{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!conda create -n ML_gpu tensorflow-gpu pytorch-gpu xarray dask matplotlib nb_conda_kernels jupyterlab cudatoolkit cupy esmtools climpred -y\n",
    "#!pip install git+https://github.com/jacobtomlinson/cupy-xarray.git"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# packages in environment at /work/mh0727/m300524/conda-envs/ML_gpu:\n",
      "#\n",
      "# Name                    Version                   Build  Channel\n",
      "cupy                      8.6.0            py39h694feb1_0    conda-forge\n",
      "cupy-xarray               0.1.0+2.g08c5779          pypi_0    pypi\n"
     ]
    }
   ],
   "source": [
    "!conda list cupy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def nvidia_smi():\n",
    "    import subprocess\n",
    "    p = subprocess.check_output('nvidia-smi').strip().decode('utf-8')\n",
    "    return p"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## climpred on CPU vs GPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import xarray as xr\n",
    "import numpy as np\n",
    "from climpred.tutorial import load_dataset\n",
    "from climpred import PerfectModelEnsemble"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "v = \"tos\"\n",
    "ds3d = load_dataset(\"MPI-PM-DP-3D\")[v]\n",
    "ds3d.lead.attrs[\"unit\"] = \"years\"\n",
    "control3d = load_dataset(\"MPI-control-3D\")[v]\n",
    "\n",
    "pm_cpu = PerfectModelEnsemble(ds3d)\n",
    "pm_cpu = pm_cpu.add_control(control3d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "numpy.ndarray"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(ds3d.data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "24.3 ms ± 147 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
     ]
    }
   ],
   "source": [
    "%timeit _ = pm_cpu.mean(['x','y'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "667 ms ± 29.6 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%timeit skill_cpu = pm_cpu.verify(metric=\"pearson_r\", comparison=\"m2m\", dim=[\"init\", \"member\"])[v]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### GPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cupy_xarray"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/mnt/lustre01/pf/zmaw/m300524/climpred/climpred/utils.py:140: UserWarning: Assuming annual resolution due to numeric inits. Change init to a datetime if it is another resolution.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "v = \"tos\"\n",
    "ds3d = load_dataset(\"MPI-PM-DP-3D\")[v].as_cupy()\n",
    "ds3d.lead.attrs[\"unit\"] = \"years\"\n",
    "control3d = load_dataset(\"MPI-control-3D\")[v].as_cupy()\n",
    "\n",
    "pm_gpu = PerfectModelEnsemble(ds3d)\n",
    "pm_gpu = pm_gpu.add_control(control3d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "cupy.core.core.ndarray"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(ds3d.data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "423 µs ± 15 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)\n"
     ]
    }
   ],
   "source": [
    "%timeit _ = pm_gpu.mean(['x','y'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "64.1 ms ± 11.6 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
     ]
    }
   ],
   "source": [
    "%timeit skill_gpu = pm_gpu.verify(metric=\"pearson_r\", comparison=\"m2m\", dim=[\"init\", \"member\"])[v]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> 🚀 GPU has the potential to increase verify by factor 10! 🚀"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### GPU Limitations\n",
    "\n",
    "with {py:meth}`.PerfectModelEnsemble.bootstrap`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "no implementation found for 'numpy.quantile' on types that implement __array_function__: [<class 'cupy.core.core.ndarray'>, <class 'numpy.ndarray'>]",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-20-45b2445bf7cd>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mpm_gpu\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbootstrap\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmetric\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"pearson_r\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcomparison\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"m2m\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"init\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"member\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0miterations\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m10\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mv\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/mnt/lustre01/pf/zmaw/m300524/climpred/climpred/classes.py\u001b[0m in \u001b[0;36mbootstrap\u001b[0;34m(self, metric, comparison, dim, reference, iterations, sig, pers_sig, **metric_kwargs)\u001b[0m\n\u001b[1;32m    860\u001b[0m             \u001b[0;34m\"init\"\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    861\u001b[0m         }\n\u001b[0;32m--> 862\u001b[0;31m         return self._apply_climpred_function(\n\u001b[0m\u001b[1;32m    863\u001b[0m             \u001b[0mbootstrap_perfect_model\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    864\u001b[0m             \u001b[0minput_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minput_dict\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/mnt/lustre01/pf/zmaw/m300524/climpred/climpred/classes.py\u001b[0m in \u001b[0;36m_apply_climpred_function\u001b[0;34m(self, func, input_dict, **kwargs)\u001b[0m\n\u001b[1;32m    545\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mcontrol\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    546\u001b[0m             \u001b[0mcontrol\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcontrol\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdrop_vars\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mctrl_vars\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 547\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mensemble\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcontrol\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    548\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    549\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_vars_to_drop\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minit\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/mnt/lustre01/pf/zmaw/m300524/climpred/climpred/bootstrap.py\u001b[0m in \u001b[0;36mbootstrap_perfect_model\u001b[0;34m(init_pm, control, metric, comparison, dim, reference, resample_dim, sig, iterations, pers_sig, reference_compute, **metric_kwargs)\u001b[0m\n\u001b[1;32m   1259\u001b[0m     )\n\u001b[1;32m   1260\u001b[0m     \u001b[0mlead_units_equal_control_time_stride\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minit_pm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcontrol\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1261\u001b[0;31m     return bootstrap_compute(\n\u001b[0m\u001b[1;32m   1262\u001b[0m         \u001b[0minit_pm\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1263\u001b[0m         \u001b[0mcontrol\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/mnt/lustre01/pf/zmaw/m300524/climpred/climpred/bootstrap.py\u001b[0m in \u001b[0;36mbootstrap_compute\u001b[0;34m(hind, verif, hist, alignment, metric, comparison, dim, reference, resample_dim, sig, iterations, pers_sig, compute, resample_uninit, reference_compute, **metric_kwargs)\u001b[0m\n\u001b[1;32m    