make_batch_generation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n"
     ]
    }
   ],
   "source": [
    "import mkl\n",
    "mkl.set_num_threads(1)\n",
    "print mkl.get_max_threads()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import numpy.random as npr\n",
    "import spglib as spg\n",
    "import ase\n",
    "from ase.io import read,write\n",
    "from ase.spacegroup import crystal\n",
    "from ase.visualize import view\n",
    "import cPickle as pck\n",
    "import pandas as pd\n",
    "from tqdm import tqdm_notebook\n",
    "import os,sys,psutil,signal\n",
    "from quippy.potential import Potential, Minim\n",
    "from ase.optimize import  FIRE\n",
    "from ase.constraints import UnitCellFilter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/musil/miniconda/envs/glosim/lib/python2.7/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "from generate_and_relax_structures import generate_crystal_step_1,generate_crystal_step_1_wrapper\n",
    "from libs.utils import unskewCell,ase2qp,qp2ase,get_standard_frame,isTooClose,isLayered\n",
    "from libs.input_structure import input2crystal,getCellParam\n",
    "from libs.LJ_pressure import (make_LJ_input,LJ_vcrelax,LJ_vcrelax_alternative,vc_relax_qp,AtomSeparator,\n",
    "                              vc_relax_ase,get_LJ_parameters,make_LJ_input)\n",
    "from libs.raw_data import z2symb,z2VdWradius,z2Covalentradius,SG2BravaisLattice,WyckTable,z2epsilon"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.insert(0,'/local/git/glosim2/')\n",
    "from libmatch.soap import get_Soaps\n",
    "from libmatch.utils import get_soapSize,get_spkit,get_spkitMax,ase2qp,qp2ase\n",
    "from libmatch.chemical_kernel import deltaKernel,PartialKernels\n",
    "from GlobalSimilarity import get_environmentalKernels,get_globalKernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "d2r = np.pi/180."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# enthalpy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "import sympy as sp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "sp.init_printing() "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "$H = U + pV$\n",
    "\n",
    "perfect gaz -> $pV=cst$\n",
    "\n",
    "$\\Delta H_f = \\int dU = \\int_{\\infty}^{r_m} F_{LJ}(r)dr = - \\int_{\\infty}^{r_m} \\frac{V_{LJ}(r)}{dr}dr  $\n",
    "\n",
    "$\\Delta H_f = - (V_{LJ}(r_m) - V_{LJ}(\\infty)) = - V_{LJ}(r_m) = -\\epsilon\\left(1^{12}-2*1^{6}\\right) =   \\epsilon$\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{1: 2.2593901062472397, 2: 0.000470937100895, 3: 1.6510281925760115, 4: 3.3580234425274806, 5: 5.855812484654403, 6: 7.429834151989738, 7: 4.896495664159516, 8: 2.58307661900519, 9: 0.8222493512744372, 10: 0.0031738489227930004, 11: 1.1141590125669882, 12: 1.5245840999870135, 13: 3.4295369047294546, 14: 4.663921447954834, 15: 3.2802914183949, 16: 2.872664683843648, 17: 1.2572066655107053, 18: 0.01006590352343, 19: 0.9224200197066228, 20: 1.8427671854363767, 21: 3.915621162305192, 22: 4.9022996552947475, 23: 5.342781125379371, 24: 4.119589993629083, 25: 2.936197658234677, 26: 4.306354136944964, 27: 4.422433959649617, 28: 4.457672477256387, 29: 3.496904658977691, 30: 1.3515007929184675, 31: 2.8186668377462145, 32: 3.8555083969759965, 33: 3.135191640014083, 34: 2.354762117722974, 35: 1.1592435865638848, 36: 0.01418067874168, 37: 0.8384694336434358, 38: 1.6997402610324284, 39: 4.401705419880929, 40: 6.322204629449886, 41: 7.597009825224208, 42: 6.82984656838506, 43: 7.026974981585283, 44: 6.742993986754256, 45: 5.762534055695307, 46: 3.903184038443979, 47: 2.952780490049627, 48: 1.1587253730696676, 49: 2.5185175818956105, 50: 3.1217180891644354, 51: 2.7403129574205733, 52: 2.0376154592620455, 53: 1.1064583600429205, 54: 0.01880129124854, 55: 0.7928666461523218, 56: 1.856240736286024, 57: 4.467000320152297, 58: 4.354029778412946, 59: 3.6990079217224006, 60: 3.388079825192078, 62: 2.1422945850939206, 63: 1.838621477482639, 64: 4.11979727902677, 65: 4.028591704044542, 66: 3.0097839744135193, 67: 3.1154995272338293, 68: 3.2792549914064653, 69: 2.4065834671446944, 70: 1.6126803940039383, 71: 4.431761802545527, 72: 6.409264496478376, 73: 8.104859049557067, 74: 8.819993671576809, 75: 8.021944890482315, 76: 8.156680398978787, 77: 6.933696552626187, 78: 5.8630674735734445, 79: 3.8161241714154888, 80: 0.6361588855010394, 81: 1.8883699729274905, 82: 2.0231054814239635, 83: 2.1723509677585184}\n"
     ]
    }
   ],
   "source": [
    "print z2epsilon"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Func"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "code_folding": [
     0
    ]
   },
   "outputs": [],
   "source": [
    "def compare(frame1,frame2):\n",
    "    centerweight  = 1.\n",
    "    gaussian_width = 0.1\n",
    "    cutoff = 3.5\n",
    "    cutoff_transition_width = 0.5\n",
    "    nmax = 8\n",
    "    lmax = 6\n",
    "    nocenters = []\n",
    "    is_fast_average = False\n",
    "\n",
    "    soap_params = {\n",
    "              'centerweight': centerweight, \n",
    "              'gaussian_width': gaussian_width,'cutoff': cutoff, \n",
    "              'cutoff_transition_width': cutoff_transition_width,\n",
    "              'nmax': nmax, 'lmax': lmax, 'is_fast_average':is_fast_average,\n",
    "              'chem_channels': True ,'nocenters': nocenters,'dispbar':True,\n",
    "                   }\n",
    "    ff = []\n",
    "    for frame in [frame1,frame2]:\n",
    "        ff.append(ase2qp(frame))\n",
    "     \n",
    "    envk = get_environmentalKernels(ff,nthreads=1, nprocess=1, nchunks=1,**soap_params)\n",
    "    gkern = get_globalKernel(envk, kernel_type='average', zeta=4, gamma=1.0, eps=1e-06, \n",
    "                             nthreads=8, normalize_global_kernel=True)\n",
    "    \n",
    "    return gkern[0,1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "code_folding": [
     0
    ]
