Array API example for GMM

array_api_gmm.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "91d24835-d13b-4ac1-b4f2-47925ddb4a2a",
   "metadata": {},
   "source": [
    "# Array API example for GMM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a849ca35-90d9-4783-810d-18250bc021ee",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import make_blobs\n",
    "from sklearn.mixture import GaussianMixture\n",
    "\n",
    "X_np, _ = make_blobs(random_state=0, n_samples=200_000, n_features=30)\n",
    "gm_np = GaussianMixture(n_components=10, random_state=0, init_params=\"random\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8eefb285-86e8-4472-ac98-bd292e7e0546",
   "metadata": {},
   "source": [
    "## Runtime for NumPy array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "a1831332-4c8e-4dcf-abdc-f202e747dad1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 30.6 s, sys: 57 s, total: 1min 27s\n",
      "Wall time: 3.41 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GaussianMixture(init_params='random', n_components=10, random_state=0)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "gm_np.fit(X_np)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f8e7614-3b48-4955-8c02-4883c91fd327",
   "metadata": {},
   "source": [
    "## Runtime for CuPy array with array_api"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "ea84e268-4ca3-46fe-b5b8-cb5b90799bbe",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import set_config\n",
    "import cupy.array_api as xp\n",
    "\n",
    "set_config(array_api_dispatch=True)\n",
    "\n",
    "X_cu = xp.asarray(X_np, copy=True)\n",
    "gm_cu = GaussianMixture(n_components=10, random_state=0, init_params=\"random\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "f828fcfc-ffc2-4ded-960f-d8151b624def",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 462 ms, sys: 136 ms, total: 598 ms\n",
      "Wall time: 597 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GaussianMixture(init_params='random', n_components=10, random_state=0)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "gm_cu.fit(X_cu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "c4ab0e4f-e5d9-472f-964a-4ad1107e9df8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "cupy.array_api._array_object.Array"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(gm_cu.means_)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8cc882ff-65aa-44b3-82f6-5bade630fec9",
   "metadata": {},
   "source": [
    "### Check that the means are the same"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "114a5700-2421-493c-99d8-3fffb6474356",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "np_means_ = np.asarray(gm_np.means_)\n",
    "\n",
    "# Is there a bettery way to convert cupy.array_api to a np.ndarray?\n",
    "cu_means_ = gm_cu.means_._array.get()\n",
    "\n",
    "np.testing.assert_allclose(np_means_, cu_means_)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "sk1-array_api (python3)",
   "language": "python",
   "name": "conda-env-sk1-array_api-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}