{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide\n",
    "!pip install -Uqq nixtla"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide \n",
    "from nixtla.utils import in_colab"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide \n",
    "IN_COLAB = in_colab()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide\n",
    "if not IN_COLAB:\n",
    "    from nixtla.utils import colab_badge\n",
    "    from dotenv import load_dotenv"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fine-tuning\n",
    "\n",
    "We can fine-tune TimeGPT by specifying the `finetune_steps` parameter."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| echo: false\n",
    "if not IN_COLAB:\n",
    "    load_dotenv()\n",
    "    colab_badge('docs/capabilities/forecast/07_finetuning')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from nixtla import NixtlaClient"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "nixtla_client = NixtlaClient(\n",
    "    # defaults to os.environ.get(\"NIXTLA_API_KEY\")\n",
    "    api_key = 'my_api_key_provided_by_nixtla'\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> 👍 Use an Azure AI endpoint\n",
    "> \n",
    "> To use an Azure AI endpoint, remember to set also the `base_url` argument:\n",
    "> \n",
    "> `nixtla_client = NixtlaClient(base_url=\"you azure ai endpoint\", api_key=\"your api_key\")`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide\n",
    "if not IN_COLAB:\n",
    "    nixtla_client = NixtlaClient()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read data\n",
    "df = pd.read_csv(\"https://raw.githubusercontent.com/Nixtla/transfer-learning-time-series/main/datasets/air_passengers.csv\")\n",
    "\n",
    "# Forecast with fine-tuning.\n",
    "# Here, we fine-tune for 5 steps\n",
    "forecast_df = nixtla_client.forecast(\n",
    "    df=df,\n",
    "    h=12,\n",
    "    finetune_steps=5,\n",
    "    time_col='timestamp',\n",
    "    target_col=\"value\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> 📘 Available models in Azure AI\n",
    ">\n",
    "> If you are using an Azure AI endpoint, please be sure to set `model=\"azureai\"`:\n",
    ">\n",
    "> `nixtla_client.forecast(..., model=\"azureai\")`\n",
    "> \n",
    "> For the public API, we support two models: `timegpt-1` and `timegpt-1-long-horizon`. \n",
    "> \n",
    "> By default, `timegpt-1` is used. Please see [this tutorial](https://docs.nixtla.io/docs/tutorials-long_horizon_forecasting) on how and when to use `timegpt-1-long-horizon`."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By default, only a small amount of finetuning is applied (`finetune_depth=1`). We can increase the intensity of finetuning by increasing the `finetune_depth` parameter. Note that increasing `finetune_depth` and `finetune_steps` increases wall time for generating predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read data\n",
    "df = pd.read_csv(\"https://raw.githubusercontent.com/Nixtla/transfer-learning-time-series/main/datasets/air_passengers.csv\")\n",
    "\n",
    "# Forecast with fine-tuning.\n",
    "# Here, we fine-tune for 5 steps\n",
    "# and we finetune more than just the last layer\n",
    "forecast_df = nixtla_client.forecast(\n",
    "    df=df,\n",
    "    h=12,\n",
    "    finetune_steps=5,\n",
    "    finetune_depth=2,\n",
    "    time_col='timestamp',\n",
    "    target_col=\"value\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For more information on fine-tuning, read our [fine-tuning tutorial](https://docs.nixtla.io/docs/tutorials-fine_tuning)."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "python3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}