   },
   "outputs": [],
   "source": [
    "def get_Nsoap(spkitMax,nmax,lmax):\n",
    "    Nsoap = 0\n",
    "    for sp1 in spkitMax:\n",
    "        for sp2 in spkitMax:\n",
    "            if sp1 == sp2:\n",
    "                Nsoap += nmax*(nmax+1)*(lmax+1) / 2 \n",
    "            elif sp1 > sp2:\n",
    "                Nsoap += nmax**2*(lmax+1)\n",
    "    return Nsoap + 1 \n",
    "def get_fingerprints(frames,soap_params,nprocess):\n",
    "    fings = get_Soaps(frames, nprocess=nprocess, **soap_params)\n",
    "    N = len(frames)\n",
    "    Nsoap = get_Nsoap(get_spkitMax(frames),soap_params['nmax'],soap_params['lmax'])\n",
    "    soaps = np.zeros((N,Nsoap))\n",
    "    ii = 0\n",
    "    for iframe,fing in enumerate(fings):\n",
    "        soaps[iframe] = fing['AVG']\n",
    "        \n",
    "    return soaps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "code_folding": [
     0,
     7,
     64,
     73,
     80
    ]
   },
   "outputs": [],
   "source": [
    "def s2hms(time):\n",
    "    m = time // 60\n",
    "    s = int(time % 60)\n",
    "    h = int(m // 60)\n",
    "    m = int(m % 60)\n",
    "    return '{:02d}:{:02d}:{:02d} (h:m:s)'.format(h,m,s)\n",
    "\n",
    "def fpsSelection(data=None,distance_func=None,threshold=3e-3,Nmin=0,Nmax=20,seed=None):\n",
    "    import numpy.random as npr\n",
    "    from tqdm import tqdm_notebook\n",
    "    nbOfFrames,Nfeature = data.shape \n",
    "    \n",
    "    if seed is None:\n",
    "        isel = 0\n",
    "    else:\n",
    "        npr.seed(seed)\n",
    "        isel = npr.randint(0,nbOfFrames)    \n",
    "    \n",
    "    ldist = 1e100*np.ones(nbOfFrames,float)\n",
    "    dsel = np.zeros(nbOfFrames,float)\n",
    "    idx_to_compute = np.ones(nbOfFrames,bool)\n",
    "    idx_to_ignore = np.zeros(nbOfFrames,bool)\n",
    "    \n",
    "    LandmarksIdx = []\n",
    "    \n",
    "    \n",
    "    nsel = 0\n",
    "    Nidx = 0\n",
    "    cond = True\n",
    "    pbar = tqdm_notebook(total=nbOfFrames)\n",
    "    while cond:\n",
    "        LandmarksIdx.append(isel)\n",
    "        Nidx += 1\n",
    "        idx_to_compute[isel] = False\n",
    "        idx_to_ignore[isel] = True\n",
    "        dsel.fill(0.)\n",
    "        \n",
    "        imax = 0       \n",
    "        distLine = distance_func(data[isel,:].reshape((1,Nfeature)),\n",
    "                                 data[idx_to_compute,:].reshape((-1,Nfeature)))\n",
    "        \n",
    "        dsel[idx_to_compute] = distLine.reshape((-1,))\n",
    "        \n",
    "        low = (dsel < ldist) * idx_to_compute\n",
    "        ldist[low] = dsel[low]\n",
    "        \n",
    "        ldist[idx_to_ignore] = 0.\n",
    "        isel = ldist.argmax()\n",
    "        \n",
    "        ids = (dsel < threshold) * idx_to_compute\n",
    "        \n",
    "        idx_to_compute[ids] = False\n",
    "        idx_to_ignore[ids] = True\n",
    "        \n",
    "        if Nidx >= Nmin :\n",
    "            if ldist[isel] < threshold:\n",
    "                cond = False\n",
    "            elif Nidx >= Nmax:\n",
    "                cond = False\n",
    "        pbar.update()\n",
    "    pbar.close()\n",
    "        \n",
    "    return LandmarksIdx\n",
    "\n",
    "def distance_func(XA,XB):\n",
    "    #mkl.set_num_threads(10)\n",
    "    kernel = np.dot(XB,XA.T)\n",
    "    # the feature vectors are normalized\n",
    "    dd = 2 - 2 * kernel\n",
    "    dd[dd<0.] = 0.\n",
    "    distance = np.sqrt(dd)\n",
    "    return distance\n",
    "\n",
    "def distance_func2(kernel):\n",
    "    # the feature vectors are normalized\n",
    "    dd = 2 - 2 * kernel\n",
    "    dd[dd<0.] = 0.\n",
    "    distance = np.sqrt(dd)\n",
    "    return distance\n",
    "\n",
    "def fpsSelection_with_restart(data=None,distance_func=None,restart_ref=None,disable_pbar=False,nthread=10,\n",
    "                              intermediate_copy=True,stride=100,threshold=3e-3,Nmin=0,Nmax=20,seed=None,fn=None):\n",
    "    import numpy.random as npr\n",
    "    import cPickle as pck\n",
    "    from tqdm import tqdm_notebook\n",
    "    try:\n",
    "        import mkl\n",
    "        mkl.set_num_threads(nthread)\n",
    "    except:\n",
    "        pass\n",
    "    \n",
    "    if fn is None:\n",
    "        fn = 'restart_ref_thr{}.pck'.format(threshold)\n",
    "    \n",
    "    nbOfFrames,Nfeature = data.shape \n",
    "    \n",
    "    if nbOfFrames == Nfeature:\n",
    "        iskernel = True\n",
    "    else:\n",
    "        iskernel = False\n",
    "    \n",
    "    if seed is None:\n",
    "        isel = 0\n",
    "    else:\n",
    "        npr.seed(seed)\n",
    "        isel = npr.randint(0,nbOfFrames)    \n",
    "    \n",
    "    dsel = np.zeros(nbOfFrames,float)\n",
    "    idx_to_compute = np.ones(nbOfFrames,bool)\n",
    "    idx_to_ignore = np.zeros(nbOfFrames,bool)\n",
    "    \n",
    "    if restart_ref is None:\n",
    "        ldist = 1e100*np.ones(nbOfFrames,float)\n",
    "        LandmarksIdx = []\n",
    "        minmax = []\n",
    "    else:\n",
    "        LandmarksIdx = restart_ref['LandmarksIdx'][:-1]\n",
    "        ldist = restart_ref['ldist']\n",
    "        isel = restart_ref['LandmarksIdx'][-1]\n",
    "        minmax = restart_ref['minmax']\n",
    "        idx_to_compute[LandmarksIdx] = False\n",
    "        idx_to_ignore[LandmarksIdx] = True\n",
    "        \n",
    "    Nidx = len(LandmarksIdx)\n",
    "    cond = True\n",
    "    pbar = tqdm_notebook(total=Nmax,disable=disable_pbar)\n",
    "    \n",
    "    while cond:\n",
    "        \n",
    "        LandmarksIdx.append(isel)\n",
    "        \n",
    "        if Nidx % (stride-1) == 0:\n",
    "            with open(fn,'wb') as f:\n",
    "                pck.dump({'LandmarksIdx':LandmarksIdx,'ldist':ldist,'minmax':minmax},f,protocol=pck.HIGHEST_PROTOCOL)\n",
