molecule-transformer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#%load_ext autoreload\n",
    "#%autoreload 2\n",
    "import os\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"]=\"0,1\"\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from pathlib import Path\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.parallel\n",
    "import torch.utils.data\n",
    "from torch.autograd import Variable\n",
    "import numpy as np\n",
    "import torch.nn.functional as F\n",
    "import multiprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#from torch.utils.data.sampler import RandomSampler\n",
    "#RandomSampler_num_samples = RandomSampler.num_samples\n",
    "#@property\n",
    "#def RandomSampler_num_samples_x10(self): return  self.RandomSampler_num_samples*10\n",
    "#RandomSampler.num_samples= RandomSampler_num_samples_x10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from torch.nn import Parameter\n",
    "\n",
    "def _load_from_state_dict(self, state_dict, prefix, local_metadata, strict,\n",
    "                          missing_keys, unexpected_keys, error_msgs):\n",
    "    r\"\"\"Copies parameters and buffers from :attr:`state_dict` into only\n",
    "    this module, but not its descendants. This is called on every submodule\n",
    "    in :meth:`~torch.nn.Module.load_state_dict`. Metadata saved for this\n",
    "    module in input :attr:`state_dict` is provided as :attr:`local_metadata`.\n",
    "    For state dicts without metadata, :attr:`local_metadata` is empty.\n",
    "    Subclasses can achieve class-specific backward compatible loading using\n",
    "    the version number at `local_metadata.get(\"version\", None)`.\n",
    "\n",
    "    .. note::\n",
    "        :attr:`state_dict` is not the same object as the input\n",
    "        :attr:`state_dict` to :meth:`~torch.nn.Module.load_state_dict`. So\n",
    "        it can be modified.\n",
    "\n",
    "    Arguments:\n",
    "        state_dict (dict): a dict containing parameters and\n",
    "            persistent buffers.\n",
    "        prefix (str): the prefix for parameters and buffers used in this\n",
    "            module\n",
    "        local_metadata (dict): a dict containing the metadata for this module.\n",
    "            See\n",
    "        strict (bool): whether to strictly enforce that the keys in\n",
    "            :attr:`state_dict` with :attr:`prefix` match the names of\n",
    "            parameters and buffers in this module\n",
    "        missing_keys (list of str): if ``strict=True``, add missing keys to\n",
    "            this list\n",
    "        unexpected_keys (list of str): if ``strict=True``, add unexpected\n",
    "            keys to this list\n",
    "        error_msgs (list of str): error messages should be added to this\n",
    "            list, and will be reported together in\n",
    "            :meth:`~torch.nn.Module.load_state_dict`\n",
    "    \"\"\"\n",
    "    for hook in self._load_state_dict_pre_hooks.values():\n",
    "        hook(state_dict, prefix, local_metadata, strict, missing_keys,\n",
    "             unexpected_keys, error_msgs)\n",
    "\n",
    "    local_name_params = itertools.chain(self._parameters.items(),\n",
    "                                        self._buffers.items())\n",
    "    local_state = {k: v.data for k, v in local_name_params if v is not None}\n",
    "\n",
    "    for name, param in local_state.items():\n",
    "        key = prefix + name\n",
    "        if key in state_dict:\n",
    "            input_param = state_dict[key]\n",
    "\n",
    "            # Backward compatibility: loading 1-dim tensor from 0.3.* to version 0.4+\n",
    "            if len(param.shape) == 0 and len(input_param.shape) == 1:\n",
    "                input_param = input_param[0]\n",
    "\n",
    "            if input_param.shape != param.shape:\n",
    "                # local shape should match the one in checkpoint\n",
    "                error_msgs.append(\n",
    "                    'size mismatch for {}: copying a param with shape {} from checkpoint, '\n",
    "                    'the shape in current model is {}.'.format(\n",
    "                        key, input_param.shape, param.shape))\n",
    "                #if not strict:\n",
    "                #    continue\n",
    "\n",
    "            if isinstance(input_param, Parameter):\n",
    "                # backwards compatibility for serialized parameters\n",
    "                input_param = input_param.data\n",
    "\n",
    "            try:\n",
    "                param.copy_(input_param)\n",
    "            except Exception:\n",
    "                error_msgs.append(\n",
    "                    'While copying the parameter named \"{}\", '\n",
    "                    'whose dimensions in the model are {} and '\n",
    "                    'whose dimensions in the checkpoint are {}.'.format(\n",
    "                        key, param.size(), input_param.size()))\n",
    "                # PG load partially\n",
    "\n",
    "                if len(input_param.size()) == 3:\n",
    "                    error_msgs.append(\n",
    "                        'Partially copying the parameter named \"{}\", '\n",
    "                        'whose dimensions in the model are {} and '\n",
    "                        'whose dimensions in the checkpoint are {}. - trying {}'\n",
    "                        .format(\n",
    "                            key, param.size(), input_param.size(),\n",
    "                            param[:input_param.size()[0], :input_param.size(\n",
    "                            )[1], :input_param.size()[2]].shape))\n",
    "                else:\n",
    "                    error_msgs.append(\n",
    "                        'Partially copying the parameter named \"{}\", '\n",
    "                        'whose dimensions in the model are {} and '\n",
    "                        'whose dimensions in the checkpoint are {}. - trying {}'\n",
    "                        .format(key, param.size(), input_param.size(),\n",
    "                                param[:input_param.size()[0]].shape))\n",
    "\n",
    "                try:\n",
    "                    new_input_param = torch.empty_like(param)\n",
    "                    new_input_param = torch.nn.init.normal_(new_input_param,\n",
    "                                                            mean=input_param.mean(),\n",
    "                                                            std=input_param.std())\n",
    "\n",
    "                    if len(input_param.size()) == 3:\n",
    "                        new_input_param[:input_param.size()[0], :input_param.\n",
    "                                        size()[1], :input_param.size(\n",
    "                                        )[2]] = input_param\n",
    "                    else:\n",
    "                        new_input_param[:input_param.size()[0]] = input_param\n",
    "                    param.copy_(new_input_param)\n",
    "                except Exception as e:\n",
    "                    assert e\n",
    "                    error_msgs.append(\n",
    "                        'Failed to load weights partially {}'.format(e))\n",
    "        elif strict:\n",
    "            missing_keys.append(key)\n",
    "\n",
    "    if strict:\n",
    "        for key in state_dict.keys():\n",
    "            if key.startswith(prefix):\n",
    "                input_name = key[len(prefix):]\n",
    "                input_name = input_name.split(\n",
    "                    '.', 1)[0]  # get the name of param/buffer/child\n",
    "                if input_name not in self._modules and input_name not in local_state:\n",
    "                    unexpected_keys.append(key)\n",
    "\n",
    "def load_state_dict(self, state_dict, strict=True):\n",
    "    r\"\"\"Copies parameters and buffers from :attr:`state_dict` into\n",
    "    this module and its descendants. If :attr:`strict` is ``True``, then\n",
    "    the keys of :attr:`state_dict` must exactly match the keys returned\n",
    "    by this module's :meth:`~torch.nn.Module.state_dict` function.\n",
    "\n",
    "    Arguments:\n",
    "        state_dict (dict): a dict containing parameters and\n",
    "            persistent buffers.\n",
    "        strict (bool, optional): whether to strictly enforce that the keys\n",
    "            in :attr:`state_dict` match the keys returned by this module's\n",
    "            :meth:`~torch.nn.Module.state_dict` function. Default: ``True``\n",
