mlip_arena/tasks/eos/run.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from mp_api.client import MPRester\n",
    "from dask.distributed import Client\n",
    "from dask_jobqueue import SLURMCluster\n",
    "from prefect import task, flow\n",
    "from prefect.task_runners import ThreadPoolTaskRunner\n",
    "from prefect_dask import DaskTaskRunner\n",
    "from pymatgen.core.structure import Structure\n",
    "from dotenv import load_dotenv\n",
    "from ase import Atoms\n",
    "from ase.io import write, read\n",
    "from pathlib import Path\n",
    "import pandas as pd\n",
    "from prefect.futures import wait\n",
    "\n",
    "from mlip_arena.tasks.eos.run import fit as EOS\n",
    "from mlip_arena.models.utils import REGISTRY, MLIPEnum\n",
    "\n",
    "load_dotenv()\n",
    "\n",
    "MP_API_KEY = os.environ.get(\"MP_API_KEY\", None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MP Database version: 2023.11.1\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bb6c1969c89840888c556f8fa59b4a67",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Retrieving SummaryDoc documents:   0%|          | 0/5135 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "with MPRester(MP_API_KEY) as mpr:\n",
    "    print(\"MP Database version:\", mpr.get_database_version())\n",
    "\n",
    "    summary_docs = mpr.materials.summary.search(\n",
    "        num_elements=(1, 2),\n",
    "        is_stable=True,\n",
    "        fields=[\"material_id\", \"structure\", \"formula_pretty\"]\n",
    "    )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "atoms_list = []\n",
    "\n",
    "for doc in summary_docs:\n",
    "\n",
    "    structure = doc.structure\n",
    "    assert isinstance(structure, Structure)\n",
    "\n",
    "    atoms = structure.to_ase_atoms()\n",
    "\n",
    "    atoms_list.append(atoms)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "write(\"all.extxyz\", atoms_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "atoms_list = read(\"all.extxyz\", index=':')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "#!/bin/bash\n",
      "\n",
      "#SBATCH -A matgen\n",
      "#SBATCH --mem=0\n",
      "#SBATCH -t 00:30:00\n",
      "#SBATCH -N 1\n",
      "#SBATCH -G 4\n",
      "#SBATCH -q debug\n",
      "#SBATCH -C gpu\n",
      "#SBATCH -J eos\n",
      "source ~/.bashrc\n",
      "module load python\n",
      "source activate /pscratch/sd/c/cyrusyc/.conda/mlip-arena\n",
      "/pscratch/sd/c/cyrusyc/.conda/mlip-arena/bin/python -m distributed.cli.dask_worker tcp://128.55.64.49:36289 --name dummy-name --nthreads 1 --memory-limit 59.60GiB --nanny --death-timeout 60\n",
      "\n"
     ]
    }
   ],
   "source": [
    "nodes_per_alloc = 1\n",
    "gpus_per_alloc = 4\n",
    "ntasks = 1\n",
    "\n",
    "cluster_kwargs = {\n",
    "    \"cores\": 1,\n",
    "    \"memory\": \"64 GB\",\n",
    "    \"shebang\": \"#!/bin/bash\",\n",
    "    \"account\": \"matgen\",\n",
    "    \"walltime\": \"00:30:00\",\n",
    "    \"job_mem\": \"0\",\n",
    "    \"job_script_prologue\": [\n",
    "        \"source ~/.bashrc\",\n",
    "        \"module load python\",\n",
    "        \"source activate /pscratch/sd/c/cyrusyc/.conda/mlip-arena\",\n",
    "    ],\n",
    "    \"job_directives_skip\": [\"-n\", \"--cpus-per-task\", \"-J\"],\n",
    "    \"job_extra_directives\": [f\"-N {nodes_per_alloc}\", f\"-G {gpus_per_alloc}\", \"-q debug\", \"-C gpu\", \"-J eos\"],\n",
    "}\n",
    "cluster = SLURMCluster(**cluster_kwargs)\n",
    "\n",
    "print(cluster.job_script())\n",
    "cluster.adapt(minimum_jobs=2, maximum_jobs=2)\n",
    "client = Client(cluster)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from prefect.concurrency.sync import concurrency\n",
    "from prefect.runtime import flow_run, task_run\n",
    "\n",
    "def postprocess(output, model: str, formula: str):\n",
    "    row = {\n",
    "        \"formula\": formula,\n",
    "        \"method\": model,\n",
    "        \"volumes\": output[\"eos\"][\"volumes\"],\n",
    "        \"energies\": output[\"eos\"][\"energies\"],\n",