    "        \n",
    "        Nidx += 1\n",
    "        idx_to_compute[isel] = False\n",
    "        idx_to_ignore[isel] = True\n",
    "        dsel.fill(0.)\n",
    "        \n",
    "        imax = 0       \n",
    "        if iskernel:\n",
    "            distLine = distance_func(data[isel,idx_to_compute])\n",
    "            dsel[idx_to_compute] = distLine\n",
    "        else:\n",
    "            if intermediate_copy:\n",
    "                distLine = distance_func(data[isel,:].reshape((1,Nfeature)),\n",
    "                                         data[idx_to_compute,:].reshape((-1,Nfeature)))\n",
    "                dsel[idx_to_compute] = distLine.reshape((-1,))\n",
    "            else:\n",
    "                distLine = distance_func(data[isel,:].reshape((1,Nfeature)),data)\n",
    "                dsel = distLine.reshape((-1,))\n",
    "        \n",
    "        \n",
    "        low = (dsel < ldist) * idx_to_compute\n",
    "        ldist[low] = dsel[low]\n",
    "        \n",
    "        ldist[idx_to_ignore] = 0.\n",
    "        isel = ldist.argmax()\n",
    "        minmax.append(ldist[isel])\n",
    "        ids = (dsel < threshold) * idx_to_compute\n",
    "        \n",
    "        idx_to_compute[ids] = False\n",
    "        idx_to_ignore[ids] = True\n",
    "        \n",
    "        if Nidx >= Nmin :\n",
    "            if ldist[isel] < threshold:\n",
    "                cond = False\n",
    "            elif Nidx >= Nmax:\n",
    "                cond = False\n",
    "        \n",
    "        pbar.update()\n",
    "    pbar.close()\n",
    "    return LandmarksIdx,minmax"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('./structures/structures_downsampled.pck','rb') as f:\n",
    "    crystals = pck.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# view(crystals)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "## downsample fps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "with open('./structures/structures_141117.pck','rb') as f:\n",
    "    crystals = pck.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "view(crystals)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e77647692e61418da31bce8eaf2d5cff",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description=u'SOAP vectors', max=855), HTML(value=u'')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "soap_params = dict(nmax= 9, cutoff= 4, gaussian_width= 0.4, lmax= 9,\n",
    "        centerweight = 1.,cutoff_transition_width = 0.5,\n",
    "        nocenters = [],is_fast_average=True,chem_channels=False,dispbar=False\n",
    "                      )\n",
    "nprocess = 4\n",
    "\n",
    "fings = get_fingerprints([ase2qp(crystal) for crystal in crystals],soap_params,nprocess)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "kernel = np.dot(fings,fings.T)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f03c3d5c21754775a5380f3feb13f88e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=855), HTML(value=u'')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "fps_ids, minmax = fpsSelection_with_restart(data=kernel,distance_func=distance_func2,restart_ref=None,disable_pbar=False,nthread=10,\n",
    "                              intermediate_copy=True,stride=10000,threshold=1e-9,Nmin=0,Nmax=kernel.shape[0],seed=None,\n",
    "                                           fn=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f0884fd6e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(minmax[:220],'ob',ms=1)\n",
    "plt.yscale('log')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "ref_cc = [crystals[it] for it in fps_ids[:220]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "with open('./structures/structures_downsampled.pck','wb') as f:\n",
    "    pck.dump(ref_cc,f,protocol=pck.HIGHEST_PROTOCOL)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Make the batch crystal generation procedure"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "## tests on some MP structures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "seed = 54565\n",
    "np.random.seed(seed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "202 ['i']\n"
     ]
    }
   ],
   "source": [
    "vdw_ratio = 1.5\n",
    "sites_z = [14]\n",
    "sg = 227 # 202\n",
    "wycks = [8] # [0]\n",
    "sg = 202\n",
    "wycks = [0]\n",
    "cc, sg, wyckoff_letters = input2crystal(sites_z ,seed ,vdw_ratio, sg=sg,wycks=wycks)\n",
    "view(cc)\n",
    "print sg,wyckoff_letters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true,
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "202 ['i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i'] [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n"
     ]
    }
   ],
   "source": [
    "sym_data = spg.get_symmetry_dataset(cc)\n",
    "print sym_data['number'],sym_data['wyckoffs'],sym_data['equivalent_atoms']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true,
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "202 ['i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'i', 'h', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h', 'i', 'i', 'i', 'i', 'h', 'h', 'i', 'i', 'i', 'i', 'i', 'i', 'h', 'i', 'h'] [0 0 2 0 0 2 0 2 8 2 0 2 8 0 0 0 2 0 0 0 2 0 8 2 2 0 0 8 2 0 2 2 2 8 2 0 8\n",
      " 2 2 2 0 2 8 2 8 2 0 0 2 8 8 0 0 2 2 0 2 8 0 8 0 0 2 0 0 2 0 2 8 2 0 2 8 0\n",
      " 0 0 2 0 0 0 2 0 8 2 2 0 0 8 2 0 2 2 2 8 2 0 8 2 2 2 0 2 8 2 8 2 0 0 2 8 8\n",
      " 0 0 2 2 0 2 8 0 8 0 0 2 0 0 2 0 2 8 2 0 2 8 0 0 0 2 0 0 0 2 0 8 2 2 0 0 8\n",
      " 2 0 2 2 2 8 2 0 8 2 2 2 0 2 8 2 8 2 0 0 2 8 8 0 0 2 2 0 2 8 0 8 0 0 2 0 0\n",
      " 2 0 2 8 2 0 2 8 0 0 0 2 0 0 0 2 0 8 2 2 0 0 8 2 0 2 2 2 8 2 0 8 2 2 2 0 2\n",
      " 8 2 8 2 0 0 2 8 8 0 0 2 2 0 2 8 0 8]\n"
     ]
    }
   ],
   "source": [
    "dd = read('/home/musil/Downloads/'+'C_mp-667273_conventional_standard.cif')\n",
    "sym_data = spg.get_symmetry_dataset(dd)\n",