899\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    900\u001b[0m     \u001b[0;31m# get confidence intervals CI\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 901\u001b[0;31m     \u001b[0minit_ci\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_distribution_to_ci\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbootstrapped_init_skill\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mci_low\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mci_high\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    902\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;34m\"uninitialized\"\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mreference\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    903\u001b[0m         \u001b[0muninit_ci\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_distribution_to_ci\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbootstrapped_uninit_skill\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mci_low\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mci_high\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/mnt/lustre01/pf/zmaw/m300524/climpred/climpred/bootstrap.py\u001b[0m in \u001b[0;36m_distribution_to_ci\u001b[0;34m(ds, ci_low, ci_high, dim)\u001b[0m\n\u001b[1;32m    211\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0missubdtype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mds\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdtype\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbool_\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    212\u001b[0m             \u001b[0mds\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mds\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mastype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfloat_\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# fails on py>36 if boolean dtype\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 213\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0mds\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mquantile\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mq\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mci_low\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mci_high\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mdim\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mskipna\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    214\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    215\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/work/mh0727/m300524/conda-envs/ML_gpu/lib/python3.9/site-packages/xarray/core/dataset.py\u001b[0m in \u001b[0;36mquantile\u001b[0;34m(self, q, dim, interpolation, numeric_only, keep_attrs, skipna)\u001b[0m\n\u001b[1;32m   5819\u001b[0m                             \u001b[0;31m# the former is often more efficient\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5820\u001b[0m                             \u001b[0mreduce_dims\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 5821\u001b[0;31m                         variables[name] = var.quantile(\n\u001b[0m\u001b[1;32m   5822\u001b[0m                             \u001b[0mq\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5823\u001b[0m                             \u001b[0mdim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mreduce_dims\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/work/mh0727/m300524/conda-envs/ML_gpu/lib/python3.9/site-packages/xarray/core/variable.py\u001b[0m in \u001b[0;36mquantile\u001b[0;34m(self, q, dim, interpolation, keep_attrs, skipna)\u001b[0m\n\u001b[1;32m   1951\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1952\u001b[0m         \u001b[0maxis\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdim\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1953\u001b[0;31m         result = apply_ufunc(\n\u001b[0m\u001b[1;32m   1954\u001b[0m             \u001b[0m_wrapper\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1955\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/work/mh0727/m300524/conda-envs/ML_gpu/lib/python3.9/site-packages/xarray/core/computation.py\u001b[0m in \u001b[0;36mapply_ufunc\u001b[0;34m(func, input_core_dims, output_core_dims, exclude_dims, vectorize, join, dataset_join, dataset_fill_value, keep_attrs, kwargs, dask, output_dtypes, output_sizes, meta, dask_gufunc_kwargs, *args)\u001b[0m\n\u001b[1;32m   1136\u001b[0m     \u001b[0;31m# feed Variables directly through apply_variable_ufunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1137\u001b[0m     \u001b[0;32melif\u001b[0m \u001b[0many\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mVariable\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0ma\u001b[0m \u001b[0;32min\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1138\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mvariables_vfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1139\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1140\u001b[0m         \u001b[0;31m# feed anything else through apply_array_ufunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/work/mh0727/m300524/conda-envs/ML_gpu/lib/python3.9/site-packages/xarray/core/computation.py\u001b[0m in \u001b[0;36mapply_variable_ufunc\u001b[0;34m(func, signature, exclude_dims, dask, output_dtypes, vectorize, keep_attrs, dask_gufunc_kwargs, *args)\u001b[0m\n\u001b[1;32m    722\u001b[0m             )\n\u001b[1;32m    723\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 724\u001b[0;31m     \u001b[0mresult_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput_data\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    725\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    726\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0msignature\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_outputs\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/work/mh0727/m300524/conda-envs/ML_gpu/lib/python3.9/site-packages/xarray/core/variable.py\u001b[0m in \u001b[0;36m_wrapper\u001b[0;34m(npa, **kwargs)\u001b[0m\n\u001b[1;32m   1948\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0m_wrapper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnpa\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1949\u001b[0m             \u001b[0;31m# move quantile axis to end. required for apply_ufunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1950\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmoveaxis\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_quantile_func\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnpa\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1951\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1952\u001b[0m         \u001b[0maxis\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdim\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<__array_function__ internals>\u001b[0m in \u001b[0;36mquantile\u001b[0;34m(*args, **kwargs)\u001b[0m\n",
      "\u001b[0;31mTypeError\u001b[0m: no implementation found for 'numpy.quantile' on types that implement __array_function__: [<class 'cupy.core.core.ndarray'>, <class 'numpy.ndarray'>]"
     ]
    }
   ],
   "source": [
    "pm_gpu.bootstrap(metric=\"pearson_r\", comparison=\"m2m\", dim=[\"init\", \"member\"], iterations=10)[v]"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}