parcels_and_dask_POC_001.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Parallelizing Parcels with Dask\n",
    "\n",
    "This will show how to distribute Parcels execution (using the Brownian motion example) to a dask cluster."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prerequisites\n",
    "\n",
    "Install Parcels and dask:\n",
    "\n",
    "```bash\n",
    "conda create -n py3_parcels_dask -c conda-forge python=3 parcels dask distributed bokeh\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## First, reproduce the brownian motion example\n",
    "\n",
    "(Note that we pass the random seed as a keyword arg.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from parcels import FieldSet, Field, ParticleSet, ScipyParticle, JITParticle, BrownianMotion2D\n",
    "import numpy as np\n",
    "from datetime import timedelta as delta\n",
    "from parcels import random"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "ptype = {'scipy': ScipyParticle, 'jit': JITParticle}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def zeros_fieldset(xdim=2, ydim=2):\n",
    "    \"\"\"Generates a zero velocity field\"\"\"\n",
    "    lon = np.linspace(-20, 20, xdim, dtype=np.float32)\n",
    "    lat = np.linspace(-20, 20, ydim, dtype=np.float32)\n",
    "\n",
    "    dimensions = {'lon': lon, 'lat': lat}\n",
    "    data = {'U': np.zeros((ydim, xdim), dtype=np.float32),\n",
    "            'V': np.zeros((ydim, xdim), dtype=np.float32)}\n",
    "    return FieldSet.from_data(data, dimensions, mesh='spherical')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def run_brownian_example(mode='scipy', npart=3000, random_seed=12345, runtime=delta(days=1)):\n",
    "    fieldset = zeros_fieldset()\n",
    "\n",
    "    # Set diffusion constants.\n",
    "    kh_zonal = 100\n",
    "    kh_meridional = 100\n",
    "\n",
    "    # Create field of Kh_zonal and Kh_meridional, using same grid as U\n",
    "    grid = fieldset.U.grid\n",
    "    fieldset.add_field(Field('Kh_zonal', kh_zonal * np.ones((2, 2)),\n",
    "                             grid=grid))\n",
    "    fieldset.add_field(Field('Kh_meridional', kh_meridional * np.ones((2, 2)),\n",
    "                             grid=grid))\n",
    "\n",
    "    # Set random seed\n",
    "    random.seed(random_seed)\n",
    "    \n",
    "    pset = ParticleSet(fieldset=fieldset, pclass=ptype[mode],\n",
    "                       lon=np.zeros(npart), lat=np.zeros(npart))\n",
    "    pset.execute(pset.Kernel(BrownianMotion2D),\n",
    "                 runtime=runtime, dt=delta(hours=1))\n",
    "\n",
    "    return pset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run the default setting once."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: Casting field data to np.float32\n",
      "INFO: Compiled random ==> /tmp/parcels-1000/random.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/8bab2818c41e5b898bc8af669c861c89.so\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 748 ms, sys: 28 ms, total: 776 ms\n",
      "Wall time: 1.88 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "pset = run_brownian_example(mode='jit')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Next step (no dask yet):  Create a list of realizations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create a list of seeds and map `run_brownian_motion` to it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0 1 2 3]\n"
     ]
    }
   ],
   "source": [
    "seeds = np.arange(4)\n",
    "print(seeds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/74d9a05b0d6153dad987f4ba8db53401.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/4efabbfec368a792a3d82d635b2bca44.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/330323f6a53c823ba595ac51dfb0cc89.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/c85aaaae4e4face49067612083956ad8.so\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 2.25 s, sys: 40 ms, total: 2.29 s\n",
      "Wall time: 5.85 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "list_of_psets = list(map(lambda s: run_brownian_example(mode='jit', random_seed=s),\n",
    "                         seeds))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Now, execute Parcels for a list of realizations using a local dask cluster"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import dask.bag as db\n",
    "from dask.distributed import Client"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "client = Client(n_workers=2, threads_per_worker=2, processes=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table style=\"border: 2px solid white;\">\n",
       "<tr>\n",
       "<td style=\"vertical-align: top; border: 0px solid white\">\n",
       "<h3>Client</h3>\n",
       "<ul>\n",
       "  <li><b>Scheduler: </b>inproc://134.245.217.57/19523/1\n",
       "  <li><b>Dashboard: </b><a href='http://localhost:8787/status' target='_blank'>http://localhost:8787/status</a>\n",
       "</ul>\n",
       "</td>\n",
       "<td style=\"vertical-align: top; border: 0px solid white\">\n",
       "<h3>Cluster</h3>\n",
       "<ul>\n",
       "  <li><b>Workers: </b>2</li>\n",
       "  <li><b>Cores: </b>4</li>\n",
       "  <li><b>Memory: </b>21.06 GB</li>\n",
       "</ul>\n",
       "</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<Client: scheduler='inproc://134.245.217.57/19523/1' processes=2 cores=4>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "client"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "_Make sure to **click on the link above!**_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "seeds_bag = db.from_sequence(seeds, partition_size=1)\n",