    "        \"K\": output[\"K\"],\n",
    "    }\n",
    "\n",
    "    fpath = Path(REGISTRY[model][\"family\"]) / f\"{model}.parquet\"\n",
    "\n",
    "    if not fpath.exists():\n",
    "        fpath.parent.mkdir(parents=True, exist_ok=True)\n",
    "        df = pd.DataFrame([row])  # Convert the dictionary to a DataFrame with a list\n",
    "    else:\n",
    "        df = pd.read_parquet(fpath)\n",
    "        new_row = pd.DataFrame([row])  # Convert dictionary to DataFrame with a list\n",
    "        df = pd.concat([df, new_row], ignore_index=True)\n",
    "\n",
    "    df.drop_duplicates(subset=[\"formula\", \"method\"], keep='last', inplace=True)\n",
    "    df.to_parquet(fpath)\n",
    "\n",
    "\n",
    "\n",
    "task_runner = DaskTaskRunner(address=client.scheduler.address)\n",
    "EOS = EOS.with_options(\n",
    "    # task_runner=task_runner, \n",
    "    log_prints=True,\n",
    "    timeout_seconds=120, \n",
    "    # result_storage=None\n",
    ")\n",
    "\n",
    "from prefect import get_client\n",
    "\n",
    "async with get_client() as client:\n",
    "    limit_id = await client.create_concurrency_limit(\n",
    "        tag=\"bottleneck\", \n",
    "        concurrency_limit=2\n",
    "    )\n",
    "\n",
    "def generate_task_run_name():\n",
    "    task_name = task_run.task_name\n",
    "\n",
    "    parameters = task_run.parameters\n",
    "\n",
    "    atoms = parameters[\"atoms\"]\n",
    "    \n",
    "    return f\"{task_name}: {atoms.get_chemical_formula()}\"\n",
    "\n",
    "@task(task_run_name=generate_task_run_name, tags=[\"bottleneck\"], timeout_seconds=150)\n",
    "def fit_one(atoms: Atoms, model: str):\n",
    "    \n",
    "    eos = EOS(\n",
    "        atoms=atoms,\n",
    "        calculator_name=model,\n",
    "        calculator_kwargs={},\n",
    "        device=None,\n",
    "        optimizer=\"QuasiNewton\",\n",
    "        optimizer_kwargs=None,\n",
    "        filter=\"FrechetCell\",\n",
    "        filter_kwargs=None,\n",
    "        criterion=dict(\n",
    "            fmax=0.1,\n",
    "        ),\n",
    "        max_abs_strain=0.1,\n",
    "        npoints=7,\n",
    "    )\n",
    "    if isinstance(eos, dict):\n",
    "        postprocess(output=eos, model=model, formula=atoms.get_chemical_formula())\n",
    "        eos[\"method\"] = model\n",
    "    \n",
    "    return eos\n",
    "    \n",
    "#https://docs-3.prefect.io/3.0/develop/task-runners#use-multiple-task-runners\n",
    "# @flow(task_runner=ThreadPoolTaskRunner(max_workers=50), log_prints=True)\n",
    "@flow(task_runner=task_runner, log_prints=True)\n",
    "def fit_all(atoms_list: list[Atoms]):\n",
    "    \n",
    "    futures = []\n",
    "    for atoms in atoms_list:\n",
    "        futures_per_atoms = []\n",
    "        for model in MLIPEnum:\n",
    "            \n",
    "            # with concurrency(\"bottleneck\", occupy=2):\n",
    "            future = fit_one.submit(atoms, model.name)\n",
    "            # if not futures_per_atoms:\n",
    "            #     if not futures:\n",
    "            #         future = fit_one.submit(atoms, model.name)\n",
    "            #     else:\n",
    "            #         future = fit_one.submit(atoms, model.name, wait_for=[futures[-1]])                    \n",
    "            # else:\n",
    "            #     future = fit_one.submit(atoms, model.name, wait_for=[future])\n",
    "            futures_per_atoms.append(future)\n",
    "            \n",
    "        futures.extend(futures_per_atoms)\n",
    "\n",
    "    return [f.result() for f in futures]\n",
    "\n",
    "\n",
    "# @task(task_run_name=generate_task_run_name, result_storage=None)\n",
    "# def fit_one(atoms: Atoms):\n",
    "    \n",
    "#     outputs = []\n",
    "#     for model in MLIPEnum:\n",
    "#         try:\n",
    "#             eos = EOS(\n",
    "#                 atoms=atoms,\n",
    "#                 calculator_name=model.name,\n",
    "#                 calculator_kwargs={},\n",