    "\n",
    "    Returns:\n",
    "        ``NamedTuple`` with ``missing_keys`` and ``unexpected_keys`` fields:\n",
    "            * **missing_keys** is a list of str containing the missing keys\n",
    "            * **unexpected_keys** is a list of str containing the unexpected keys\n",
    "    \"\"\"\n",
    "    missing_keys = []\n",
    "    unexpected_keys = []\n",
    "    error_msgs = []\n",
    "\n",
    "    # copy state_dict so _load_from_state_dict can modify it\n",
    "    metadata = getattr(state_dict, '_metadata', None)\n",
    "    state_dict = state_dict.copy()\n",
    "    if metadata is not None:\n",
    "        state_dict._metadata = metadata\n",
    "\n",
    "    def load(module, prefix=''):\n",
    "        local_metadata = {} if metadata is None else metadata.get(\n",
    "            prefix[:-1], {})\n",
    "        module._load_from_state_dict(state_dict, prefix, local_metadata, True,\n",
    "                                     missing_keys, unexpected_keys, error_msgs)\n",
    "        for name, child in module._modules.items():\n",
    "            if child is not None:\n",
    "                load(child, prefix + name + '.')\n",
    "                \n",
    "    load(self)\n",
    "\n",
    "    if strict:\n",
    "        if len(unexpected_keys) > 0:\n",
    "            error_msgs.insert(\n",
    "                0, 'Unexpected key(s) in state_dict: {}. '.format(', '.join(\n",
    "                    '\"{}\"'.format(k) for k in unexpected_keys)))\n",
    "        if len(missing_keys) > 0:\n",
    "            error_msgs.insert(\n",
    "                0, 'Missing key(s) in state_dict: {}. '.format(', '.join(\n",
    "                    '\"{}\"'.format(k) for k in missing_keys)))\n",
    "\n",
    "    if strict and len(error_msgs) > 0:\n",
    "        raise RuntimeError(\n",
    "            'Error(s) in loading state_dict for {}:\\n\\t{}'.format(\n",
    "                self.__class__.__name__, \"\\n\\t\".join(error_msgs)))\n",
    "    return None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "nn.Module._load_from_state_dict = _load_from_state_dict\n",
    "nn.Module.load_state_dict = load_state_dict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from fastai import *\n",
    "from fastai.basic_train import *\n",
    "from fastai.basic_data import *\n",
    "from fastai.data_block import *\n",
    "from fastai.torch_core import *\n",
    "from fastai.train import *\n",
    "from fastai.callback import *\n",
    "from fastai.callbacks import *\n",
    "from fastai.distributed import *\n",
    "from fastai.layers import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ext = '_ext' # or ''\n",
    "multi = '_molecule'\n",
    "train_fname = Path('train.npz')\n",
    "type_index=types = ['1JHC', '2JHH', '1JHN', '2JHN', '2JHC', '3JHH', '3JHC', '3JHN'] \n",
    "atoms = 'CFHNO'\n",
    "xyz = ['x', 'y', 'z']\n",
    "max_atoms = 29 #int(s.atom_index.max() + 1)\n",
    "\n",
    "fname_ext = lambda fname,ext,multi: f'{str(fname)[:-4]}{ext}{multi}{str(fname)[-4:]}'\n",
    "s  = pd.read_csv('structures.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def process_group(process_args,i):\n",
    "    m_group,has_y,x0_xyz,x0_type,x0_ext,x0_atom,x0_id,m0,y0_scalar = process_args\n",
    "    x0_atom[:] = -1\n",
    "    x0_type[:] = -1\n",
    "    x0_id[:] = -1\n",
    "    y0_scalar[:] = np.nan\n",
    "    (m_name, m_index) ,m_group = m_group\n",
    "    ss = s[s.molecule_name==m_name]\n",
    "    n_atoms = len(ss)\n",
    "    ii = 0\n",
    "    x0_xyz[ii] = 0.\n",
    "    x0_type[ii] = -1\n",
    "    x0_ext[ii] =  True\n",
    "    x0_xyz[ii,:,:n_atoms] = ss[xyz].values.T  # xyz \n",
    "    x0_type[ii,0,m_group['atom_index_1'], m_group['atom_index_0']] = m_group['type_idx'].T\n",
    "    x0_ext[ii,0, m_group['atom_index_0']] = m_group['ext'].T\n",
    "    x0_atom[ii,:,:n_atoms] = ss['atom_idx'].T\n",
    "    x0_id[ii,0,m_group['atom_index_1'], m_group['atom_index_0']] = m_group['id'].T\n",
    "    m0[ii] = m_index\n",
    "    res = [x0_xyz,x0_type,x0_ext,x0_atom,x0_id,m0]\n",
    "    if has_y:\n",
    "        y0_scalar[0,0,m_group['atom_index_1'], m_group['atom_index_0']] =  m_group['scalar_coupling_constant']\n",
    "        res.append(y0_scalar)\n",
    "    return i,res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def preprocess(fname, type_index=None, ext=''):\n",
    "    t  = pd.read_csv(fname_ext(fname,ext,''))\n",
    "    has_y = 'scalar_coupling_constant' in t.columns\n",
    "    t['molecule_index'] = pd.factorize(t['molecule_name'])[0] + t['id'].min()\n",
    "    # make sure we use the same indexes in train/test (test needs to provide type_index)\n",
    "    if type_index is not None:\n",
    "        t['type_idx'] = t['type'].apply(lambda x: type_index.index(x) ) # pd.factorize(pd.concat([pd.Series(type_index),t['type']]))[0][len(type_index):]\n",
    "    else:\n",
    "        t['type_idx'] = pd.factorize(t['type'])[0]\n",
    "\n",
    "    s['atom_idx'] = s['atom'].apply(lambda x: atoms.index(x) )\n",
    "\n",
    "    max_items = t['molecule_name'].nunique()\n",
    "    x_xyz   = np.zeros((max_items,len(xyz),  max_atoms), dtype=np.float32)\n",
    "    x_type  = np.empty((max_items,1,         max_atoms,max_atoms), dtype=np.int8)\n",
    "    x_ext   = np.zeros((max_items,1,         max_atoms), dtype=np.bool_)\n",
    "    x_atom  = np.empty((max_items,1,         max_atoms), dtype=np.int8)\n",
    "    x_ids   = np.zeros((max_items,           max_atoms,max_atoms), dtype=np.int32)\n",
    "    x_atom[:] = -1\n",
    "    x_type[:] = -1\n",
    "    x_ids[:] = -1\n",
    "    m = np.zeros((max_items,), dtype=np.int32)\n",
    "    if has_y:\n",
    "        y_scalar   = np.empty((max_items,1 ,max_atoms,max_atoms), dtype=np.float32)\n",
    "        y_scalar[:] = np.nan\n",
    "        \n",
    "    x0_xyz   = np.zeros((1,len(xyz),  max_atoms), dtype=np.float32)        \n",
    "    x0_type  = np.empty((1,1,         max_atoms,max_atoms), dtype=np.int8)\n",
    "    x0_ext   = np.zeros((1,1,         max_atoms), dtype=np.bool_)\n",
    "    x0_atom  = np.empty((1,1,         max_atoms), dtype=np.int8)\n",
    "    x0_id    = np.zeros((1,1,         max_atoms,max_atoms), dtype=np.int32)\n",
    "    y0_scalar   = np.empty((1,1 ,max_atoms,max_atoms), dtype=np.float32)\n",
    "    m0 = np.zeros((1,), dtype=np.int32)\n",
    "    \n",
    "    process_args = [(m_group,has_y,x0_xyz,x0_type,x0_ext,x0_atom,x0_id,m0,y0_scalar) \n",
    "                    for m_group in t.groupby(['molecule_name', 'molecule_index'])]\n",
    "    res = parallel(process_group,process_args,max_workers=multiprocessing.cpu_count())\n",
    "    for i,r in progress_bar(res):\n",
    "        x_xyz[i],x_type[i],x_ext[i],x_atom[i],x_ids[i],m[i] = r[:6]\n",
    "        if has_y: y_scalar[i]=r[6]\n",
    "        \n",
    "    res = [x_xyz,x_type,x_ext,x_atom,x_ids,m]\n",
    "    if has_y: res.append(y_scalar)\n",
    "    return res"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define where you want to use original training set '' or extended ''"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load preprocessed or preprocess and save for later"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_fname = Path('train.npz')\n",
    "types = ['1JHC', '2JHH', '1JHN', '2JHN', '2JHC', '3JHH', '3JHC', '3JHN'] \n",
    "atoms = 'CFHNO'\n",
    "\n",
    "try:\n",
    "    npzfile = np.load(fname_ext(train_fname,ext,multi))\n",
    "    x_xyz   = npzfile['x_xyz']\n",
    "    x_type  = npzfile['x_type']\n",
    "    x_ext   = npzfile['x_ext']\n",
    "    x_atom  = npzfile['x_atom']\n",
    "\n",
    "    y_scalar    = npzfile['y_scalar']\n",
    "    m = npzfile['m']\n",
    "    max_items, max_atoms = x_xyz.shape[0], x_xyz.shape[-1]\n",
    "except:\n",
    "    x_xyz,x_type,x_ext,x_atom,x_ids,m,y_scalar=preprocess(train_fname.with_suffix('.csv'),type_index=types,ext=ext)\n",
    "    np.savez(fname_ext(train_fname,ext,multi), \n",
    "             x_xyz=x_xyz,\n",
    "             x_type=x_type,\n",
    "             x_ext=x_ext,\n",
    "             x_atom=x_atom,\n",
    "             x_ids=x_ids,\n",
    "             y_scalar=y_scalar,\n",
    "             m=m)\n",
    "n_types = int(x_type[~np.isnan(x_type)].max() + 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(85003, 3, 29),\n",
       " (85003, 1, 29, 29),\n",
       " (85003, 1, 29),\n",
       " (85003, 1, 29),\n",
       " (85003, 1, 29, 29),\n",