    "print sym_data['number'],sym_data['wyckoffs'],sym_data['equivalent_atoms']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[8]\n",
      "227 ['a']\n"
     ]
    }
   ],
   "source": [
    "vdw_ratio = 10\n",
    "sites_z = [14]\n",
    "sg = 227 # 202\n",
    "wycks = [8] # [0]\n",
    "cc, sg, wyckoff_letters = input2crystal(sites_z ,seed ,vdw_ratio, sg=sg,wycks=wycks)\n",
    "view(cc)\n",
    "print sg,wyckoff_letters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12 ['i', 'i', 'i', 'i'] [0 0 0 0]\n"
     ]
    }
   ],
   "source": [
    "dd = read('/home/musil/Downloads/'+'C_mp-632329_symmetrized.cif')\n",
    "sym_data = spg.get_symmetry_dataset(dd)\n",
    "print sym_data['number'],sym_data['wyckoffs'],sym_data['equivalent_atoms']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12 ['i']\n"
     ]
    }
   ],
   "source": [
    "vdw_ratio = 1.5\n",
    "sites_z = [14]\n",
    "sg = 12 # 202\n",
    "wycks = [1] # [0]\n",
    "cc, sg, wyckoff_letters = input2crystal(sites_z ,vdw_ratio= vdw_ratio, sg=sg,wycks=wycks)\n",
    "view(cc)\n",
    "print sg,wyckoff_letters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "cc = new_crystals[55]\n",
    "view(cc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "from quippy.potential import Potential\n",
    "dd = ase2qp(cc)\n",
    "\n",
    "LJ_parameters = get_LJ_parameters(dd)\n",
    "max_cutoff = LJ_parameters['cutoffs'].max()\n",
    "param_str = make_LJ_input(dd, LJ_parameters)\n",
    "pot = Potential('IP LJ', param_str=param_str)\n",
    "dd.set_calculator(pot)\n",
    "dd.set_cutoff(max_cutoff, 0.5)\n",
    "\n",
    "sep = AtomSeparator(dd)\n",
    "sep.run(Nmax=20)\n",
    "\n",
    "# ee = vc_relax_qp(dd,isotropic_external_pressure=1e-2, fmax=5e-3, steps=1e4)\n",
    "ee = vc_relax_ase(dd, isotropic_external_pressure=1e-2,\n",
    "                         fmax=5e-3, steps=5e5)\n",
    "view(unskewCell(ee))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "ee = LJ_vcrelax_alternative(cc,isotropic_external_pressure=1e-2,debug=False)\n",
    "view(ee)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1 1 2]\n"
     ]
    }
   ],
   "source": [
    "aspect_ratio = 0.75\n",
    "length = ee.get_cell_lengths_and_angles()[:3]\n",
    "\n",
    "rr = np.ones(3, int)\n",
    "\n",
    "for it in range(3):\n",
    "    mask = length / length[it] < aspect_ratio\n",
    "    if np.any(mask):\n",
    "        rr[it] = 2\n",
    "print rr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "view(dd)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## run"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from libs.LJ_pressure import vc_relax_ase,vc_relax_qp,get_LJ_parameters,make_LJ_input,AtomSeparator\n",
    "from libs.utils import stdchannel_to_null"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(10)\n",
    "sites_z = [14]\n",
    "crystal, sg, wyckoff_letters = input2crystal(sites_z)\n",
    "    \n",
    "crystal = unskewCell(crystal)\n",
    "dd = ase2qp(crystal)\n",
    "\n",
    "# get the string to setup the quippy LJ potential (parameters and species)\n",
    "LJ_parameters = get_LJ_parameters(dd)\n",
    "max_cutoff = max(LJ_parameters['cutoffs'].values())*1.1\n",
    "thr = min(LJ_parameters['cutoffs'].values())\n",
    "param_str = make_LJ_input(dd, LJ_parameters)\n",
    "\n",
    "pot = Potential('IP LJ', param_str=param_str)\n",
    "\n",
    "sites_z = np.unique(crystal.get_atomic_numbers())\n",
    "thr = np.min([z2Covalentradius[z] for z in sites_z])\n",
    "\n",
    "dd.set_calculator(pot)\n",
    "dd.set_cutoff(max_cutoff, 0.5)\n",
    "\n",
    "sep = AtomSeparator(dd)\n",
    "sep.run(Nmax=20)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "vc_relax_ase(dd, fmax=5e-1, steps=1e4)\n",
    "\n",
    "vc_relax_ase(dd, isotropic_external_pressure=20,\n",
    "                             relax_positions=False,fmax=5e-5, steps=1e3)\n",
    "view(dd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6\n"
     ]
    }
   ],
   "source": [
    "for iii in range(100):\n",
    "    if isLayered(dd,cutoff=thr*1.5, aspect_ratio=0.75):\n",
    "        vc_relax_ase(dd, fmax=5e-1, steps=1e4)\n",
    "\n",
    "        vc_relax_ase(dd, isotropic_external_pressure=20,\n",
    "                     relax_positions=False,fmax=5e-5, steps=1e3)\n",
    "    else:\n",
    "        break\n",
    "print iii\n",
    "view(dd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Atoms object at 0x7fadc31cdd50 fpointer=(1789871280, 22080)>"
      ]
     },
     "execution_count": 121,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vc_relax_ase(dd, fmax=5e-4, steps=1e4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[13.61955182 13.61955182 13.61955182 90.         90.         90.        ]\n"
     ]
    }
   ],
   "source": [
    "print dd.get_cell_lengths_and_angles()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [],
   "source": [
    "view(dd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(229, ['l'])"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sg, wyckoff_letters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "229 ['l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l'] [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n"
     ]
    }
   ],
   "source": [
    "data = spg.get_symmetry_dataset(dd)\n",
    "print data['number'],data['wyckoffs'],data['equivalent_atoms']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_crystal(sites_z):\n",
    "    crystal, sg, wyckoff_letters = input2crystal(sites_z)\n",
    "    \n",
    "    crystal = unskewCell(crystal)\n",
    "    \n",
    "    crystal = LJ_vcrelax_alternative(crystal, isotropic_external_pressure=20, debug=True)\n",
    "    \n",
    "#     thr = np.min([z2Covalentradius[z] for z in sites_z])\n",
    "\n",
    "#     if isLayered(crystal,cutoff=thr*1.5, aspect_ratio=0.75):\n",
    "#         crystal = LJ_vcrelax_alternative(crystal, isotropic_external_pressure=2, debug=True)\n",
