param_search_example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Running experiments on a cluster with `param_search`\n",
    "\n",
    "This Python package makes it easy to launch lots of jobs with different parameters to a computing cluster. It also has features to monitor the status of running or completed jobs, aggregate their output metrics, and create some basic visualizations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/net/pulsar/home/koes/mtr22/ipython-notebooks'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import sys, os\n",
    "sys.path.append('../param_search')\n",
    "import param_search\n",
    "os.getcwd()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1) Create a template job script\n",
    "This is the script that you want to evaluate using ranges of different parameters. The parameters are specified as Python string formatting fields, i.e. using curly braces. There are also two optional built-in fields, `job_name` and `job_params`. The `job_name` field will be filled in with a parameterized job name, and the `job_params` field will be replaced by a comment block that displays what parameters were used to create each script."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "template_file = 'my_job.sh'\n",
    "template = '''\\\n",
    "#!/bin/bash\n",
    "#SBATCH --job-name={job_name}\n",
    "#SBATCH -o %J.stdout\n",
    "#SBATCH -e %J.stderr\n",
    "{job_params}\n",
    "product=$(python3 -c \"print({term1}*{term2})\")\n",
    "quotient=$(python3 -c \"print({term1}/{term2})\")\n",
    "echo \"product quotient\\n$product $quotient\" > $SLURM_JOB_ID.metrics\n",
    "'''\n",
    "with open(template_file, 'w') as f:\n",
    "    f.write(template)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2) Define the parameter search space\n",
    "\n",
    "Use the `ParamSpace` object to define the ranges of values for the parameters you want to explore. It is a subclass of `OrderedDict` that expects each key to be assigned as either a single value or iterable of values (strings are considered single values). Then, iterating over the `ParamSpace` produces parameter assignments from the Cartesian product over the defined parameter ranges."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Params([('term1', 0), ('term2', 0)]),\n",
       " Params([('term1', 0), ('term2', 1)]),\n",
       " Params([('term1', 0), ('term2', 2)]),\n",
       " Params([('term1', 0), ('term2', 3)]),\n",
       " Params([('term1', 1), ('term2', 0)]),\n",
       " Params([('term1', 1), ('term2', 1)]),\n",
       " Params([('term1', 1), ('term2', 2)]),\n",
       " Params([('term1', 1), ('term2', 3)]),\n",
       " Params([('term1', 2), ('term2', 0)]),\n",
       " Params([('term1', 2), ('term2', 1)]),\n",
       " Params([('term1', 2), ('term2', 2)]),\n",
       " Params([('term1', 2), ('term2', 3)]),\n",
       " Params([('term1', 3), ('term2', 0)]),\n",
       " Params([('term1', 3), ('term2', 1)]),\n",
       " Params([('term1', 3), ('term2', 2)]),\n",
       " Params([('term1', 3), ('term2', 3)])]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "param_space = param_search.ParamSpace(\n",
    "    term1=range(4),\n",
    "    term2=range(4),\n",
    ")\n",
    "list(param_space)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3) Setup parameterized job scripts\n",
    "The `setup` function fills in the parameter fields in the template script with each parameter assignment in the parameter space, then writes out each of these parameterized job scripts to a separate subdirectory. The subdirectories are named according to a Python formatting string `name_format`, which is also used to fill in the `job_name` field in the template script."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['my_experiment/my_job_0_0/my_job.sh',\n",
       " 'my_experiment/my_job_0_1/my_job.sh',\n",
       " 'my_experiment/my_job_0_2/my_job.sh',\n",
       " 'my_experiment/my_job_0_3/my_job.sh',\n",
       " 'my_experiment/my_job_1_0/my_job.sh',\n",
       " 'my_experiment/my_job_1_1/my_job.sh',\n",
       " 'my_experiment/my_job_1_2/my_job.sh',\n",
       " 'my_experiment/my_job_1_3/my_job.sh',\n",
       " 'my_experiment/my_job_2_0/my_job.sh',\n",
       " 'my_experiment/my_job_2_1/my_job.sh',\n",
       " 'my_experiment/my_job_2_2/my_job.sh',\n",
       " 'my_experiment/my_job_2_3/my_job.sh',\n",
       " 'my_experiment/my_job_3_0/my_job.sh',\n",
       " 'my_experiment/my_job_3_1/my_job.sh',\n",
       " 'my_experiment/my_job_3_2/my_job.sh',\n",
       " 'my_experiment/my_job_3_3/my_job.sh']"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "job_files = param_search.setup(\n",
    "    expt_dir='my_experiment',\n",
    "    name_format='my_job_{term1}_{term2}',\n",
    "    template_file=template_file,\n",
    "    param_space=param_space\n",
    ")\n",
    "job_files"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4) Submit job scripts to the cluster\n",
    "The `submit` function submits a list of job scripts to a computing cluster with Slurm or Torque. It returns the job IDs associated with each submitted job."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[7629409,\n",