       " (85003,)]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "[v.shape for v in [x_xyz,x_type,x_ext,x_atom,\n",
    "                   y_scalar, m]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_xyz_mean, x_xyz_std = Tensor(x_xyz.mean(axis=(0,2),keepdims=True)), Tensor(x_xyz.std(axis=(0,2),keepdims=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([1, 3, 1])"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_xyz_std.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Fastai classes (this should should be done into its own `application` but who has time?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class MoleculeItem(ItemBase):\n",
    "    def __init__(self,i,xyz,type,ext,atom):#,qm9_mulliken,coulomb): \n",
    "        self.i,self.xyz,self.type,self.ext, self.atom = \\\n",
    "            i,xyz,type,ext,atom\n",
    "        self.data = [Tensor(xyz), LongTensor((type)), \n",
    "                     Tensor(ext), LongTensor((atom))]\n",
    "    def __str__(self):\n",
    "        # TODO: count n_atoms correctly. \n",
    "        n_atoms = np.count_nonzero(np.sum(np.absolute(self.xyz), axis=0))+1\n",
    "        n_couplings = np.sum((self.type!=-1))\n",
    "        return f'{self.i} {n_atoms} atoms {n_couplings} couplings'\n",
    "    \n",
    "    def apply_tfms(self, tfms:Collection, **kwargs):\n",
    "        x = self.clone()\n",
    "        for t in tfms:\n",
    "            if t: x.data = t(x.data)\n",
    "        return x\n",
    "    \n",
    "    def clone(self):\n",
    "        return self.__class__(self.i,self.xyz,self.type,self.ext,self.atom)#,self.qm9_mulliken,self.coulomb)\n",
    "    \n",
    "class ScalarCouplingItem(ItemBase):\n",
    "    def __init__(self,scalar,**kwargs): \n",
    "        self.scalar = scalar#,magnetic,mulliken,dipole,potential\n",
    "        self.data = Tensor(scalar) #, Tensor(magnetic), Tensor(dipole), Tensor(potential))\n",
    "    def __str__(self):\n",
    "        #res, spacer, n_couplings = '', '', 0\n",
    "        n_couplings = (~torch.isnan(self.data)).sum()\n",
    "        return f'{n_couplings}'\n",
    "    def apply_tfms(self, tfms:Collection, **kwargs):\n",
    "        y = self.clone()\n",
    "        for t in tfms:\n",
    "            if 'label_smoothing' == t.__name__:\n",
    "                if t: y.data = t(y.data)\n",
    "        return y\n",
    "    def clone(self): return self.__class__(self.scalar)#,self.magnetic,self.mulliken,self.dipole,self.potential)\n",
    "    def __hash__(self): return hash(str(self))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class LMAEMaskedLoss(Module):\n",
    "    def __init__(self,types_w = [1]*n_types, return_all=False, proxy_log=torch.log, exclude_ext=False):\n",
    "        self.types_w = types_w\n",
    "        self.return_all = return_all\n",
    "        self.proxy_log = proxy_log\n",
    "        self.exclude_ext = exclude_ext\n",
    "        self.loss = torch.nn.L1Loss()\n",
    "    \n",
    "    def forward(self, input_outputs, t_scalar):#, t_magnetic, t_dipole, t_potential):    \n",
    "        t_mask, type, ext, p_scalar = input_outputs\n",
    "        loss = 0.\n",
    "        n = 0\n",
    "        j_loss = [0] * n_types\n",
    "        p_scalar = p_scalar.squeeze(-1)\n",
    "        #print(t_scalar.shape, type.shape)\n",
    "        t_scalar = t_scalar.view(t_scalar.shape[0],-1)[t_mask] #t_scalar[~torch.isnan(t_scalar)].unsqueeze(-1)\n",
    "        #t_scalar = t_scalar.view(-1,1)\n",
    "        #print(t_scalar.shape)\n",
    "        for t in range(n_types):\n",
    "            #if self.exclude_ext:\n",
    "            #    print(type.shape)\n",
    "            mask = (type == t) #if not self.exclude_ext else ((type == t) & (ext == 0)).squeeze(1)\n",
    "            if mask.sum() > 0:\n",
    "               # print(mask.shape)\n",
    "                _output,_target = p_scalar[mask], t_scalar[mask]\n",
    "               # print(mask.shape,p_scalar.shape,t_scalar.shape)\n",
    "\n",
    "                # LMAE scalar\n",
    "#                s_loss = self.proxy_log((_output - _target).abs().mean()+1e-9)\n",
    "                s_loss = self.proxy_log(torch.nn.L1Loss()(_output,_target)+1e-9)\n",
    "                loss += self.types_w[t] * s_loss\n",
    "                j_loss[t] += s_loss\n",
    "                n+=1\n",
    "        loss /= n\n",
    "        \n",
    "        return loss if not self.return_all else (loss, *j_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class ScalarCouplingList(ItemList):\n",
    "    def __init__(self, items:Iterator, **kwargs):\n",
    "        super().__init__(items, **kwargs)\n",
    "        self.loss_func = LMAEMaskedLoss\n",
    "\n",
    "    def get(self, i):\n",
    "        o = super().get(i)\n",
    "        #print(len(o),o[0].shape,o[0].dtype)\n",
    "        return ScalarCouplingItem(np.array(o, dtype=np.float32))\n",
    "\n",
    "    def reconstruct(self,t): return 0; # TODO for viz !!!! ScalarCouplingItem(t.numpy())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Quaterions allow us to rotate 3d points randoming with a nice uniform distribution of 3 numbers hece we use them, however it's still to be seen if are useful here"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#https://github.com/facebookresearch/QuaterNet/blob/master/common/quaternion.py\n",
    "def qrot(q, v):\n",
    "    \"\"\"\n",
    "    Rotate vector(s) v about the rotation described by quaternion(s) q.\n",
    "    Expects a tensor of shape (*, 4) for q and a tensor of shape (*, 3) for v,\n",
    "    where * denotes any number of dimensions.\n",
    "    Returns a tensor of shape (*, 3).\n",
    "    \"\"\"\n",
    "    assert q.shape[-1] == 4\n",
    "    assert v.shape[-1] == 3\n",
    "    assert q.shape[:-1] == v.shape[:-1]\n",
    "    \n",
    "    original_shape = list(v.shape)\n",
    "    q = q.view(-1, 4)\n",
    "    v = v.view(-1, 3)\n",
    "    \n",
    "    qvec = q[:, 1:]\n",
    "    uv = torch.cross(qvec, v, dim=1)\n",
    "    uuv = torch.cross(qvec, uv, dim=1)\n",
    "    return (v + 2 * (q[:, :1] * uv + uuv)).view(original_shape)\n",
    "\n",
    "def random_rotation(data):\n",
    "    x_xyz = data[0].transpose(0,1)\n",
    "    r = torch.rand(3)\n",
    "    sq1_v1,sqv1,v2_2pi,v3_2pi = torch.sqrt(1-r[:1]),torch.sqrt(r[:1]),2*math.pi*r[1:2],2*math.pi*r[2:3]\n",
    "    q = torch.cat([sq1_v1*torch.sin(v2_2pi), sq1_v1*torch.cos(v2_2pi), \n",
    "                   sqv1  *torch.sin(v3_2pi), sqv1  *torch.cos(v3_2pi)], dim=0).unsqueeze(0)\n",
    "    x_xyz = qrot(q.expand(x_xyz.shape[0],-1), x_xyz).squeeze(0).transpose(0,1)\n",
    "    return (x_xyz, *data[1:])\n",
    "\n",
    "def normalize(data):\n",
    "    sq = False\n",
    "    if data[0].ndim < 3:\n",
    "        data[0].unsqueeze_(0)\n",
    "        #data[4].unsqueeze_(0)\n",
    "        sq = True\n",
    "    x_xyz      = (data[0] - x_xyz_mean)          / x_xyz_std\n",
    "    #x_mulliken = (data[4] - x_qm9_mulliken_mean) / x_qm9_mulliken_std\n",
    "    if sq:\n",
    "        x_xyz.squeeze_(0)\n",
    "        #x_mulliken.squeeze_(0)\n",
    "    return (x_xyz, *data[1:])\n",
    "\n",
    "def canonize(data):\n",
    "    xyz,type,ext,atom = data\n",
    "    mask = (atom == -1).squeeze(0)\n",
    "    i_max_atom = torch.nonzero(atom != -1).max() + 1\n",
    "    mask_atoms = torch.ones((  max_atoms,  max_atoms), dtype=torch.uint8)\n",
    "    zeros      = torch.zeros((i_max_atom, i_max_atom), dtype=torch.uint8)\n",
    "    mask_atoms[:zeros.shape[0],:zeros.shape[1]] = zeros\n",
    "    n_atoms = mask.sum()\n",
    "    xyz[:,mask], type[:,mask],ext[:,mask],atom[:,mask] = 0,-1,1,-1\n",
    "    return (xyz,type,ext,atom,mask_atoms.unsqueeze(0),n_atoms)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Build `data` bunch etc. for fastai"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data = ItemList(items=(MoleculeItem(i,*v) for i,v in \n",
    "                       enumerate(zip(x_xyz,x_type,x_ext,x_atom))),\n",
    "                label_cls=ScalarCouplingItem)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "_, idx_valid_split = train_test_split(range(m.max()+1), test_size=0.1, random_state=13)\n",
    "idx_valid_split = np.argwhere(np.isin(m, idx_valid_split)).flatten()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true,
    "pixiedust": {
     "displayParams": {}
    }
   },
   "outputs": [],
   "source": [