    "psets_jit_bag = db.map(lambda s: run_brownian_example(mode='jit', random_seed=s),\n",
    "                       seeds_bag)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/dee94dfa7edd708cf3eee30f83f87732.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/ebe079b32cad3e55cfbbf5a2a80c33f2.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/db1158fbc8444ce392fb212e0bbe7ffe.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/598b1ddd22d171d9594f803f5e0b48dc.so\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 2.93 s, sys: 124 ms, total: 3.05 s\n",
      "Wall time: 3.69 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "psets_jit = psets_jit_bag.compute()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Some longer integration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "psets_jit_long_bag = db.map(lambda s: run_brownian_example(mode='jit', random_seed=s,\n",
    "                                                           runtime=delta(days=10),\n",
    "                                                           npart=int(1e5)),\n",
    "                            seeds_bag)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/6225d33ce72a0ac407d98209270dff3f.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/3c3b97f9a2d4e6cfdda9911f46fab3c9.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/df8d25e1c966f0fdc79fb89cc8e0e9f8.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/352ebd0c80d39bda328b771ca01ba454.so\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 2min 44s, sys: 1min 6s, total: 3min 51s\n",
      "Wall time: 2min 11s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "psets_jit_long = psets_jit_long_bag.compute();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/c18eedd1070e8f2be92728c96858b77b.so\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 20.3 s, sys: 272 ms, total: 20.6 s\n",
      "Wall time: 20.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "pset = run_brownian_example(mode='jit', random_seed=0,\n",
    "                            runtime=delta(days=10), npart=int(1e5));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/52d6a1186ee413ffc8aa48dda45d74fe.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/2605570d8ae634bc878cedec40dd07b3.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/780253500d442926130d806a0d0e766c.so\n",
      "INFO: Compiled JITParticleBrownianMotion2D ==> /tmp/parcels-1000/671a3cb47ad9a8e167875b5c56736508.so\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 20s, sys: 2.17 s, total: 1min 22s\n",
      "Wall time: 1min 21s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "list_of_psets_long = list(map(lambda s: run_brownian_example(mode='jit', random_seed=s,\n",
    "                                                             runtime=delta(days=10),\n",
    "                                                             npart=int(1e5)),\n",
    "                              seeds))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Discussion\n",
    "\n",
    "- Note how we set `processes=False` in creation of the dask cluster.  Running different processes fails because the ctypes used in Parcels cannot be pickled.  (Test by re-running with `processes=True`.)\n",
    "\n",
    "- For the brownian motion example, we cannot circumvent this pickling issue by using Scipy particles, because `parcels.random` seems to imply some compilation.\n",
    "\n",
    "- Note that the approach outlined here could also be achived with Python's own `multiprocessing` toolbox.  But as there, inter-process communication is done by pickling name spaces as well, we're down to making sure that the C types can be pickled again.  On the up-side, getting Parcels to work with dask very likely also paves the way for many other (and still to come?) distributed computing tools that rely on some serialization for inter-worker communication."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Next steps\n",
    "\n",
    "- _**How does this scale?**_  — We'd probably like to allow for much longer run times and some machine larger than my desktop computer.\n",
    "\n",
    "- _**How to split tasks across dask resources?**_ — The obvious way is to go for splitting a particle set across different workers and re-merge them after the execution."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Boilerplate for the sake of (some) reproducibility"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# packages in environment at /home/wrath/miniconda3/envs/py3_parcels_dask:\n",
      "#\n",
      "appdirs                   1.4.3                      py_0    conda-forge\n",
      "attrs                     17.4.0                     py_0    conda-forge\n",
      "backcall                  0.1.0                      py_0    conda-forge\n",
      "binutils_impl_linux-64    2.28.1               had2808c_3  \n",
      "binutils_linux-64         7.2.0                        26  \n",
      "bleach                    2.1.3                      py_0    conda-forge\n",
      "bokeh                     0.12.15                  py36_0    conda-forge\n",
      "bottleneck                1.2.1                    py36_1    conda-forge\n",
      "bzip2                     1.0.6                         1    conda-forge\n",
      "ca-certificates           2018.4.16                     0    conda-forge\n",