       " 7629410,\n",
       " 7629411,\n",
       " 7629412,\n",
       " 7629413,\n",
       " 7629414,\n",
       " 7629415,\n",
       " 7629416,\n",
       " 7629417,\n",
       " 7629418,\n",
       " 7629419,\n",
       " 7629420,\n",
       " 7629421,\n",
       " 7629422,\n",
       " 7629423,\n",
       " 7629424]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "job_ids = param_search.submit(job_files)\n",
    "job_ids"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5) Monitor the status of queued and running jobs\n",
    "The `status` function checks the current status of a list of jobs in a Slurm or Torque computing cluster."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>job_id</th>\n",
       "      <th>queue</th>\n",
       "      <th>job_name</th>\n",
       "      <th>user</th>\n",
       "      <th>job_state</th>\n",
       "      <th>runtime</th>\n",
       "      <th>walltime</th>\n",
       "      <th>node_id</th>\n",
       "      <th>work_dir</th>\n",
       "      <th>array_idx</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7629411</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_0_2</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>CG</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>7629412</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_0_3</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>CG</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>7629413</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_1_0</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>CG</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7629423</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_3_2</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>PD</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>(None)</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>7629424</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_3_3</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>PD</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>(Priority)</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>7629409</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_0_0</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>7629410</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_0_1</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>7629414</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_1_1</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>7629415</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_1_2</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>7629416</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_1_3</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>7629417</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_2_0</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>7629418</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_2_1</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>7629419</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_2_2</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>7629420</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_2_3</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>7629421</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_3_0</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>7629422</td>\n",
       "      <td>any_cpu</td>\n",
       "      <td>my_job_3_1</td>\n",
       "      <td>mtr22</td>\n",
       "      <td>R</td>\n",
       "      <td>0:00</td>\n",
       "      <td>28-00:00:00</td>\n",
       "      <td>n029</td>\n",
       "      <td>/net/pulsar/home/koes/mtr22/ipython-notebooks/...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     job_id    queue    job_name   user job_state runtime     walltime  \\\n",
       "0   7629411  any_cpu  my_job_0_2  mtr22        CG    0:00  28-00:00:00   \n",
       "1   7629412  any_cpu  my_job_0_3  mtr22        CG    0:00  28-00:00:00   \n",
       "2   7629413  any_cpu  my_job_1_0  mtr22        CG    0:00  28-00:00:00   \n",
       "3   7629423  any_cpu  my_job_3_2  mtr22        PD    0:00  28-00:00:00   \n",
       "4   7629424  any_cpu  my_job_3_3  mtr22        PD    0:00  28-00:00:00   \n",
       "5   7629409  any_cpu  my_job_0_0  mtr22         R    0:00  28-00:00:00   \n",
       "6   7629410  any_cpu  my_job_0_1  mtr22         R    0:00  28-00:00:00   \n",
       "7   7629414  any_cpu  my_job_1_1  mtr22         R    0:00  28-00:00:00   \n",
       "8   7629415  any_cpu  my_job_1_2  mtr22         R    0:00  28-00:00:00   \n",
       "9   7629416  any_cpu  my_job_1_3  mtr22         R    0:00  28-00:00:00   \n",
       "10  7629417  any_cpu  my_job_2_0  mtr22         R    0:00  28-00:00:00   \n",
       "11  7629418  any_cpu  my_job_2_1  mtr22         R    0:00  28-00:00:00   \n",
       "12  7629419  any_cpu  my_job_2_2  mtr22         R    0:00  28-00:00:00   \n",
       "13  7629420  any_cpu  my_job_2_3  mtr22         R    0:00  28-00:00:00   \n",