    "data = data.split_by_idx(idx_valid_split)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def l(o): return np.array(y_scalar[o.i], dtype=np.float32)\n",
    "data = data.label_from_func(func=l,label_cls=ScalarCouplingList)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tfms = [normalize, canonize]\n",
    "tta_tfms = list(tfms)\n",
    "tta_tfms.insert(0,random_rotation)\n",
    "data = data.transform((tta_tfms, tfms))#.transform_y(([label_smoothing], None))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DataBunch;\n",
       "\n",
       "Train: LabelList (76502 items)\n",
       "x: ItemList\n",
       "0 6 atoms 20 couplings,1 5 atoms 12 couplings,2 4 atoms 2 couplings,3 4 atoms 4 couplings,5 7 atoms 20 couplings\n",
       "y: ScalarCouplingList\n",
       "20,12,2,4,20\n",
       "Path: .;\n",
       "\n",
       "Valid: LabelList (8501 items)\n",
       "x: ItemList\n",
       "4 9 atoms 54 couplings,9 7 atoms 18 couplings,37 9 atoms 26 couplings,43 14 atoms 88 couplings,47 14 atoms 90 couplings\n",
       "y: ScalarCouplingList\n",
       "54,18,26,88,90\n",
       "Path: .;\n",
       "\n",
       "Test: None"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data=data.databunch()\n",
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-0.7525, -0.0535,  0.7733,  1.6618,  1.5748,  2.1949,  0.5579, -1.2907,\n",
       "         -0.4500,  0.4930,  0.9964,  1.5785,  2.4198,  0.7214,  0.1886,  0.0000,\n",
       "          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,\n",
       "          0.0000,  0.0000,  0.0000,  0.0000,  0.0000],\n",
       "        [ 0.1721,  0.1563, -0.5128, -0.2695,  0.6941,  1.1445,  0.9783,  0.5648,\n",
       "         -0.0066, -1.1574, -0.4511, -0.5532, -0.4465,  1.2539,  1.4654,  0.0000,\n",
       "          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,\n",
       "          0.0000,  0.0000,  0.0000,  0.0000,  0.0000],\n",
       "        [-0.9535, -0.0066, -0.2540,  0.5572,  0.7089,  1.1751,  0.1711, -0.7663,\n",
       "          0.8162, -0.1242, -1.1766,  1.4313,  0.2387, -0.6937,  0.7038,  0.0000,\n",
       "          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,\n",
       "          0.0000,  0.0000,  0.0000,  0.0000,  0.0000]])"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.train_ds[1000][0].data[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1],\n",
       "         [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "          1, 1, 1, 1, 1, 1]]], dtype=torch.uint8)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.train_ds[1000][0].data[4]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#data.train_ds.filter_by_func(lambda item, _: len(item.data[1].cpu().numpy()[item.data[1].cpu().numpy()==0]) == 0)\n",
    "#data.valid_ds.filter_by_func(lambda item, _: len(item.data[1].cpu().numpy()[item.data[1].cpu().numpy()==0]) == 0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model\n",
    "Whole model here, self-contained (needs some cleanup)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class LMAEMetric(LearnerCallback):\n",
    "    _order=-20 # Needs to run before the recorder\n",
    "    def __init__(self, learn, val_only=True):\n",
    "        super().__init__(learn)\n",
    "        self.val_only=val_only\n",
    "        self.metric = LMAEMaskedLoss(return_all=True, exclude_ext=True)\n",
    "\n",
    "    def on_train_begin(self, **kwargs):\n",
    "        if not self.val_only: self.learn.recorder.add_metric_names(['tLMAE'])\n",
    "        self.learn.recorder.add_metric_names(['👉🏻LMAE👈🏻'] + [f'lmae{i}' for i in range(n_types)])\n",
    "            \n",
    "    def on_batch_end(self, train, last_output, last_target, **kwargs):\n",
    "        if self.val_only and train: return \n",
    "        preds,targs = self.preds[int(train)], self.targs[int(train)] # 0 val 1 train\n",
    "        \n",
    "        if preds is None:\n",
    "            targs, preds = last_target.detach(),[t.detach() for t in listify(last_output)]\n",
    "        else:\n",
    "            targs = torch.cat([targs,last_target.detach()], dim=0)\n",
    "            for i,(o,t) in enumerate(zip(last_output, last_target)):\n",
    "                preds[i] = torch.cat([preds[i], o.detach()], dim=0)\n",
    "        self.preds[int(train)], self.targs[int(train)] = preds,targs\n",
    "        \n",
    "    def on_epoch_begin(self, **kwargs):\n",
    "        self.targs, self.preds = [None, None], [None, None]\n",
    "\n",
    "    def on_epoch_end(self, last_metrics, **kwargs):\n",
    "        if True: #(kwargs['epoch'] % max_atoms) == 0:\n",
    "            mets = []\n",
    "            if self.preds[1]: mets.append(self.metric.forward(self.preds[1], self.targs[1])[0]) # just tLMAE\n",
    "            if self.preds[0]: \n",
    "                mets.extend(self.metric.forward(self.preds[0], self.targs[0]))\n",
    "            return add_metrics(last_metrics, mets) if mets else None\n",
    "        else: return None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Activation = Enum('Activation', 'ReLU Swish GeLU')\n",
    "\n",
    "class PositionalEncoding(Module):\n",
    "    \"Encode the position with a sinusoid.\"\n",
    "    def __init__(self, d:int): self.register_buffer('freq', 1 / (10000 ** (torch.arange(0., d, 2.)/d)))\n",
    "\n",
    "    def forward(self, pos:Tensor):\n",
    "        inp = torch.ger(pos, self.freq)\n",
    "        enc = torch.cat([inp.sin(), inp.cos()], dim=-1)\n",
    "        return enc\n",
    "    \n",
    "class GeLU(Module):\n",
    "    def forward(self, x): return 0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3))))\n",
    "\n",
    "class Swish(Module):\n",
    "    def forward(self, x): return x * torch.sigmoid(x)\n",
    "\n",
    "_activ_func = {Activation.ReLU:nn.ReLU(inplace=True), Activation.GeLU:GeLU(), Activation.Swish: Swish()}\n",
    "\n",
    "def feed_forward(d_model:int,d_model_out:int, d_ff:int, ff_p:float=0.,\n",
    "                 act:Activation=Activation.ReLU, double_drop:bool=True):\n",
    "    if d_model==d_model_out:\n",
    "        layers = [nn.Linear(d_model, d_ff), _activ_func[act]]\n",
    "        if double_drop: layers.append(nn.Dropout(ff_p))\n",
    "        return SequentialEx(*layers,nn.Linear(d_ff,d_model_out),nn.Dropout(ff_p),MergeLayer(),nn.LayerNorm(d_model_out))\n",
    "    else:\n",
    "        return SequentialEx(nn.Linear(d_model,d_model_out))#,nn.Dropout(ff_p)),nn.LayerNorm(d_model_out))\n",
    "\n",
    "class MultiHeadAttention(Module):\n",
    "    \"MutiHeadAttention.\"\n",
    "    def __init__(self, n_heads:int, d_model:int, d_head:int=None, resid_p:float=0., attn_p:float=0., bias:bool=True,\n",
    "                 scale:bool=True):\n",
    "        d_head = ifnone(d_head, d_model//n_heads)\n",
    "        self.n_heads,self.d_head,self.scale = n_heads,d_head,scale\n",
    "        self.attention = nn.Linear(d_model, 3 * n_heads * d_head, bias=bias)\n",
    "        self.out = nn.Linear(n_heads * d_head, d_model, bias=bias)\n",
    "        self.drop_att,self.drop_res = nn.Dropout(attn_p),nn.Dropout(resid_p)\n",
    "        self.ln = nn.LayerNorm(d_model)\n",
    "\n",
    "    def forward(self, x:Tensor, mask:Tensor=None, **kwargs):\n",
    "        return self.ln(x + self.drop_res(self.out(self._apply_attention(x, mask=mask, **kwargs))))\n",
    "\n",
    "    def _apply_attention(self, x:Tensor, mask:Tensor=None):\n",
    "        bs,x_len = x.size(0),x.size(1)\n",
    "        wq,wk,wv = torch.chunk(self.attention(x), 3, dim=-1)\n",
    "        wq,wk,wv = map(lambda x:x.view(bs, x.size(1), self.n_heads, self.d_head), (wq,wk,wv))\n",
    "        wq,wk,wv = wq.permute(0, 2, 1, 3),wk.permute(0, 2, 3, 1),wv.permute(0, 2, 1, 3)\n",
    "        attn_score = torch.matmul(wq, wk)\n",
    "        if self.scale: attn_score.div_(self.d_head ** 0.5)\n",
    "        if mask is not None:\n",
    "            minus_inf = -65504 if attn_score.dtype == torch.float16 else -1e9\n",
    "            attn_score = attn_score.masked_fill(mask, minus_inf).type_as(attn_score)\n",
    "        attn_prob = self.drop_att(F.softmax(attn_score, dim=-1))\n",
    "        attn_vec = torch.matmul(attn_prob, wv)\n",
    "        return attn_vec.permute(0, 2, 1, 3).contiguous().contiguous().view(bs, x_len, -1)\n",
    "\n",
    "def _line_shift(x:Tensor, mask:bool=False):\n",
    "    \"Shift the line i of `x` by p-i elements to the left, is `mask` puts 0s on the diagonal.\"\n",
    "    bs,nh,n,p = x.size()\n",
    "    x_pad = torch.cat([x.new_zeros(bs,nh,n,1), x], dim=3)\n",
