🎉 Sklearn pipelines wrapper

docs/source/notebooks/mmgp_pipeline.ipynb

@@ -0,0 +1,323 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Exemple of pipeline PCA-GP-PCA type"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "from pathlib import Path\n",
+    "import numpy as np\n",
+    "\n",
+    "from sklearn.base import BaseEstimator, ClassifierMixin, RegressorMixin, TransformerMixin, MetaEstimatorMixin\n",
+    "from sklearn.compose import ColumnTransformer, TransformedTargetRegressor\n",
+    "from sklearn.decomposition import PCA\n",
+    "from sklearn.pipeline import Pipeline\n",
+    "from sklearn.preprocessing import MinMaxScaler, StandardScaler\n",
+    "from sklearn.datasets import make_regression\n",
+    "from sklearn.gaussian_process import GaussianProcessRegressor\n",
+    "\n",
+    "from plaid.containers.dataset import Dataset\n",
+    "from plaid.containers.sample import Sample\n",
+    "from plaid.problem_definition import ProblemDefinition\n",
+    "from plaid.wrappers.sklearn import WrappedSklearnTransform, WrappedSklearnRegressor"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Load Rotor37 PLAID dataset"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# data_dir = Path(os.environ['AIRFRANS_PLAID_DATASET_PATH'])\n",
+    "data_dir = Path('/gpfs_new/cold-data/InputData/public_datasets/data_challenge/Rotor37')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "ename": "FileNotFoundError",
+     "evalue": "\"\\gpfs_new\\cold-data\\InputData\\public_datasets\\data_challenge\\Rotor37\\dataset\" is not a directory or does not exist. Abort",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[31m---------------------------------------------------------------------------\u001b[39m",
+      "\u001b[31mFileNotFoundError\u001b[39m                         Traceback (most recent call last)",
+      "\u001b[36mCell\u001b[39m\u001b[36m \u001b[39m\u001b[32mIn[3]\u001b[39m\u001b[32m, line 3\u001b[39m\n\u001b[32m      1\u001b[39m \u001b[38;5;66;03m# dset = Dataset(data_dir/'dataset', processes_number=2)\u001b[39;00m\n\u001b[32m      2\u001b[39m dset = Dataset()\n\u001b[32m----> \u001b[39m\u001b[32m3\u001b[39m \u001b[43mdset\u001b[49m\u001b[43m.\u001b[49m\u001b[43m_load_from_dir_\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdata_dir\u001b[49m\u001b[43m/\u001b[49m\u001b[33;43m'\u001b[39;49m\u001b[33;43mdataset\u001b[39;49m\u001b[33;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43mids\u001b[49m\u001b[43m \u001b[49m\u001b[43m=\u001b[49m\u001b[43m \u001b[49m\u001b[43mnp\u001b[49m\u001b[43m.\u001b[49m\u001b[43marange\u001b[49m\u001b[43m(\u001b[49m\u001b[32;43m13\u001b[39;49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m      4\u001b[39m problem_def = ProblemDefinition(data_dir/\u001b[33m'\u001b[39m\u001b[33mproblem_definition\u001b[39m\u001b[33m'\u001b[39m)\n",
+      "\u001b[36mFile \u001b[39m\u001b[32m~\\Code\\travaux\\plaid\\src\\plaid\\containers\\dataset.py:1006\u001b[39m, in \u001b[36mDataset._load_from_dir_\u001b[39m\u001b[34m(self, savedir, ids, verbose, processes_number)\u001b[39m\n\u001b[32m   1004\u001b[39m savedir = Path(savedir)\n\u001b[32m   1005\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m savedir.is_dir():\n\u001b[32m-> \u001b[39m\u001b[32m1006\u001b[39m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mFileNotFoundError\u001b[39;00m(\n\u001b[32m   1007\u001b[39m         \u001b[33mf\u001b[39m\u001b[33m'\u001b[39m\u001b[33m\"\u001b[39m\u001b[38;5;132;01m{\u001b[39;00msavedir\u001b[38;5;132;01m}\u001b[39;00m\u001b[33m\"\u001b[39m\u001b[33m is not a directory or does not exist. Abort\u001b[39m\u001b[33m'\u001b[39m\n\u001b[32m   1008\u001b[39m     )\n\u001b[32m   1010\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m processes_number < -\u001b[32m1\u001b[39m:\n\u001b[32m   1011\u001b[39m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\u001b[33m\"\u001b[39m\u001b[33mNumber of processes cannot be < -1\u001b[39m\u001b[33m\"\u001b[39m)\n",
+      "\u001b[31mFileNotFoundError\u001b[39m: \"\\gpfs_new\\cold-data\\InputData\\public_datasets\\data_challenge\\Rotor37\\dataset\" is not a directory or does not exist. Abort"