      "cachetools                2.0.1                      py_0    conda-forge\n",
      "certifi                   2018.4.16                py36_0    conda-forge\n",
      "cgen                      2017.1                   py36_0    conda-forge\n",
      "click                     6.7                        py_1    conda-forge\n",
      "cloudpickle               0.5.2                      py_0    conda-forge\n",
      "curl                      7.59.0                        1    conda-forge\n",
      "cycler                    0.10.0                   py36_0    conda-forge\n",
      "cytoolz                   0.9.0.1                  py36_0    conda-forge\n",
      "dask                      0.17.3                     py_0    conda-forge\n",
      "dask-core                 0.17.3                     py_0    conda-forge\n",
      "dbus                      1.11.0                        0    conda-forge\n",
      "decorator                 4.3.0                      py_0    conda-forge\n",
      "distributed               1.21.7                   py36_0    conda-forge\n",
      "entrypoints               0.2.3                    py36_1    conda-forge\n",
      "expat                     2.2.5                         0    conda-forge\n",
      "ffmpeg                    3.2.4                         3    conda-forge\n",
      "fontconfig                2.12.6                        0    conda-forge\n",
      "freetype                  2.8.1                         0    conda-forge\n",
      "gcc_impl_linux-64         7.2.0                habb00fd_3  \n",
      "gcc_linux-64              7.2.0                        26  \n",
      "gettext                   0.19.8.1                      0    conda-forge\n",
      "glib                      2.55.0                        0    conda-forge\n",
      "gmp                       6.1.2                         0    conda-forge\n",
      "gst-plugins-base          1.8.0                         0    conda-forge\n",
      "gstreamer                 1.8.0                         1    conda-forge\n",
      "h5netcdf                  0.5.1                      py_0    conda-forge\n",
      "h5py                      2.7.1                    py36_2    conda-forge\n",
      "hdf4                      4.2.13                        0    conda-forge\n",
      "hdf5                      1.10.1                        2    conda-forge\n",
      "heapdict                  1.0.0                    py36_0    conda-forge\n",
      "html5lib                  1.0.1                      py_0    conda-forge\n",
      "icu                       58.2                          0    conda-forge\n",
      "intel-openmp              2018.0.0                      8  \n",
      "ipykernel                 4.8.2                    py36_0    conda-forge\n",
      "ipython                   6.3.1                    py36_0    conda-forge\n",
      "ipython_genutils          0.2.0                    py36_0    conda-forge\n",
      "jedi                      0.12.0                   py36_0    conda-forge\n",
      "jinja2                    2.10                     py36_0    conda-forge\n",
      "jpeg                      9b                            2    conda-forge\n",
      "jsonschema                2.6.0                    py36_1    conda-forge\n",
      "jupyter_client            5.2.3                    py36_0    conda-forge\n",
      "jupyter_core              4.4.0                      py_0    conda-forge\n",
      "jupyterlab                0.32.1                   py36_0    conda-forge\n",
      "jupyterlab_launcher       0.10.5                   py36_0    conda-forge\n",
      "krb5                      1.14.6                        0    conda-forge\n",
      "libffi                    3.2.1                         3    conda-forge\n",
      "libgcc                    7.2.0                h69d50b8_2  \n",
      "libgcc-ng                 7.2.0                hdf63c60_3  \n",
      "libgfortran               3.0.0                         1  \n",
      "libgfortran-ng            7.2.0                hdf63c60_3  \n",
      "libiconv                  1.15                          0    conda-forge\n",
      "libnetcdf                 4.6.1                         2    conda-forge\n",
      "libpng                    1.6.34                        0    conda-forge\n",
      "libsodium                 1.0.16                        0    conda-forge\n",
      "libssh2                   1.8.0                         2    conda-forge\n",
      "libstdcxx-ng              7.2.0                hdf63c60_3  \n",
      "libxcb                    1.13                          0    conda-forge\n",
      "libxml2                   2.9.8                         0    conda-forge\n",
      "locket                    0.2.0                    py36_1    conda-forge\n",
      "markupsafe                1.0                      py36_0    conda-forge\n",
      "matplotlib                2.0.2            py36h2acb4ad_1  \n",
      "mistune                   0.8.3                    py36_1    conda-forge\n",
      "mkl                       2018.0.2                      1  \n",
      "mkl_fft                   1.0.2                    py36_0    conda-forge\n",
      "mkl_random                1.0.1                    py36_0    conda-forge\n",
      "more-itertools            4.1.0                      py_0    conda-forge\n",
      "msgpack-python            0.5.6                    py36_0    conda-forge\n",
      "nbconvert                 5.3.1                      py_1    conda-forge\n",