    "    x_shift = x_pad.view(bs,nh,p + 1,n)[:,:,1:].view_as(x)\n",
    "    if mask: x_shift.mul_(torch.tril(x.new_ones(n,p), p-n)[None,None,])\n",
    "    return x_shift\n",
    "\n",
    "class DecoderLayer(Module):\n",
    "    \"Basic block of a Transformer model.\"\n",
    "    #Can't use Sequential directly cause more than one input...\n",
    "    def __init__(self, n_heads:int, d_model:int, d_model_out:int,d_head:int, d_inner:int, \n",
    "                 resid_p:float=0., attn_p:float=0.,ff_p:float=0.,bias:bool=True, scale:bool=True, \n",
    "                 act:Activation=Activation.ReLU, double_drop:bool=True, attn_cls:Callable=MultiHeadAttention):\n",
    "        self.mhra = attn_cls(n_heads, d_model, d_head, resid_p=resid_p, attn_p=attn_p, bias=bias, scale=scale)\n",
    "        self.ff   = feed_forward(d_model, d_model_out, d_inner, ff_p=ff_p, act=act, double_drop=double_drop)\n",
    "\n",
    "    def forward(self, x:Tensor, mask:Tensor=None, **kwargs): return self.ff(self.mhra(x, mask=mask, **kwargs))\n",
    "\n",
    "class Transformer(Module):\n",
    "    \"Transformer model: https://arxiv.org/abs/1706.03762.\"\n",
    "    def __init__(self, n_layers:int, n_heads:int, d_model:int, d_model_out:int,d_head:int, d_inner:int,\n",
    "                 resid_p:float=0., attn_p:float=0., ff_p:float=0., embed_p:float=0., bias:bool=True, scale:bool=True,\n",
    "                 act:Activation=Activation.ReLU, double_drop:bool=True, attn_cls:Callable=MultiHeadAttention,\n",
    "                 learned_pos_enc:bool=True, mask:bool=True, dense_out:bool=False,final_p:float=0.):\n",
    "        self.mask = mask\n",
    "        self.drop_final = nn.Dropout(final_p)\n",
    "        self.dense_out = dense_out\n",
    "        self.layers = nn.ModuleList([DecoderLayer(\n",
    "            n_heads, d_model, d_model_out if k == n_layers-1 else d_model, d_head, d_inner, \n",
    "            resid_p=resid_p, attn_p=attn_p, ff_p=ff_p, bias=bias, scale=scale, act=act, double_drop=double_drop, \n",
    "            attn_cls=attn_cls) for k in range(n_layers)])\n",
    "\n",
    "    def reset(self): pass\n",
    "\n",
    "    def forward(self, x, mask):\n",
    "        inp = x\n",
    "        if self.dense_out: out = x.new()\n",
    "        for i,layer in enumerate(self.layers):\n",
    "            inp = layer(inp, mask=mask)\n",
    "            inp_d = self.drop_final(inp)\n",
    "            if self.dense_out: out = torch.cat([out, inp_d], dim=-1)\n",
    "        return out if self.dense_out else inp_d\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "class MoleculeTransformer(Module):\n",
    "    def __init__(self,n_layers,n_heads,d_model,embed_p:float=0,final_p:float=0, scalar_p:float=0,\n",
    "                 d_head=None,deep_decoder=False, dense_out=False, **kwargs):\n",
    "        \n",
    "        self.d_model = d_model\n",
    "        d_head = ifnone(d_head, d_model//n_heads)\n",
    "        self.transformer = Transformer(n_layers=n_layers,n_heads=n_heads,d_model=d_model,d_model_out=d_model,\n",
    "                                       d_head=d_head,final_p=final_p,dense_out=dense_out,**kwargs)\n",
    "        \n",
    "        channels_out = d_model*n_layers if dense_out else d_model\n",
    "        channels_out_scalar = channels_out# + n_types + 1\n",
    "        if deep_decoder:\n",
    "            sl = [int(channels_out_scalar/(2**d)) for d in range(int(math.ceil(np.log2(channels_out_scalar/4)-1)))]\n",
    "            self.scalar = nn.Sequential(*(list(itertools.chain.from_iterable(\n",
    "                [[nn.Conv1d(sl[i],sl[i+1],1),nn.ReLU(),nn.BatchNorm1d(sl[i+1])] for i in range(len(sl)-1)])) + \n",
    "                [nn.Conv1d(sl[-1], 1, 1)]))\n",
    "        else:\n",
    "            self.scalar = nn.Conv1d(channels_out_scalar,max_atoms,1)\n",
    "        self.n_type_embedding=n_type_embedding = int((d_model/2)/max_atoms)\n",
    "        n_atom_embedding = d_model - 3 - max_atoms - max_atoms*n_type_embedding\n",
    "\n",
    "        self.atom_embedding = nn.Embedding(len(atoms)+1,n_atom_embedding)\n",
    "        self.type_embedding = nn.Embedding(n_types+1,n_type_embedding)\n",
    "        self.drop_type, self.drop_atom = nn.Dropout(embed_p),nn.Dropout(embed_p)\n",
    "        self.pos_enc = PositionalEncoding(d_model)\n",
    "        \n",
    "        pos = torch.arange(0, max_atoms)\n",
    "        self.pos_one_hot = torch.zeros(1,max_atoms,max_atoms).scatter_(1,pos.expand(1,max_atoms,max_atoms),1.)\n",
    "        # bs,ref0,ref1\n",
    "        # bs,ref0,h\n",
    "        self.drop_mask_type = torch.distributions.Bernoulli(torch.ones(max_atoms,max_atoms)*scalar_p)\n",
    "        \n",
    "    def forward(self,xyz,type,ext,atom,mask_atoms,n_atoms):\n",
    "        bs, _, n_pts = xyz.shape\n",
    "        # idea: count number of t > 0 and wipe out eg 10% of them, to simulate dropout on y (it's equivalent)        \n",
    "        if self.training:\n",
    "            type = type.masked_fill(self.drop_mask_type.sample().to(dtype=torch.uint8,device=xyz.device),-1)\n",
    "\n",
    "        t = self.drop_type(self.type_embedding((type+1).squeeze(1))).transpose(1,2).contiguous().view(bs,n_pts,-1) \n",
    "        # b,1,ref1,ref0,n_type_embedding =>  => b,ref0,ref1,n_type_embedding => b,ref0,ref1 * n_type_embedding\n",
    "        # t_one_hot=torch.zeros(bs,n_types+1,n_pts,n_pts,device=type.device,dtype=x.dtype).scatter_(1,type+1,1.)\n",
    "        # bn,n_types+1,max_atoms(target),max_atoms(reference)\n",
    "        #t_one_hot=t_one_hot.view(bs,-1,n_pts)\n",
    "        \n",
    "        a = self.drop_atom(self.atom_embedding((atom+1).squeeze(1)))\n",
    "\n",
    "        pos_one_hot = torch.eye(max_atoms,device=xyz.device,dtype=xyz.dtype).expand(bs,max_atoms,max_atoms) * \\\n",
    "                     (1-mask_atoms.squeeze(1).type_as(xyz)) # eye matrix masked by positive mask\n",
    "        x = torch.cat([pos_one_hot,xyz.transpose(1,2),a,t], dim=-1) \n",
    "        \n",
    "        x *= math.sqrt(self.d_model) # bn,max_atoms,d_model\n",
    "        x = self.transformer(x,mask_atoms).transpose(1,2) #  bn,max_atoms,d_model > bn,d_model,max_atoms\n",
    "        \n",
    "        scalar = self.scalar(x).view(bs,-1)\n",
    "        \n",
    "        type = type.transpose(2,3).contiguous().view(bs,-1) \n",
    "        # b,1,ref1,ref0,n_type_embedding -> b,1,ref0,ref1*n_type_embedding\n",
    "        \n",
    "        ext  = ext.view(bs,1,-1) #not used for molecular transformer right now\n",
    "        mask= (type != -1).squeeze(1) #b,ref0,ref1*n_type_embedding\n",
    "\n",
    "        return mask.detach(),type[mask].detach(),ext.detach(),scalar[mask].unsqueeze(-1) \n",
    "        # we only pass non-fake couplings to loss fn\n",
    "    \n",
    "    def reset(self): pass"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This callback allows to insert multiple stateful (not averaged) metrics in one pass. Addditionally we could add metrics for train if we want to."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Model instantiation: where's all your TPUs/GPUs when you need a decent hyperparam sweep?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "net, learner = None,None\n",
    "gc.collect()\n",
    "torch.cuda.empty_cache()\n",
    "\n",
    "net = MoleculeTransformer(n_layers=6, n_heads=max_atoms*8,d_model=256,d_inner=2048,\n",
    "                      resid_p=0., attn_p=0., ff_p=0., embed_p=0., final_p=0.,scalar_p=0.5,\n",
    "                      deep_decoder=False, dense_out=False)\n",
    "learner = Learner(data,net, loss_func=LMAEMaskedLoss(),)\n",
    "learner.callbacks.extend([\n",
    "    SaveModelCallback(learner, monitor='👉🏻LMAE👈🏻', mode='min'),\n",
    "    LMAEMetric(learner)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MoleculeTransformer\n",
       "======================================================================\n",
       "Layer (type)         Output Shape         Param #    Trainable \n",
       "======================================================================\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "Linear               [29, 696]            178,872    True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            59,648     True      \n",
       "______________________________________________________________________\n",
       "Dropout              [232, 29, 29]        0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 2048]           526,336    True      \n",