    "    \n",
    "    return crystal\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(10)\n",
    "cc = generate_crystal([14])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[13.53493621 13.53493621 13.53493621 90.         90.         90.        ]\n"
     ]
    }
   ],
   "source": [
    "print cc.get_cell_lengths_and_angles()\n",
    "view(cc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "229 ['l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l', 'l'] [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n"
     ]
    }
   ],
   "source": [
    "data = spg.get_symmetry_dataset(cc)\n",
    "print data['number'],data['wyckoffs'],data['equivalent_atoms']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "code_folding": [
     10,
     22,
     51
    ]
   },
   "outputs": [],
   "source": [
    "import multiprocessing as mp\n",
    "\n",
    "def func_wrapper(inp):\n",
    "    #np.random.rand()*\n",
    "    current = mp.current_process()\n",
    "    name = int(current._identity[0])\n",
    "    \n",
    "    return generate_crystal(**inp)\n",
    "        \n",
    "class mp_gen(object):\n",
    "    def __init__(self, chunks, nprocess,dispbar=False):\n",
    "        super(mp_gen, self).__init__()\n",
    "        \n",
    "            \n",
    "        self.func_wrap = func_wrapper\n",
    "        self.dispbar = dispbar\n",
    "        self.parent_id = os.getpid()\n",
    "        self.nprocess = nprocess\n",
    "\n",
    "        \n",
    "        self.chunks = chunks\n",
    "\n",
    "    def run(self):\n",
    "        Nit = len(self.chunks)\n",
    "        pbar = tqdm_notebook(total=Nit,desc='',disable=self.dispbar)\n",
    "        results = []\n",
    "        if self.nprocess > 1:\n",
    "            pool = mp.Pool(self.nprocess, initializer=self.worker_init,\n",
    "                                maxtasksperchild=10)\n",
    "\n",
    "            for res in pool.imap_unordered(self.func_wrap, self.chunks):\n",
    "                results.append(res)\n",
    "                pbar.update()\n",
    "\n",
    "            pool.close()\n",
    "            pool.join()\n",
    "\n",
    "        elif self.nprocess == 1:\n",
    "            for chunk in self.chunks:\n",
    "                res = self.func_wrap(chunk)\n",
    "                results.append(res)\n",
    "\n",
    "                pbar.update()\n",
    "        else:\n",
    "            print 'Nproces: ',self.nprocess\n",
    "            raise NotImplementedError('need at least 1 process')\n",
    "\n",
    "        pbar.close()\n",
    "\n",
    "        return results\n",
    "    # clean kill of the pool in interactive sessions\n",
    "    def worker_init(self):\n",
    "        \n",
    "        def sig_int(signal_num, frame):\n",
    "            print('signal: %s' % signal_num)\n",
    "            parent = psutil.Process(self.parent_id)\n",
    "            for child in parent.children():\n",
    "                if child.pid != os.getpid():\n",
    "                    print(\"killing child: %s\" % child.pid)\n",
    "                    child.kill()\n",
    "            print(\"killing parent: %s\" % self.parent_id)\n",
    "            parent.kill()\n",
    "            print(\"suicide: %s\" % os.getpid())\n",
    "            psutil.Process(os.getpid()).kill()\n",
    "\n",
    "        signal.signal(signal.SIGINT, sig_int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f68c8846a1f841f89dabc8e209bd3936",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=10), HTML(value=u'')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sites_z = [14]\n",
    "nprocess = 2\n",
    "seed = 10\n",
    "np.random.seed(seed)\n",
    "pool = mp_gen([dict(sites_z=sites_z) for it in range(10)],nprocess=1,dispbar=False)\n",
    "\n",
    "new_cc = pool.run()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "view(new_cc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "view(new_cc[6])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "cc = new_cc[6]\n",
    "ncc = LJ_vcrelax_alternative(cc,isotropic_external_pressure=20,debug=False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "view(ncc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1 1 1]\n",
      "[1, 2, 3]\n",
      "[2, 3]\n",
      "[3]\n",
      "[]\n",
      "[]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "False"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "thr = np.min([z2Covalentradius[z] for z in sites_z])\n",
    "isLayered(ncc, cutoff=thr * 1.5, aspect_ratio=0.75,debug=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[False False False] False\n",
      "[False False False] False\n",
      "[False False False] True\n"
     ]
    }
   ],
   "source": [
    "aspect_ratio=0.75\n",
    "crystal = ncc\n",
    "length = crystal.get_cell_lengths_and_angles()[:3]\n",
    "scaled = crystal.get_scaled_positions()\n",
    "rr = np.ones(3, int)\n",
    "\n",
    "for it in range(3):\n",
    "    min,max = np.min(scaled[:,it]),np.max(scaled[:,it])\n",
    "    mask2 = (max-min) < (1-aspect_ratio)\n",
    "    mask = length / length[it] < aspect_ratio\n",
    "    print mask,mask2\n",
    "    if np.any(mask) or mask2:\n",
    "        rr[it] = 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9999999999999175\n"
     ]
    }
   ],
   "source": [
    "print  np.max(scaled[:,0])-np.min(scaled[:,0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_density(crystal):\n",
    "    Na = crystal.get_number_of_atoms()\n",
    "    V = crystal.get_volume()\n",
    "    dens = float(Na)/V\n",