    "#                 device=None,\n",
    "#                 optimizer=\"QuasiNewton\",\n",
    "#                 optimizer_kwargs=None,\n",
    "#                 filter=\"FrechetCell\",\n",
    "#                 filter_kwargs=None,\n",
    "#                 criterion=dict(\n",
    "#                     fmax=0.1,\n",
    "#                 ),\n",
    "#                 max_abs_strain=0.1,\n",
    "#                 npoints=7,\n",
    "#             )\n",
    "#             if isinstance(eos, dict):\n",
    "#                 postprocess(output=eos, model=model.name, formula=atoms.get_chemical_formula())\n",
    "#                 eos[\"method\"] = model.name\n",
    "#                 outputs.append(eos)\n",
    "#         except:\n",
    "#             continue\n",
    "    \n",
    "#     return outputs\n",
    "\n",
    "# # https://orion-docs.prefect.io/latest/concepts/task-runners/#using-multiple-task-runners\n",
    "# @flow(task_runner=DaskTaskRunner(address=client.scheduler.address), log_prints=True, result_storage=None)\n",
    "# def fit_all(atoms_list: list[Atoms]):\n",
    "    \n",
    "#     futures = []\n",
    "#     for atoms in atoms_list:\n",
    "#         future = fit_one.submit(atoms)\n",
    "#         futures.append(future)\n",
    "            \n",
    "#     wait(futures)\n",
    "    \n",
    "#     return [f.result(raise_on_failure=False) for f in futures]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">18:53:47.335 | <span style=\"color: #008080; text-decoration-color: #008080\">INFO</span>    | prefect.engine - Created flow run<span style=\"color: #800080; text-decoration-color: #800080\"> 'vengeful-malkoha'</span> for flow<span style=\"color: #800080; text-decoration-color: #800080; font-weight: bold\"> 'fit-all'</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "18:53:47.335 | \u001b[36mINFO\u001b[0m    | prefect.engine - Created flow run\u001b[35m 'vengeful-malkoha'\u001b[0m for flow\u001b[1;35m 'fit-all'\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">18:53:47.341 | <span style=\"color: #008080; text-decoration-color: #008080\">INFO</span>    | prefect.engine - View at <span style=\"color: #0000ff; text-decoration-color: #0000ff\">https://app.prefect.cloud/account/f7d40474-9362-4bfa-8950-ee6a43ec00f3/workspace/d4bb0913-5f5e-49f7-bfc5-06509088baeb/runs/flow-run/909d2bc4-695f-4eeb-8b7c-7660397a0692</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "18:53:47.341 | \u001b[36mINFO\u001b[0m    | prefect.engine - View at \u001b[94mhttps://app.prefect.cloud/account/f7d40474-9362-4bfa-8950-ee6a43ec00f3/workspace/d4bb0913-5f5e-49f7-bfc5-06509088baeb/runs/flow-run/909d2bc4-695f-4eeb-8b7c-7660397a0692\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">18:53:47.654 | <span style=\"color: #008080; text-decoration-color: #008080\">INFO</span>    | prefect.task_runner.dask - Connecting to existing Dask cluster SLURMCluster(df8c3d55, 'tcp://128.55.64.49:36289', workers=0, threads=0, memory=0 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "18:53:47.654 | \u001b[36mINFO\u001b[0m    | prefect.task_runner.dask - Connecting to existing Dask cluster SLURMCluster(df8c3d55, 'tcp://128.55.64.49:36289', workers=0, threads=0, memory=0 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fit_all(atoms_list)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```\n",
    "Note that, because the DaskTaskRunner uses multiprocessing, calls to flows in scripts must be guarded with if __name__ == \"__main__\": or you will encounter warnings and errors.\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# import os\n",
    "# import tempfile\n",
    "# import shutil\n",
    "# from contextlib import contextmanager\n",
    "\n",
    "# @contextmanager\n",
    "# def twd():\n",
    "    \n",
    "#     pwd = os.getcwd()\n",
    "#     temp_dir = tempfile.mkdtemp()\n",
    "    \n",
    "#     try:\n",
    "#         os.chdir(temp_dir)\n",
    "#         yield\n",
    "#     finally:\n",
    "#         os.chdir(pwd)\n",
    "#         shutil.rmtree(temp_dir)\n",