       "______________________________________________________________________\n",
       "ReLU                 [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            524,544    True      \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "MergeLayer           [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 696]            178,872    True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            59,648     True      \n",
       "______________________________________________________________________\n",
       "Dropout              [232, 29, 29]        0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 2048]           526,336    True      \n",
       "______________________________________________________________________\n",
       "ReLU                 [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            524,544    True      \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "MergeLayer           [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 696]            178,872    True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            59,648     True      \n",
       "______________________________________________________________________\n",
       "Dropout              [232, 29, 29]        0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 2048]           526,336    True      \n",
       "______________________________________________________________________\n",
       "ReLU                 [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            524,544    True      \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "MergeLayer           [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 696]            178,872    True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            59,648     True      \n",
       "______________________________________________________________________\n",
       "Dropout              [232, 29, 29]        0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 2048]           526,336    True      \n",
       "______________________________________________________________________\n",
       "ReLU                 [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            524,544    True      \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "MergeLayer           [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 696]            178,872    True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            59,648     True      \n",
       "______________________________________________________________________\n",
       "Dropout              [232, 29, 29]        0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 2048]           526,336    True      \n",
       "______________________________________________________________________\n",
       "ReLU                 [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            524,544    True      \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "MergeLayer           [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 696]            178,872    True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            59,648     True      \n",
       "______________________________________________________________________\n",
       "Dropout              [232, 29, 29]        0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Linear               [29, 2048]           526,336    True      \n",
       "______________________________________________________________________\n",
       "ReLU                 [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 2048]           0          False     \n",
       "______________________________________________________________________\n",
       "Linear               [29, 256]            524,544    True      \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "MergeLayer           [29, 256]            0          False     \n",
       "______________________________________________________________________\n",
       "LayerNorm            [29, 256]            512        True      \n",
       "______________________________________________________________________\n",
       "Conv1d               [29, 29]             7,453      True      \n",
       "______________________________________________________________________\n",
       "Embedding            [29, 108]            648        True      \n",
       "______________________________________________________________________\n",
       "Embedding            [29, 29, 4]          36         True      \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 29, 4]          0          False     \n",
       "______________________________________________________________________\n",
       "Dropout              [29, 108]            0          False     \n",
       "______________________________________________________________________\n",
       "\n",
       "Total params: 7,750,681\n",
       "Total trainable params: 7,750,681\n",
       "Total non-trainable params: 0\n",
       "Optimized with 'torch.optim.adam.Adam', betas=(0.9, 0.99)\n",
       "Using true weight decay as discussed in https://www.fast.ai/2018/07/02/adam-weight-decay/ \n",
       "Loss function : LMAEMaskedLoss\n",
       "======================================================================\n",
       "Callbacks functions applied \n",
       "    SaveModelCallback\n",
       "    LMAEMetric"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learner.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MoleculeTransformer(\n",
       "  (transformer): Transformer(\n",
       "    (drop_final): Dropout(p=0.0)\n",
       "    (layers): ModuleList(\n",
       "      (0): DecoderLayer(\n",
       "        (mhra): MultiHeadAttention(\n",
       "          (attention): Linear(in_features=256, out_features=696, bias=True)\n",
       "          (out): Linear(in_features=232, out_features=256, bias=True)\n",
       "          (drop_att): Dropout(p=0.0)\n",
       "          (drop_res): Dropout(p=0.0)\n",
       "          (ln): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "        )\n",
       "        (ff): SequentialEx(\n",
       "          (layers): ModuleList(\n",
       "            (0): Linear(in_features=256, out_features=2048, bias=True)\n",
       "            (1): ReLU(inplace)\n",
       "            (2): Dropout(p=0.0)\n",
       "            (3): Linear(in_features=2048, out_features=256, bias=True)\n",
       "            (4): Dropout(p=0.0)\n",
       "            (5): MergeLayer()\n",
       "            (6): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (1): DecoderLayer(\n",
       "        (mhra): MultiHeadAttention(\n",
       "          (attention): Linear(in_features=256, out_features=696, bias=True)\n",
       "          (out): Linear(in_features=232, out_features=256, bias=True)\n",
       "          (drop_att): Dropout(p=0.0)\n",
       "          (drop_res): Dropout(p=0.0)\n",
       "          (ln): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "        )\n",
       "        (ff): SequentialEx(\n",
       "          (layers): ModuleList(\n",
       "            (0): Linear(in_features=256, out_features=2048, bias=True)\n",
       "            (1): ReLU(inplace)\n",
       "            (2): Dropout(p=0.0)\n",
       "            (3): Linear(in_features=2048, out_features=256, bias=True)\n",
       "            (4): Dropout(p=0.0)\n",
       "            (5): MergeLayer()\n",
       "            (6): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (2): DecoderLayer(\n",
       "        (mhra): MultiHeadAttention(\n",
       "          (attention): Linear(in_features=256, out_features=696, bias=True)\n",
       "          (out): Linear(in_features=232, out_features=256, bias=True)\n",
       "          (drop_att): Dropout(p=0.0)\n",
       "          (drop_res): Dropout(p=0.0)\n",
       "          (ln): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "        )\n",
       "        (ff): SequentialEx(\n",
       "          (layers): ModuleList(\n",
       "            (0): Linear(in_features=256, out_features=2048, bias=True)\n",
       "            (1): ReLU(inplace)\n",
       "            (2): Dropout(p=0.0)\n",
       "            (3): Linear(in_features=2048, out_features=256, bias=True)\n",
       "            (4): Dropout(p=0.0)\n",
       "            (5): MergeLayer()\n",
       "            (6): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (3): DecoderLayer(\n",
       "        (mhra): MultiHeadAttention(\n",