    "    return dens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f0b85a3fc10>]"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAGsdJREFUeJzt3X/sXXV9x/Hne0Vw4kSlXxftj7WOmlijQ/qlmsy5TAIWkvF1GcZCjF1DUpPZZG4Kw7kJZZppyYYuMidDKrDF4tjMmsjGDCxbYtC1VUQrQUqH8qVEvqUMR43Wwnt/nFN7ub33e8+559fncz6vR/LN9/44997P/dzzeZ/PeX8+5xxzd0REJA2/0HUBRESkPQr6IiIJUdAXEUmIgr6ISEIU9EVEEqKgLyKSEAV9EZGEKOiLiCREQV9EJCGndF2AYUuXLvVVq1Z1XQwRkajs3bv3kLvPTFouuKC/atUq9uzZ03UxRESiYmbfL7Kc0jsiIglR0BcRSYiCvohIQhT0RUQSoqAvIpIQBX0RkYQo6IuIJKSXQf/QIbjuuux/Hz8v9HL0nepZYtbLoL9jB1x5Zfa/j58Xejn6TvUsMSt0RK6ZbQA+BSwBbnL3jw89/1bgk8AbgI3ufkf++NnAZ4CXAM8CH3P32+sr/mibNz//f98+L/Ry9J3qWWJm7r74AmZLgO8B5wPzwG7gUnf/7sAyq8gC+weBXQNB/zWAu/tDZvYqYC/wWnf/33GfNzs76zoNg4hIOWa2191nJy1XJL2zHtjv7gfc/SiwE5gbXMDdH3H3+4Hnhh7/nrs/lN8+CDwBTDwhkIj0QxfjH3V/Zt/GcIoE/WXAowP35/PHSjGz9cCpwMNlXysSk74FiSq6GP+o+zP7NoZTJKdvIx5bPCc0/AZmrwRuAza5+3Mjnt8CbAFYuXJlmbeWIYcOZSvn5s2wdGnXpUnT8SABcMUV3Zala2XHP+pYf+secynyfjG1uyI9/XlgxcD95cDBoh9gZi8Bvgz8qbt/bdQy7n6ju8+6++zMjLI/VfStVxKjzZth+3YN9EIWAK+4onggrGP9LfuZRd8Pxu/BxdTuivT0dwNrzGw18BiwEbisyJub2anAl4Bb3f0fpy6lFKaZJd0bDBIhC7F3GvL6u9ge3PHyXnxxtmEIqU5P4u4T/4CLyGbwPAx8OH/sWuDi/Pa5ZHsER4AngX354+8GfgbcN/B39mKftW7dOheR5m3f7g7Zf5lsYSGrq4WF8ct0WafAHi8QzydO2WybpmyKtCPEnn7suqzTolM2FfRFRHqgznn6IiLSEwr6IiIJUdAXEUmIgr5IBHSUr9RFQV+kQ0WDeUwH/3S5gdLGcbJCp1YWkWYUPWVDyActDevyNBQ6BcZkCvoiHSoazGM5yhea20AVmQMf08axK5qnL9IgHQBVn+uuy3rx27c3vwE8dAg+/ens9tatcfx2Refpq6dfkRq1LCaWdEMM63GbvfgdO2Dbtuz26aeH/duVpaBfUSyNWroRS7pheD0OcSPQZopr82Y4cuTE7T5R0K8olkYtz9dWUIslFz+8HqfemVm6FK65ZvRzIW4Qy1DQryiWRi3PVyaoxd7Iixhej9WZGS/2DaKCviSpTFAb1cj7viFQZ2a82DeICvqSpDJBbVQjj723J9OLfYOooC8ywahGHntvr2/6vudVJ52GQWQKdV+HtSt1nLagynsMvrbK+8R0moquJdvTV89ApJ40VZVB8cHXwvRlKbvnlXL7TzboKycrUk+aqsqg+KjXTlOWsnn2lNt/sqdhSHlLL9KVce2u7fbYx/avyyVO0JecrMhi6jrV8Lj3Kfv+49pd2zn5lNt/sukdkRTUlcYY9z51vb9mQ7Un2Z6+hK2Ji2GUec++XIxj8+bsrJRVg+m496nr/cv2vOv+fdpY34JZp9w9qL9169a5yPbt7pD97+I9m/j8kCwsZN9tYaHaMl2p+/dpY31rep0C9niBGKv0jgSp6O5+mQG5MimEWNIN0w5IFknLhDzDpe7fp4nfe/g9g1mnimwZ2vxTT1/K6HuPfJJpv3/sPX05GerpSwqC6T11ZNrvP2pe+/BeQ+znmJHRNJArUUt56h3U+/3rnjZZdeAyhsHaGBUK+ma2wcweNLP9ZnbViOffambfMLNjZnbJ0HObzOyh/G9TXQUXkXqNm4kzbbCsuhGpeyOk8/NkJqZ3zGwJcANwPjAP7DazXe7+3YHFfgD8HvDBode+HLgamAUc2Ju/9ql6ii8idRmXzpl2QLdq6i2GwdoYFcnprwf2u/sBADPbCcwBPw/67v5I/txzQ699O/AVdz+cP/8VYAPwhcolF5FSpp3pU+e4QZuvb/r9YlUkvbMMeHTg/nz+WBFVXis9pLzqZE3V0bTpjdTHTfqmSE/fRjxW9CxthV5rZluALQArV64s+NYSo5DnfoeiqTpSekOgWE9/HlgxcH85cLDg+xd6rbvf6O6z7j47MzNT8K0lRnUdtl+3NvZAin7GcB3VVTb12AWKBf3dwBozW21mpwIbgV0F3/8u4AIze5mZvQy4IH9MAtdUEAw18LQxs6PoZwzXkWadSJ0mBn13PwZsJQvWDwBfdPd9ZnatmV0MYGbnmtk88E7gs2a2L3/tYeDPyTYcu4Frjw/qSthSCzRt7IFs3gxXXw1HjpTbmI4qW6hjI02dyjnU7xulIofttvmn0zCEQYfgN6Ou00aEevqJpr5fqN83JOg0DFKFprc1o67B1FAHZZv6fqF+3xgle7lEKaePl5dLhX67NOhyiVKrrnP8yulOr+vfTsKi9I4U0vXuteb3T6/r307CovSOREEpCpHFKb0jjegqzRLq/P4UKdUWNwX9gMTQmMrkh2P4PlKexgjippx+QGLIW5fJD8fwffqmjTSYxgji1vugH1MuOIbGVGb+fgzfp2/a2NDGcgxHTG2/Tb0P+jH1NmNpTEX17fvEQBvaE2Jq+23qfdBXI5AYTdtL1Yb2BLX90Xo/kBvKrI9QBjVDKYcsToOl1YXS9kPT+55+KELZ1QylHNNKJU+rXqo0RUG/JaE04lDKMa3YN1qLGd6g9e37SRh6n94JxWAjLppeaSIVE/sub6hX3lpM0d+x7pROm6m8Jj9r0nsrZVlSkfMvt/nX9/PplzkvuM4h3g9Ff8e6r2HQ5vrT5GdNem+1kww6n36YyqRXYk/FSKbo71h3SqfN9afJz5r03mon5eiEaxKlVAZ0RYrSCdekE23lVzWlUWQ6Su/IVMb1tNuaXaNdepHpqKefqKo98nE97bZm17Q9CymmGSIxlVXap55+oqr2yMf1tBcbjIw5Dx/T8QExlVXap6CfqKrpkWlmmsQcjGJKJ8VU1r4LsaOj2TtS2rQr8qjXhdgoROpy3XVZR2f79uY7OkVn76inL6VN22MftXcQYu9fGyKpS4h7XQr6UtqkFblM0OyyUXQ9A0kbl/4L8RxKvQv6akjNm7QilwmaXTaKceVsa0MU4l6O9F/vgr4aUvdC3KUdZZoZSG18vkiTCg3kmtkG4FPAEuAmd//40POnAbcC64AngXe5+yNm9gLgJuAcsg3Mre7+F4t9VtWBXPX0pS5alyQmtZ2GwcyWADcAFwJrgUvNbO3QYpcDT7n7WcD1wCfyx98JnOburyfbILzXzFYV/RLTiP3UwePogJv26VQP/aN2VOyI3PXAfnc/4O5HgZ3A3NAyc8At+e07gPPMzAAHTjezU4BfBI4CP6ql5C0KYUVRAGpfjOful8WpHRXL6S8DHh24Pw+8adwy7n7MzJ4GziTbAMwBjwMvAv7Q3Q9XLXTb6honqJIuUP63fSHOvIhRSGkytaNiPX0b8djwQMC4ZdYDzwKvAlYDHzCzV5/0AWZbzGyPme1ZWFgoUKT2HDoER47A1VdXX1Gq9DL6mrZqQgh7ZqHqom5C6l2rHRUL+vPAioH7y4GD45bJUzlnAIeBy4B/c/efufsTwFeBkwYa3P1Gd59199mZmZny36JBO3bAtm0nbldpLEoXtCOkIBOaLupG631YiqR3dgNrzGw18BiwkSyYD9oFbALuBS4B7nF3N7MfAG8zs78nS++8GfhkXYVvw/EV9ciR6ikepQvaoV348bqoG633YSk6ZfMismC9BLjZ3T9mZteSXZNxl5m9ELgNeCNZD3+jux8wsxcDO8hm/Riww92vW+yzQj33Tkh5yZiUqTfVscj0aj33jrvfCdw59NhHBm7/hGx65vDrnhn1eIzUW5lOmUHwOgbMteEQWVzvjshNQUyBre0LweuIbJlWTO2qCl05KxfTjI8yg3Fdf68ysyXqmFmhQcM4dL1ejpLKBAD19HMx9RDL9Ihj+l7Dpul5KQ0XhxDXy1QmACjo52L6wcsEtpi+17AQA0PK6kx/hLheptJh0JWzJFip5Fhj0eZVoKS82k64Js0JMa/Zpknfv2qOP/X6rVuM4yVaB06moN+hVAaOxmn6+6dev3WL8RQGWgdOppx+h0LMa7ap7u8/nA5KvX5F68AoyulLbyjnLClTTl9Kiz3/GWPOeRqx/07SLQV9+blY8p/jgl6MOedpxPI7TRLTxiumsk6inL78XJf5zzLTM6vM3+/DNNC+5KljOg4jprJOoqAfoK4CU5cHp5RpVFWCXh8ab18OIopp4xVTWSdR0G9A1aDdh8BUVplGVSboaUZPuGLaeMVU1kmU029A1ZxrUwOSo/KSoeQqm8rHD/8WqeT95YRQ1vFQqKffgKq9yaZ6FaP2IPq+VxFzz74P4w8h6Ps6XpaCfgNC3RUcFQBjDopFhPpbFFEmWGkDMV7f1/GydHCWVKJg05wydasD06TWyyWKjKNd5+akcgptaZeCvlQSQ7BJYW8k5jSWtCu52Tsaya9XDLNh+nIEq0gdkgv6CgDpieGcPFU6I+rIdCu2+k8u6McQAKReRfdGumy8VTojXXRkYgt0TYqtI5lcTl+5Txk0mO/vclC6ythIF+MqGsA/YVL9hzamlFzQFxk0GLy6HJSu0hnpoiMTwwB+WybVf2gbSAV9Sdpg8NJeYHGqq+JC20Aml9MfNG1eso18pnKm7Yhh9pHELbR1LOmgP+0ATBsDN0U/QxsHESmjUHrHzDYAnwKWADe5+8eHnj8NuBVYBzwJvMvdH8mfewPwWeAlwHPAue7+k7q+QBXT7na1sbs27jOGB4VCyxeKSODcfdE/skD/MPBq4FTgW8DaoWV+H/jb/PZG4Pb89inA/cCv5ffPBJYs9nnr1q1zGW/7dnfI/ru7LyxktxcWui2XnEy/TfNUxycAe3xCPHf3Qumd9cB+dz/g7keBncDc0DJzwC357TuA88zMgAuA+939W/kG5kl3f7b0lmlKw6mPPqRCho8zCC1fKCfENn87Rqrj8oqkd5YBjw7cnwfeNG4Zdz9mZk+T9epfA7iZ3QXMADvdfXvlUhc0nProQypEsybiEdqsjb4YTHGqjssrEvRtxGPD52Met8wpwFuAc4EfA3fnp/+8+3kvNtsCbAFYuXJlgSIVM7xCxHYQRSxUb6NpA92M4c5bX+q4rXZUJL0zD6wYuL8cODhuGTM7BTgDOJw//p/ufsjdfwzcCZwz/AHufqO7z7r77MzMTPlvMcZw6mNSKiSmXcWQUlUx1ZvEr6+nUmmrHRXp6e8G1pjZauAxsoHay4aW2QVsAu4FLgHucffjaZ0rzexFwFHgN4Hr6yp83WLaVQwpVRVTvUn8+roH1VY7KnTlLDO7CPgk2Uyem939Y2Z2Ldlo8S4zeyFwG/BGsh7+Rnc/kL/23cCHyNI9d7r7lYt9VlNXzupbCqJv30dEqil65axkLpeY0uXktEEQCV/d7VSXSxySUgoipNSPiIzWVTtNJuj3NQ84St82cNpzkT7qqp0mfe6dvurbAVuhzw6aZiZVSLOvpBtdtdNkevoSj+Gefeh7LtPspisFJ11R0E9YqGmT4YAYcmru0CE4cgSuvrrcRin0DZn0l4J+wkLtbcYUEHfsgG3bsllhZTacIW/IpN8U9BMWanCNKSCGWoci4yQzT1+kDaGmzKT/is7T1+ydQGg2Rz+EPtNIuhVCO1d6JxCh5telHKV7ZDEhtHMF/UB0HSzaTkv0NQ0S03iEtK/oZVCbpPROILo+oGowLdHGLqjSINJXi7Wfce28zfagnn5HQuvpDvZA2tgF7XrPRqQp07SfNtuDgn5HQsjtDRpMS7SxAioNIn01Tftpsz1oymZHQuvpi0jcNGWzhC6mUXWdw5c4hTDlT+KmoI8GFSUeWlelKuX00aCixEPrqlSlnL6MpXEHkXgopy+V879KJYg0q4sxGqV3ahZS77jqtFClEkSa1cXUbQX9moU0/75q0NZc+uJC2thLPLroWCno1yyk3rGCdntC2thLPLpoowr6NVOgTVNIG3spLsU9NA3kitRAB9vFKcXJCgr6NUr9aMmuvn/q9S7T27w5u75xSntoCvo1SrHXMKir7596vcv0UtxDU06/Rqnndbv6/qnXu7SvylhA1+MIhXr6ZrbBzB40s/1mdtWI508zs9vz579uZquGnl9pZs+Y2QfrKXaYUuw1DOrq+6de79K+KnuXXe+ZTuzpm9kS4AbgfGAe2G1mu9z9uwOLXQ485e5nmdlG4BPAuwaevx741/qKLSLSnSp7l13vmRbp6a8H9rv7AXc/CuwE5oaWmQNuyW/fAZxnZgZgZu8ADgD76ily/2ggUiQuVfYuu94zLRL0lwGPDtyfzx8buYy7HwOeBs40s9OBPwa2VS9qf3W9uyci6SgykGsjHhs+Nee4ZbYB17v7M3nHf/QHmG0BtgCsXLmyQJH6pevdPRFJR5Ge/jywYuD+cuDguGXM7BTgDOAw8CZgu5k9Arwf+BMz2zr8Ae5+o7vPuvvszMxM6S8RqqJpm65390QkHUV6+ruBNWa2GngM2AhcNrTMLmATcC9wCXCPZyfq/43jC5jZNcAz7v7pGsodBZ2PRURCMzHou/uxvHd+F7AEuNnd95nZtcAed98FfA64zcz2k/XwNzZZ6FgobSMiodGVs0bo+uAJEZGydOWsCkKbTaMpneHQbyGxU9AfYbGTMJVt9HUEidA2QinTbyGx07l3RljsnPhlB2frGMzV2EA49FtI7BT0Syrb6OsIEn24MEtfxkn68FtI2hT0Syrb6BUkMpq+KhIGBf0p9aXn2halRUTCoIHcKcU2oNf1rBMddSwSBvX0pxRbz1XpFREBBf2pxZarj20jJSLNUNBPRGwbKRFphnL6IiIJUdAXEUmIgn4PdD0zR0TioaDfA7FNHxWR7mggtwc0M0dEilJPvwd04JPUSenCflPQH0MrvqSqqXSh2lQYlN4ZQ0ewSqqaSheqTYVBQX8M5cklVU0dyNfHNhXjiRcV9MfQEawi9epjm4px70VBX0RkSjHuvSjoi4hMKca9F83eERFJiIK+iEhCFPRFRBKioC8ikhAFfRGRhCjoi4gkpFDQN7MNZvagme03s6tGPH+amd2eP/91M1uVP36+me01s2/n/99Wb/FFRKSMiUHfzJYANwAXAmuBS81s7dBilwNPuftZwPXAJ/LHDwG/7e6vBzYBt9VVcBERKa9IT389sN/dD7j7UWAnMDe0zBxwS377DuA8MzN3/6a7H8wf3we80MxOq6PgIiJSXpGgvwx4dOD+fP7YyGXc/RjwNHDm0DK/C3zT3X86/AFmtsXM9pjZnoWFhaJlFxGRkooEfRvxmJdZxsxeR5byee+oD3D3G9191t1nZ2ZmChRJRKR+KZzzv0jQnwdWDNxfDhwct4yZnQKcARzO7y8HvgS8x90frlpgEZGmpHC96SInXNsNrDGz1cBjwEbgsqFldpEN1N4LXALc4+5uZi8Fvgx8yN2/Wl+xRUTqF+NZM8ua2NPPc/RbgbuAB4Avuvs+M7vWzC7OF/sccKaZ7Qf+CDg+rXMrcBbwZ2Z2X/73itq/RS6FXTMRaU4K15sudGpld78TuHPosY8M3P4J8M4Rr/so8NGKZSwsxgsaxCDGqwOJyGi9Op9+CrtmXdDGVKQ/ehX0Y7ygQQy0MRXpD517RyZKIc8p6Ul1DFBBX0SSlML0zFF6ld4RESkq1bSlgr6IJCnVMUCld0REEqKgLyKSEAV9EZGEKOiLiCREQV9EJCEK+iIiCVHQFxFJiLkPXwSrW2a2AHy/wlssJbsgu5ygOhlN9XIy1cloMdTLr7j7xEsPBhf0qzKzPe4+23U5QqI6GU31cjLVyWh9qheld0REEqKgLyKSkD4G/Ru7LkCAVCejqV5OpjoZrTf10rucvoiIjNfHnr6IiIzRm6BvZhvM7EEz229mV3Vdni6Z2SNm9m0zu8/M9uSPvdzMvmJmD+X/X9Z1OZtkZjeb2RNm9p2Bx0bWgWX+Ol937jezc7orebPG1Ms1ZvZYvr7cZ2YXDTz3obxeHjSzt3dT6maZ2Qoz+w8ze8DM9pnZH+SP93J96UXQN7MlwA3AhcBa4FIzW9ttqTr3W+5+9sA0s6uAu919DXB3fr/PPg9sGHpsXB1cCKzJ/7YAn2mpjF34PCfXC8D1+fpytrvfCZC3oY3A6/LX/E3e1vrmGPABd38t8Gbgffl37+X60ougD6wH9rv7AXc/CuwE5jouU2jmgFvy27cA7+iwLI1z9/8CDg89PK4O5oBbPfM14KVm9sp2StquMfUyzhyw091/6u7/A+wna2u94u6Pu/s38tv/BzwALKOn60tfgv4y4NGB+/P5Y6ly4N/NbK+Zbckf+2V3fxyylRx4RWel6864OtD6A1vzVMXNA6m/5OrFzFYBbwS+Tk/Xl74EfRvxWMrTkn7d3c8h2w19n5m9tesCBS719eczwK8CZwOPA3+ZP55UvZjZi4F/At7v7j9abNERj0VTL30J+vPAioH7y4GDHZWlc+5+MP//BPAlsl3yHx7fBc3/P9FdCTszrg6SXn/c/Yfu/qy7Pwf8HSdSOMnUi5m9gCzg/4O7/3P+cC/Xl74E/d3AGjNbbWankg0+7eq4TJ0ws9PN7JeO3wYuAL5DVh+b8sU2Af/STQk7Na4OdgHvyWdlvBl4+vhufQqG8tG/Q7a+QFYvG83sNDNbTTZw+d9tl69pZmbA54AH3P2vBp7q5/ri7r34Ay4Cvgc8DHy46/J0WA+vBr6V/+07XhfAmWQzEB7K/7+867I2XA9fIEtV/IysZ3b5uDog212/IV93vg3Mdl3+luvltvx7308W0F45sPyH83p5ELiw6/I3VCdvIUvP3A/cl/9d1Nf1RUfkiogkpC/pHRERKUBBX0QkIQr6IiIJUdAXEUmIgr6ISEIU9EVEEqKgLyKSEAV9EZGE/D8o8LGbvn/qUgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f0b830fa610>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot([get_density(cc) for cc in crystals],'ob',ms=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "view(crystals[it] for it in [0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.         1.         1.14006449]\n",
      "[1.         1.         1.14006449]\n",
      "[0.87714336 0.87714336 1.        ]\n"
     ]
    }
   ],
   "source": [
    "for it in range(3):\n",
    "    print length / length[it]\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "code_folding": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2358af04da4a412e8427a9316abce2a9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=500), HTML(value=u'')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "signal: 2\n",
      "killing child: 29580\n",
      "killing child: 29582\n",
      "killing child: 29583\n"
     ]
    }
   ],
   "source": [
    "#%%time\n",
    "\n",
    "sites_z = [14]\n",
    "nprocess = 1\n",
    "\n",
    "pool = mp_gen([sites_z for it in range(500)],nprocess=nprocess,dispbar=False)\n",
    "\n",
    "new_cc = pool.run()\n",
    "# for it in tqdm_notebook(range(500)):\n",
    "#     cc, sg, wyckoff_letters = input2crystal(sites_z)\n",
    "#     ee = LJ_vcrelax_alternative(cc,isotropic_external_pressure=1e-2,debug=True)\n",
    "#     new_cc.append(ee)\n",
    "\n",
    "soap_params = dict(nmax= 9, cutoff= 4, gaussian_width= 0.4, lmax= 9,\n",
    "        centerweight = 1.,cutoff_transition_width = 0.5,\n",
    "        nocenters = [],is_fast_average=True,chem_channels=False,dispbar=True\n",
    "                      )\n",
    "\n",
    "\n",
    "new_crystals = []\n",
    "new_crystals.extend(crystals)\n",
    "new_crystals.extend(new_cc)\n",
    "\n",
    "fings = get_fingerprints([ase2qp(crystal) for crystal in new_crystals],soap_params,nprocess)\n",
    "kernel = np.dot(fings,fings.T)\n",
    "\n",
    "fps_ids, minmax = fpsSelection_with_restart(data=kernel,distance_func=distance_func2,restart_ref=None,\n",
    "                                            disable_pbar=True,nthread=10,\n",
    "                              intermediate_copy=True,stride=10000,threshold=5e-3,Nmin=0,\n",
    "                                            Nmax=kernel.shape[0],seed=None,fn=None)\n",
    "\n",
    "new_crystals = [new_crystals[it] for it in fps_ids]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "from libs.io import Frame_Dataset_h5\n",
    "from glob import glob"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "fns = glob('./structures/relaxed_structures_r*')\n",
    "readers = []\n",
    "for fn in fns:\n",
    "    readers.append(Frame_Dataset_h5(fn,mode='r',disable_pbar=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "frames = []\n",
    "for reader in readers:\n",