      "nbformat                  4.4.0                    py36_0    conda-forge\n",
      "ncurses                   5.9                          10    conda-forge\n",
      "netcdf4                   1.3.1                    py36_3    conda-forge\n",
      "notebook                  5.4.1                    py36_0    conda-forge\n",
      "numpy                     1.14.2           py36hdbf6ddf_1  \n",
      "openssl                   1.0.2o                        0    conda-forge\n",
      "packaging                 17.1                       py_0    conda-forge\n",
      "pandas                    0.22.0                   py36_1    conda-forge\n",
      "pandoc                    2.2                           0    conda-forge\n",
      "pandocfilters             1.4.2                    py36_0    conda-forge\n",
      "parcels                   1.0.2                    py36_0    conda-forge\n",
      "parso                     0.2.0                      py_0    conda-forge\n",
      "partd                     0.3.8                    py36_0    conda-forge\n",
      "pcre                      8.41                          1    conda-forge\n",
      "pexpect                   4.5.0                    py36_0    conda-forge\n",
      "pickleshare               0.7.4                    py36_0    conda-forge\n",
      "pip                       9.0.3                    py36_0    conda-forge\n",
      "pluggy                    0.6.0                      py_0    conda-forge\n",
      "progressbar2              3.37.0                   py36_0    conda-forge\n",
      "prompt_toolkit            1.0.15                   py36_0    conda-forge\n",
      "psutil                    5.4.5                    py36_0    conda-forge\n",
      "ptyprocess                0.5.2                    py36_0    conda-forge\n",
      "py                        1.5.3                      py_0    conda-forge\n",
      "pygments                  2.2.0                    py36_0    conda-forge\n",
      "pymbolic                  2017.1                   py36_0    conda-forge\n",
      "pyparsing                 2.2.0                    py36_0    conda-forge\n",
      "pyqt                      5.6.0                    py36_5    conda-forge\n",
      "pytest                    3.5.1                    py36_0    conda-forge\n",
      "pytest-runner             4.2                        py_0    conda-forge\n",
      "python                    3.6.5                         1    conda-forge\n",
      "python-dateutil           2.7.2                      py_0    conda-forge\n",
      "python-utils              2.3.0                    py36_0    conda-forge\n",
      "pytools                   2018.4                     py_0    conda-forge\n",
      "pytz                      2018.4                     py_0    conda-forge\n",
      "pyyaml                    3.12                     py36_1    conda-forge\n",
      "pyzmq                     17.0.0                   py36_4    conda-forge\n",
      "qt                        5.6.2                         7    conda-forge\n",
      "readline                  7.0                           0    conda-forge\n",
      "scipy                     1.0.1            py36hfc37229_0  \n",
      "send2trash                1.5.0                      py_0    conda-forge\n",
      "setuptools                39.1.0                   py36_0    conda-forge\n",
      "simplegeneric             0.8.1                    py36_0    conda-forge\n",
      "sip                       4.18                     py36_1    conda-forge\n",
      "six                       1.11.0                   py36_1    conda-forge\n",
      "sortedcontainers          1.5.10                   py36_0    conda-forge\n",
      "sqlite                    3.20.1                        2    conda-forge\n",
      "tblib                     1.3.2                    py36_0    conda-forge\n",
      "terminado                 0.8.1                    py36_0    conda-forge\n",
      "testpath                  0.3.1                    py36_0    conda-forge\n",
      "tk                        8.6.7                         0    conda-forge\n",
      "toolz                     0.9.0                      py_0    conda-forge\n",
      "tornado                   5.0.2                    py36_0    conda-forge\n",
      "traitlets                 4.3.2                    py36_0    conda-forge\n",
      "wcwidth                   0.1.7                    py36_0    conda-forge\n",
      "webencodings              0.5                      py36_0    conda-forge\n",
      "wheel                     0.31.0                   py36_0    conda-forge\n",
      "x264                      20131218                      0    conda-forge\n",
      "xarray                    0.10.3                   py36_0    conda-forge\n",
      "xorg-libxau               1.0.8                         3    conda-forge\n",
      "xorg-libxdmcp             1.1.2                         3    conda-forge\n",
      "xz                        5.2.3                         0    conda-forge\n",
      "yaml                      0.1.7                         0    conda-forge\n",
      "zeromq                    4.2.5                         1    conda-forge\n",
      "zict                      0.1.3                      py_0    conda-forge\n",
      "zlib                      1.2.11                        0    conda-forge\n"
     ]
    }
   ],
   "source": [
    "!conda list"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}