       "          (attention): Linear(in_features=256, out_features=696, bias=True)\n",
       "          (out): Linear(in_features=232, out_features=256, bias=True)\n",
       "          (drop_att): Dropout(p=0.0)\n",
       "          (drop_res): Dropout(p=0.0)\n",
       "          (ln): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "        )\n",
       "        (ff): SequentialEx(\n",
       "          (layers): ModuleList(\n",
       "            (0): Linear(in_features=256, out_features=2048, bias=True)\n",
       "            (1): ReLU(inplace)\n",
       "            (2): Dropout(p=0.0)\n",
       "            (3): Linear(in_features=2048, out_features=256, bias=True)\n",
       "            (4): Dropout(p=0.0)\n",
       "            (5): MergeLayer()\n",
       "            (6): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (4): DecoderLayer(\n",
       "        (mhra): MultiHeadAttention(\n",
       "          (attention): Linear(in_features=256, out_features=696, bias=True)\n",
       "          (out): Linear(in_features=232, out_features=256, bias=True)\n",
       "          (drop_att): Dropout(p=0.0)\n",
       "          (drop_res): Dropout(p=0.0)\n",
       "          (ln): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "        )\n",
       "        (ff): SequentialEx(\n",
       "          (layers): ModuleList(\n",
       "            (0): Linear(in_features=256, out_features=2048, bias=True)\n",
       "            (1): ReLU(inplace)\n",
       "            (2): Dropout(p=0.0)\n",
       "            (3): Linear(in_features=2048, out_features=256, bias=True)\n",
       "            (4): Dropout(p=0.0)\n",
       "            (5): MergeLayer()\n",
       "            (6): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (5): DecoderLayer(\n",
       "        (mhra): MultiHeadAttention(\n",
       "          (attention): Linear(in_features=256, out_features=696, bias=True)\n",
       "          (out): Linear(in_features=232, out_features=256, bias=True)\n",
       "          (drop_att): Dropout(p=0.0)\n",
       "          (drop_res): Dropout(p=0.0)\n",
       "          (ln): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "        )\n",
       "        (ff): SequentialEx(\n",
       "          (layers): ModuleList(\n",
       "            (0): Linear(in_features=256, out_features=2048, bias=True)\n",
       "            (1): ReLU(inplace)\n",
       "            (2): Dropout(p=0.0)\n",
       "            (3): Linear(in_features=2048, out_features=256, bias=True)\n",
       "            (4): Dropout(p=0.0)\n",
       "            (5): MergeLayer()\n",
       "            (6): LayerNorm(torch.Size([256]), eps=1e-05, elementwise_affine=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "  )\n",
       "  (scalar): Conv1d(256, 29, kernel_size=(1,), stride=(1,))\n",
       "  (atom_embedding): Embedding(6, 108)\n",
       "  (type_embedding): Embedding(9, 4)\n",
       "  (drop_type): Dropout(p=0.0)\n",
       "  (drop_atom): Dropout(p=0.0)\n",
       "  (pos_enc): PositionalEncoding()\n",
       ")"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learner.model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#sub_fname = \"loss-1.4032val-1.4586\" # uncomment or set None to skip loading trained net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NOT loaded!  name 'sub_fname' is not defined\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    print(f\"Attempting to load: {sub_fname}... \", end=\"\")\n",
    "    learner.load(sub_fname, strict=False,with_opt=False)\n",
    "    print(\"Loaded\")\n",
    "except Exception as e:\n",
    "    print(\"NOT loaded! \", e)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# neural weight manual transplant \n",
    "if False:\n",
    "    for name,child in learner.model.named_children():\n",
    "        print(\"CHILD: \",name)\n",
    "        if not (name in ['scalar', 'magnetic', 'dipole', 'potential']):\n",
    "            print(\"FREEZING\")\n",
    "            for param in child.parameters(): param.requires_grad = False\n",
    "        else:\n",
    "            for name,param in child.named_parameters(): \n",
    "                param.requires_grad = True\n",
    "    learner.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner = learner.to_parallel()#.to_fp16()#loss_scale=128, dynamic=False)\n",
    "data.batch_size = int(2048//2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Real loss func. Need to test different auxiliary tasks weights: `magnetic_w`, `dipole_w`, `potential`, weights of indivial `lmae`s: `types_w`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.loss_func = LMAEMaskedLoss(types_w = [1] + [1] * (n_types-1))#, proxy_log=lambda x: x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
      "Min numerical gradient: 6.31E-07\n",
      "Min loss divided by 10: 1.58E-03\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "! rm models/bestmodel.pth\n",
    "learner.lr_find()\n",
    "learner.recorder.plot(suggestion=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>👉🏻LMAE👈🏻</th>\n",
       "      <th>lmae0</th>\n",
       "      <th>lmae1</th>\n",
       "      <th>lmae2</th>\n",
       "      <th>lmae3</th>\n",
       "      <th>lmae4</th>\n",
       "      <th>lmae5</th>\n",
       "      <th>lmae6</th>\n",
       "      <th>lmae7</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.923160</td>\n",
       "      <td>1.900068</td>\n",
       "      <td>1.912084</td>\n",
       "      <td>4.546370</td>\n",
       "      <td>2.508763</td>\n",
       "      <td>3.840623</td>\n",
       "      <td>1.114149</td>\n",
       "      <td>1.128598</td>\n",
       "      <td>1.156542</td>\n",
       "      <td>1.077546</td>\n",
       "      <td>-0.075916</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1.896336</td>\n",
       "      <td>1.850479</td>\n",
       "      <td>1.861903</td>\n",
       "      <td>4.557377</td>\n",
       "      <td>2.257590</td>\n",
       "      <td>3.837995</td>\n",
       "      <td>1.112092</td>\n",
       "      <td>1.016414</td>\n",
       "      <td>1.170805</td>\n",
       "      <td>1.026037</td>\n",
       "      <td>-0.083090</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>1.859745</td>\n",
       "      <td>1.796040</td>\n",
       "      <td>1.807070</td>\n",
       "      <td>4.558851</td>\n",
       "      <td>2.019801</td>\n",
       "      <td>3.836298</td>\n",
       "      <td>1.117859</td>\n",
       "      <td>0.945417</td>\n",
       "      <td>1.104309</td>\n",
       "      <td>0.974670</td>\n",
       "      <td>-0.100642</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>1.827177</td>\n",
       "      <td>1.750030</td>\n",
       "      <td>1.759744</td>\n",
       "      <td>4.556438</td>\n",
       "      <td>1.687487</td>\n",
       "      <td>3.835960</td>\n",
       "      <td>1.098810</td>\n",
       "      <td>0.926714</td>\n",
       "      <td>1.098469</td>\n",
       "      <td>0.966579</td>\n",
       "      <td>-0.092510</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>1.798473</td>\n",
       "      <td>1.710036</td>\n",
       "      <td>1.718670</td>\n",
       "      <td>4.552663</td>\n",
       "      <td>1.439460</td>\n",
       "      <td>3.832593</td>\n",
       "      <td>1.071974</td>\n",
       "      <td>0.915737</td>\n",
       "      <td>1.079052</td>\n",
       "      <td>0.955354</td>\n",
       "      <td>-0.097473</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>1.771950</td>\n",
       "      <td>1.678105</td>\n",
       "      <td>1.686272</td>\n",
       "      <td>4.548733</td>\n",
       "      <td>1.240661</td>\n",
       "      <td>3.830760</td>\n",
       "      <td>1.050669</td>\n",
       "      <td>0.904645</td>\n",
       "      <td>1.081646</td>\n",
       "      <td>0.943697</td>\n",
       "      <td>-0.110636</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>1.741322</td>\n",
       "      <td>1.651284</td>\n",
       "      <td>1.659049</td>\n",
       "      <td>4.545418</td>\n",
       "      <td>1.076431</td>\n",
       "      <td>3.830232</td>\n",
       "      <td>1.024180</td>\n",
       "      <td>0.885470</td>\n",
       "      <td>1.084973</td>\n",
       "      <td>0.941262</td>\n",
       "      <td>-0.115577</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>1.718283</td>\n",
       "      <td>1.636681</td>\n",
       "      <td>1.643380</td>\n",
       "      <td>4.539898</td>\n",
       "      <td>0.978736</td>\n",
       "      <td>3.827724</td>\n",