+     ]
+    }
+   ],
+   "source": [
+    "# dset = Dataset(data_dir/'dataset', processes_number=2)\n",
+    "dset = Dataset()\n",
+    "dset._load_from_dir_(data_dir/'dataset',ids = np.arange(13))\n",
+    "problem_def = ProblemDefinition(data_dir/'problem_definition')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(f\"{dset.get_scalar_names()=}\")\n",
+    "print(f\"{dset.get_field_names()=}\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## PCA-GP-PCA as a sklearn `Pipeline`"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 1. Define the PCA for the shape embedding\n",
+    "\n",
+    "In this example we only apply PCA to the first 8 columns\n",
+    "\n",
+    "The last two columns are unchanged"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "NB_PCA_MODES = 8\n",
+    "from sklearn.decomposition import PCA\n",
+    "pca = WrappedSklearnTransform(\n",
+    "            PCA(NB_PCA_MODES),\n",
+    "            in_keys='field::all',\n",
+    "            # in_keys=['omega', 'compression_rate'],\n",
+    "            out_keys=[f'scalar::pca{i_mode}' for i_mode in range(NB_PCA_MODES)],\n",
+    "        )"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pca.fit(dset, problem_def)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "feats_to_reduce = list(range(8))\n",
+    "preprocessor = ColumnTransformer(\n",
+    "    transformers=[\n",
+    "        (\n",
+    "            \"pca\",\n",
+    "            PCA(n_components=8),\n",
+    "            feats_to_reduce,\n",
+    "        ),\n",
+    "    ],\n",
+    "    remainder=\"passthrough\",\n",
+    ")\n",
+    "preprocessor"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 2. Define the output scaler for the output fields (MinMaxScaler + PCA)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "postprocessor = Pipeline(\n",
+    "    [\n",
+    "        (\"scaler\", MinMaxScaler()),\n",
+    "        (\"pca\", PCA(n_components=9)),\n",
+    "    ]\n",
+    ")\n",
+    "postprocessor"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 3. Define the regressor\n",
+    "\n",
+    "Y = GP(transformer(X)) where transformer(X) = postprocessor(X)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "regressor = TransformedTargetRegressor(\n",
+    "    regressor=GaussianProcessRegressor(n_restarts_optimizer=3),\n",
+    "    check_inverse=False,\n",
+    "    transformer=postprocessor,\n",
+    ")\n",
+    "regressor"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 4. Combine to make the pipeline"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model = Pipeline(\n",
+    "    steps=[\n",
+    "        (\"preprocessor\", preprocessor),\n",
+    "        (\"scaler\", StandardScaler()),\n",
+    "        (\"regressor\", regressor),\n",
+    "    ]\n",
+    ")\n",
+    "model"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Fit the model"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.fit(dset, problem_def)\n",
+    "model"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Predict on the training data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "y_pred = model.predict(dset)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Other way to define the pipeline"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 1. Define the regressor"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "regressor = Pipeline(\n",
+    "    steps=[\n",
+    "        (\"preprocessor\", preprocessor),\n",
+    "        (\"scaler\", StandardScaler()),\n",
+    "        (\"regressor\", GaussianProcessRegressor(n_restarts_optimizer=3)),\n",
+    "    ]\n",
+    ")\n",
+    "regressor"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 2. Combine to make the pipeline"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model = TransformedTargetRegressor(\n",
+    "    regressor=regressor,\n",
+    "    check_inverse=False,\n",
+    "    transformer=postprocessor,\n",
+    ")\n",
+    "model"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "plaid_dev",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}