       "14  7629421  any_cpu  my_job_3_0  mtr22         R    0:00  28-00:00:00   \n",
       "15  7629422  any_cpu  my_job_3_1  mtr22         R    0:00  28-00:00:00   \n",
       "\n",
       "       node_id                                           work_dir  array_idx  \n",
       "0         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "1         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "2         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "3       (None)  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "4   (Priority)  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "5         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "6         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "7         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "8         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "9         n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "10        n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "11        n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "12        n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "13        n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "14        n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  \n",
       "15        n029  /net/pulsar/home/koes/mtr22/ipython-notebooks/...        NaN  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "qstat = param_search.status(job=job_ids)\n",
    "qstat"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6) Check completed jobs for errors\n",
    "The `errors` function reads stderr files for the most recent submission of a list of job scripts and parses them for error messages."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "................"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['ZeroDivisionError: division by zero',\n",
       " None,\n",
       " None,\n",
       " None,\n",
       " 'ZeroDivisionError: division by zero',\n",
       " None,\n",
       " None,\n",
       " None,\n",
       " 'ZeroDivisionError: division by zero',\n",
       " None,\n",
       " None,\n",
       " None,\n",
       " 'ZeroDivisionError: division by zero',\n",
       " None,\n",
       " None,\n",
       " None]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "param_search.errors(job_files)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7) Aggregate output metrics of completed jobs\n",
    "The `metrics` function reads csv-style output files for a list of job scripts, concatenates them, and indexes them with the job parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>product</th>\n",
       "      <th>quotient</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>term1</th>\n",
       "      <th>term2</th>\n",
       "      <th>job_name</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th>my_job_0_0</th>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_0_0</th>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_0_0</th>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_0_0</th>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_0_0</th>\n",
       "      <td>0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th rowspan=\"5\" valign=\"top\">3</th>\n",
       "      <th>my_job_3_3</th>\n",
       "      <td>9</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_3_3</th>\n",
       "      <td>9</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_3_3</th>\n",
       "      <td>9</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_3_3</th>\n",
       "      <td>9</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>my_job_3_3</th>\n",
       "      <td>9</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>329 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                        product  quotient\n",
       "term1 term2 job_name                     \n",
       "0     0     my_job_0_0        0       NaN\n",
       "            my_job_0_0        0       NaN\n",
       "            my_job_0_0        0       NaN\n",
       "            my_job_0_0        0       NaN\n",
       "            my_job_0_0        0       NaN\n",
       "...                         ...       ...\n",
       "3     3     my_job_3_3        9       1.0\n",
       "            my_job_3_3        9       1.0\n",
       "            my_job_3_3        9       1.0\n",
       "            my_job_3_3        9       1.0\n",
       "            my_job_3_3        9       1.0\n",
       "\n",
       "[329 rows x 2 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics = param_search.metrics(job_files)\n",
    "metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8) Plot metrics with respect to parameters\n",
    "The `plot` function allows you to visualize the effects of parameters on job output metrics. It plots each metric against each parameter on separate axes, grouping by all of the other parameters. This default behavior can be altered by specifying different values for the `x`, `y`, and `hue` arguments."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = param_search.plot(metrics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}