    "\n",
    "# with twd():\n",
    "\n",
    "# fit_all(atoms_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "df = pd.read_parquet('mace-mp/MACE-MP(M).parquet')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>formula</th>\n",
       "      <th>method</th>\n",
       "      <th>volumes</th>\n",
       "      <th>energies</th>\n",
       "      <th>K</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Ac2O3</td>\n",
       "      <td>MACE-MP(M)</td>\n",
       "      <td>[82.36010147441682, 85.41047560309894, 88.4608...</td>\n",
       "      <td>[-39.47541427612305, -39.65580749511719, -39.7...</td>\n",
       "      <td>95.755459</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Ac6In2</td>\n",
       "      <td>MACE-MP(M)</td>\n",
       "      <td>[278.3036976131417, 288.61124196918433, 298.91...</td>\n",
       "      <td>[-31.21324348449707, -31.40914535522461, -31.5...</td>\n",
       "      <td>33.370214</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Ac6Tl2</td>\n",
       "      <td>MACE-MP(M)</td>\n",
       "      <td>[278.30267000598286, 288.6101763025008, 298.91...</td>\n",
       "      <td>[-29.572534561157227, -29.833026885986328, -30...</td>\n",
       "      <td>29.065081</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Ac3Sn</td>\n",
       "      <td>MACE-MP(M)</td>\n",
       "      <td>[135.293532345587, 140.30440391394214, 145.315...</td>\n",
       "      <td>[-17.135194778442383, -17.228239059448242, -17...</td>\n",
       "      <td>30.622045</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>AcAg</td>\n",
       "      <td>MACE-MP(M)</td>\n",
       "      <td>[55.376437498321394, 57.4274166649259, 59.4783...</td>\n",
       "      <td>[-7.274301528930664, -7.346108913421631, -7.39...</td>\n",
       "      <td>40.212164</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Ac4</td>\n",
       "      <td>MACE-MP(M)</td>\n",
       "      <td>[166.09086069175856, 172.2423740507126, 178.39...</td>\n",
       "      <td>[-16.326059341430664, -16.406923294067383, -16...</td>\n",
       "      <td>25.409891</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Ac16S24</td>\n",
       "      <td>MACE-MP(M)</td>\n",
       "      <td>[1006.5670668063424, 1043.84732853991, 1081.12...</td>\n",
       "      <td>[-249.4179229736328, -250.7970733642578, -251....</td>\n",
       "      <td>61.734158</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   formula      method                                            volumes  \\\n",
       "1    Ac2O3  MACE-MP(M)  [82.36010147441682, 85.41047560309894, 88.4608...   \n",
       "2   Ac6In2  MACE-MP(M)  [278.3036976131417, 288.61124196918433, 298.91...   \n",
       "3   Ac6Tl2  MACE-MP(M)  [278.30267000598286, 288.6101763025008, 298.91...   \n",
       "4    Ac3Sn  MACE-MP(M)  [135.293532345587, 140.30440391394214, 145.315...   \n",
       "5     AcAg  MACE-MP(M)  [55.376437498321394, 57.4274166649259, 59.4783...   \n",
       "6      Ac4  MACE-MP(M)  [166.09086069175856, 172.2423740507126, 178.39...   \n",
       "7  Ac16S24  MACE-MP(M)  [1006.5670668063424, 1043.84732853991, 1081.12...   \n",
       "\n",
       "                                            energies          K  \n",
       "1  [-39.47541427612305, -39.65580749511719, -39.7...  95.755459  \n",
       "2  [-31.21324348449707, -31.40914535522461, -31.5...  33.370214  \n",
       "3  [-29.572534561157227, -29.833026885986328, -30...  29.065081  \n",
       "4  [-17.135194778442383, -17.228239059448242, -17...  30.622045  \n",
       "5  [-7.274301528930664, -7.346108913421631, -7.39...  40.212164  \n",
       "6  [-16.326059341430664, -16.406923294067383, -16...  25.409891  \n",
       "7  [-249.4179229736328, -250.7970733642578, -251....  61.734158  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "mlip-arena",
   "language": "python",
   "name": "mlip-arena"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.8"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {},
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}