automatic chunk size

anomaly-chunksize.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "import dask.array as da\n",
    "import numpy as np\n",
    "import dask\n",
    "import xarray as xr\n",
    "import math\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def climatology(ds):\n",
    "    seasonal_clim = ds.groupby(\"time.season\").mean(dim=\"time\")\n",
    "    return seasonal_clim\n",
    "def anomaly(ds):\n",
    "    seasonal_clim = climatology(ds)\n",
    "    seasonal_anom = ds.groupby(\"time.season\") - seasonal_clim\n",
    "    return seasonal_anom"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "dask.config.set({\"array.chunk-size\": '256MB'})\n",
    "timesteps=20834\n",
    "lat=320\n",
    "lon=384\n",
    "random_data =da.random.RandomState(0).standard_normal(shape, chunks='auto'  )\n",
    "lats = xr.DataArray(np.linspace(start=-90, stop=90, num=lat), dims=[\"lat\"])\n",
    "lons = xr.DataArray(np.linspace(start=-180, stop=180, num=lon), dims=[\"lon\"])\n",
    "times = xr.DataArray(pd.date_range(start=\"1980-01-01\", freq=\"1D\", periods=timesteps), dims=[\"time\"])\n",
    "ds = xr.DataArray(\n",
    "        random_data,\n",
    "        dims=[\"time\", \"lon\", \"lat\"],\n",
    "       coords={\"time\": times, \"lon\": lons, \"lat\": lats},\n",
    "        name=\"sst\",\n",
    ").to_dataset()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<xarray.Dataset>\n",
       "Dimensions:  (lat: 320, lon: 384, time: 20834)\n",
       "Coordinates:\n",
       "  * time     (time) datetime64[ns] 1980-01-01 1980-01-02 ... 2037-01-14\n",
       "  * lon      (lon) float64 -180.0 -179.1 -178.1 -177.2 ... 178.1 179.1 180.0\n",
       "  * lat      (lat) float64 -90.0 -89.44 -88.87 -88.31 ... 88.31 88.87 89.44 90.0\n",
       "Data variables:\n",
       "    sst      (time, lon, lat) float64 dask.array<shape=(20834, 384, 320), chunksize=(317, 192, 160)>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When specifying chunk size as 370MB in dask, dask array creates chunksize 187MB."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<xarray.Dataset>\n",
       "Dimensions:  (lat: 320, lon: 384, time: 20834)\n",
       "Coordinates:\n",
       "  * lon      (lon) float64 -180.0 -179.1 -178.1 -177.2 ... 178.1 179.1 180.0\n",
       "  * lat      (lat) float64 -90.0 -89.44 -88.87 -88.31 ... 88.31 88.87 89.44 90.0\n",
       "  * time     (time) datetime64[ns] 1980-01-01 1980-01-02 ... 2037-01-14\n",
       "    season   (time) <U3 'DJF' 'DJF' 'DJF' 'DJF' ... 'DJF' 'DJF' 'DJF' 'DJF'\n",
       "Data variables:\n",
       "    sst      (time, lon, lat) float64 dask.array<shape=(20834, 384, 320), chunksize=(60, 192, 160)>"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "anomaly(ds)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pangeobench",
   "language": "python",
   "name": "pangeobench"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}

anomaly-chunksizesimplifyed.ipynb

auto_chunkstest.ipynb

automatic_chunksize.ipynb