      self.training_length=training_length,\n",
        "      self.input_chunk_length=input_chunk_length,\n",
        "      self.force_reset=force_reset,\n",
        "      self.save_checkpoints=save_checkpoints\n",
        "      \n",
        "      super(RNNModel, self).__init__()\n",
        "\n",
        "    def _instantiate_model(self):\n",
        "        # import inside method to avoid hard dependency\n",
        "        from darts.models import RNNModel as _RNNModel\n",
        "        return _RNNModel(\n",
        "            model=self.model,\n",
        "            hidden_dim=self.hidden_dim,\n",
        "            dropout=self.dropout,\n",
        "            batch_size=self.batch_size,\n",
        "            n_epochs=self.n_epochs,\n",
        "            optimizer_kwargs=self.optimizer_kwargs, \n",
        "            model_name=self.model_name,\n",
        "            log_tensorboard=self.log_tensorboard,\n",
        "            random_state=self.random_state,\n",
        "            training_length=self.training_length,\n",
        "            input_chunk_length=self.input_chunk_length,\n",
        "            force_reset=self.force_reset,\n",
        "            save_checkpoints=self.save_checkpoints\n",
        "          )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Br19Z6Vg-Ito"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}