    "    frames.extend(reader.load_frames().values())\n",
    "#view(frames)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "inputs = []\n",
    "for reader in readers:\n",
    "    inputs.extend(reader.get_inputs().values())\n",
    "#view([cc for (cc,_) in inputs])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['wyckoffs_spg', 'wyckoff_letters', 'sites_z', 'equivalent_atoms_spg', 'sg_spg', 'sg']\n"
     ]
    }
   ],
   "source": [
    "inp = inputs[0][1]\n",
    "print inp.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "227 ['b', 'b', 'b', 'b', 'b', 'b', 'b', 'b'] [0 0 0 0 0 0 0 0]\n"
     ]
    }
   ],
   "source": [
    "cc = crystals[5]\n",
    "sym_data = spg.get_symmetry_dataset(cc)\n",
    "print sym_data['number'],sym_data['wyckoffs'],sym_data['equivalent_atoms']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "view(get_standard_frame(cc,to_primitive=True,symprec=1e-5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "new_frames = []\n",
    "QMAT_id = 1\n",
    "for ff,inp in zip(frames,inputs):\n",
    "    equivalent_atoms = inp[1]['equivalent_atoms_spg']\n",
    "    #wyckoffs = inp[1]['wyckoffs_spg']\n",
    "    if len(np.unique(equivalent_atoms)) == QMAT_id:\n",
    "        new_frames.append(ff)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "new_crystals = []\n",
    "new_crystals.extend(new_frames)\n",
    "new_crystals.extend(crystals)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "10f130fd873e481b87be2e6fc0d41a66",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description=u'SOAP vectors', max=9724), HTML(value=u'')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "soap_params = dict(nmax= 9, cutoff= 4, gaussian_width= 0.4, lmax= 9,\n",
    "        centerweight = 1.,cutoff_transition_width = 0.5,\n",
    "        nocenters = [],is_fast_average=True,chem_channels=False,dispbar=False\n",
    "                      )\n",
    "nprocess = 4\n",
    "\n",
    "fings = get_fingerprints([ase2qp(crystal) for crystal in new_crystals],soap_params,nprocess)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "kernel = np.dot(fings,fings.T)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7f193847c9b34fe894ec8124a11b6914",
       "version_major": 2,
       "version_minor": 0
      },
      "text/html": [
       "<p>Failed to display Jupyter Widget of type <code>HBox</code>.</p>\n",
       "<p>\n",
       "  If you're reading this message in the Jupyter Notebook or JupyterLab Notebook, it may mean\n",
       "  that the widgets JavaScript is still loading. If this message persists, it\n",
       "  likely means that the widgets JavaScript library is either not installed or\n",
       "  not enabled. See the <a href=\"https://ipywidgets.readthedocs.io/en/stable/user_install.html\">Jupyter\n",
       "  Widgets Documentation</a> for setup instructions.\n",
       "</p>\n",
       "<p>\n",
       "  If you're reading this message in another frontend (for example, a static\n",
       "  rendering on GitHub or <a href=\"https://nbviewer.jupyter.org/\">NBViewer</a>),\n",
       "  it may mean that your frontend doesn't currently support widgets.\n",
       "</p>\n"
      ],
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=9724), HTML(value=u'')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "fps_ids, minmax = fpsSelection_with_restart(data=kernel,distance_func=distance_func2,restart_ref=None,disable_pbar=False,nthread=10,\n",
    "                              intermediate_copy=True,stride=10000,threshold=1e-9,Nmin=0,Nmax=kernel.shape[0],seed=None,\n",
    "                                           fn=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff2f41ffd50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(minmax[:],'ob',ms=1)\n",
    "plt.yscale('log')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "minmax = np.array(minmax)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(231,)"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "minmax[minmax > 1e-3].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "import hdbscan"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [],
   "source": [
    "distance = distance_func2(kernel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "tsne = TSNE(n_components=2, perplexity=30.0, early_exaggeration=12.0, \n",
    "     learning_rate=200.0, n_iter=1000, metric='precomputed')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = tsne.fit_transform(distance)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "182\n"
     ]
    }
   ],
   "source": [
    "clusterer = hdbscan.HDBSCAN(algorithm='best', alpha=1.0, approx_min_span_tree=True,\n",
    "    gen_min_span_tree=False, leaf_size=40, \n",
    "    metric='precomputed', min_cluster_size=10, min_samples=None, p=None)\n",
    "clusterer.fit(distance)\n",
    "print len(np.unique(clusterer.labels_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(9724, 2)"
      ]
     },
     "execution_count": 99,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [],
   "source": [
    "ids = np.array(fps_ids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7ff2a2bd33d0>"
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff2a4c1cbd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "color_palette = sns.color_palette('deep', len(np.unique(clusterer.labels_)))\n",
    "cluster_colors = [color_palette[x] if x >= 0\n",
    "                  else (0.5, 0.5, 0.5)\n",
    "                  for x in clusterer.labels_]\n",
    "cluster_member_colors = [sns.desaturate(x, p) for x, p in\n",
    "                         zip(cluster_colors, clusterer.probabilities_)]\n",
    "plt.scatter(*X.T, s=50, linewidth=0, c=cluster_member_colors, alpha=0.25)\n",
    "plt.scatter(*X[ids[:300],:].T, s=50, linewidth=0, c='b', alpha=0.25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2 glosim",
   "language": "python",
   "name": "glosim"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}