       "      <td>0.993823</td>\n",
       "      <td>0.887674</td>\n",
       "      <td>1.086348</td>\n",
       "      <td>0.920729</td>\n",
       "      <td>-0.087894</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>1.690624</td>\n",
       "      <td>1.613813</td>\n",
       "      <td>1.619510</td>\n",
       "      <td>4.537863</td>\n",
       "      <td>0.866910</td>\n",
       "      <td>3.821663</td>\n",
       "      <td>0.964368</td>\n",
       "      <td>0.872888</td>\n",
       "      <td>1.078340</td>\n",
       "      <td>0.927007</td>\n",
       "      <td>-0.112956</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>1.658581</td>\n",
       "      <td>1.583372</td>\n",
       "      <td>1.588312</td>\n",
       "      <td>4.534563</td>\n",
       "      <td>0.756868</td>\n",
       "      <td>3.813275</td>\n",
       "      <td>0.928321</td>\n",
       "      <td>0.854972</td>\n",
       "      <td>1.031071</td>\n",
       "      <td>0.916354</td>\n",
       "      <td>-0.128929</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>1.635653</td>\n",
       "      <td>1.576292</td>\n",
       "      <td>1.580750</td>\n",
       "      <td>4.531662</td>\n",
       "      <td>0.735059</td>\n",
       "      <td>3.810500</td>\n",
       "      <td>0.897579</td>\n",
       "      <td>0.847236</td>\n",
       "      <td>1.018001</td>\n",
       "      <td>0.922411</td>\n",
       "      <td>-0.116451</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>1.605840</td>\n",
       "      <td>1.543092</td>\n",
       "      <td>1.545585</td>\n",
       "      <td>4.525040</td>\n",
       "      <td>0.665089</td>\n",
       "      <td>3.784478</td>\n",
       "      <td>0.753761</td>\n",
       "      <td>0.841625</td>\n",
       "      <td>0.998922</td>\n",
       "      <td>0.903637</td>\n",
       "      <td>-0.107871</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>1.578403</td>\n",
       "      <td>1.523698</td>\n",
       "      <td>1.525374</td>\n",
       "      <td>4.517993</td>\n",
       "      <td>0.654212</td>\n",
       "      <td>3.768739</td>\n",
       "      <td>0.709732</td>\n",
       "      <td>0.820317</td>\n",
       "      <td>0.963693</td>\n",
       "      <td>0.887964</td>\n",
       "      <td>-0.119659</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>1.554033</td>\n",
       "      <td>1.492192</td>\n",
       "      <td>1.493657</td>\n",
       "      <td>4.510228</td>\n",
       "      <td>0.582665</td>\n",
       "      <td>3.730137</td>\n",
       "      <td>0.608148</td>\n",
       "      <td>0.799588</td>\n",
       "      <td>0.952835</td>\n",
       "      <td>0.884790</td>\n",
       "      <td>-0.119139</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>1.528659</td>\n",
       "      <td>1.469144</td>\n",
       "      <td>1.470126</td>\n",
       "      <td>4.500204</td>\n",
       "      <td>0.526358</td>\n",
       "      <td>3.688287</td>\n",
       "      <td>0.577462</td>\n",
       "      <td>0.792270</td>\n",
       "      <td>0.935497</td>\n",
       "      <td>0.867491</td>\n",
       "      <td>-0.126560</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>1.505099</td>\n",
       "      <td>1.450884</td>\n",
       "      <td>1.451472</td>\n",
       "      <td>4.489229</td>\n",
       "      <td>0.511316</td>\n",
       "      <td>3.653171</td>\n",
       "      <td>0.524555</td>\n",
       "      <td>0.776972</td>\n",
       "      <td>0.932698</td>\n",
       "      <td>0.875827</td>\n",
       "      <td>-0.151992</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>1.479709</td>\n",
       "      <td>1.432040</td>\n",
       "      <td>1.432857</td>\n",
       "      <td>4.476369</td>\n",
       "      <td>0.484057</td>\n",
       "      <td>3.605438</td>\n",
       "      <td>0.537327</td>\n",
       "      <td>0.763003</td>\n",
       "      <td>0.912035</td>\n",
       "      <td>0.856149</td>\n",
       "      <td>-0.171526</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>1.454451</td>\n",
       "      <td>1.399424</td>\n",
       "      <td>1.400345</td>\n",
       "      <td>4.462112</td>\n",
       "      <td>0.446165</td>\n",
       "      <td>3.538923</td>\n",
       "      <td>0.392389</td>\n",
       "      <td>0.761661</td>\n",
       "      <td>0.916879</td>\n",
       "      <td>0.846439</td>\n",
       "      <td>-0.161806</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>1.434157</td>\n",
       "      <td>1.378786</td>\n",
       "      <td>1.379282</td>\n",
       "      <td>4.448698</td>\n",
       "      <td>0.418973</td>\n",
       "      <td>3.502500</td>\n",
       "      <td>0.400068</td>\n",
       "      <td>0.732957</td>\n",
       "      <td>0.887636</td>\n",
       "      <td>0.847447</td>\n",
       "      <td>-0.204028</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>1.404399</td>\n",
       "      <td>1.360919</td>\n",
       "      <td>1.361939</td>\n",
       "      <td>4.437576</td>\n",
       "      <td>0.387527</td>\n",
       "      <td>3.433676</td>\n",
       "      <td>0.406672</td>\n",
       "      <td>0.709840</td>\n",
       "      <td>0.874388</td>\n",
       "      <td>0.837441</td>\n",
       "      <td>-0.191606</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>1.375556</td>\n",
       "      <td>1.336550</td>\n",
       "      <td>1.338030</td>\n",
       "      <td>4.424209</td>\n",
       "      <td>0.380232</td>\n",
       "      <td>3.384842</td>\n",
       "      <td>0.335051</td>\n",
       "      <td>0.704217</td>\n",
       "      <td>0.859839</td>\n",
       "      <td>0.833234</td>\n",
       "      <td>-0.217383</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>1.350494</td>\n",
       "      <td>1.290910</td>\n",
       "      <td>1.293265</td>\n",
       "      <td>4.402764</td>\n",
       "      <td>0.291132</td>\n",
       "      <td>3.284025</td>\n",
       "      <td>0.269266</td>\n",
       "      <td>0.674918</td>\n",
       "      <td>0.831945</td>\n",
       "      <td>0.807365</td>\n",
       "      <td>-0.215294</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>1.320101</td>\n",
       "      <td>1.265374</td>\n",
       "      <td>1.269257</td>\n",
       "      <td>4.387833</td>\n",
       "      <td>0.312718</td>\n",
       "      <td>3.184059</td>\n",
       "      <td>0.257383</td>\n",
       "      <td>0.673300</td>\n",
       "      <td>0.794573</td>\n",
       "      <td>0.803550</td>\n",
       "      <td>-0.259359</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>1.276994</td>\n",
       "      <td>1.207309</td>\n",
       "      <td>1.211983</td>\n",
       "      <td>4.360477</td>\n",
       "      <td>0.236214</td>\n",
       "      <td>3.003839</td>\n",
       "      <td>0.227907</td>\n",
       "      <td>0.635018</td>\n",
       "      <td>0.706500</td>\n",
       "      <td>0.781361</td>\n",
       "      <td>-0.255452</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>1.229009</td>\n",
       "      <td>1.165063</td>\n",
       "      <td>1.171527</td>\n",
       "      <td>4.322108</td>\n",
       "      <td>0.239637</td>\n",
       "      <td>2.754847</td>\n",
       "      <td>0.204880</td>\n",
       "      <td>0.634630</td>\n",
       "      <td>0.661185</td>\n",
       "      <td>0.779927</td>\n",
       "      <td>-0.224996</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>1.182553</td>\n",
       "      <td>1.113301</td>\n",
       "      <td>1.118619</td>\n",
       "      <td>4.293884</td>\n",
       "      <td>0.183382</td>\n",
       "      <td>2.532362</td>\n",
       "      <td>0.176095</td>\n",
       "      <td>0.599156</td>\n",
       "      <td>0.625382</td>\n",
       "      <td>0.760567</td>\n",
       "      <td>-0.221874</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>1.132429</td>\n",
       "      <td>1.067570</td>\n",
       "      <td>1.071469</td>\n",
       "      <td>4.267035</td>\n",
       "      <td>0.200034</td>\n",
       "      <td>2.274621</td>\n",
       "      <td>0.135760</td>\n",
       "      <td>0.591674</td>\n",
       "      <td>0.597163</td>\n",
       "      <td>0.759582</td>\n",
       "      <td>-0.254119</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>1.089079</td>\n",
       "      <td>1.033457</td>\n",
       "      <td>1.035177</td>\n",
       "      <td>4.237643</td>\n",
       "      <td>0.162712</td>\n",
       "      <td>2.065954</td>\n",
       "      <td>0.127348</td>\n",
       "      <td>0.585096</td>\n",
       "      <td>0.582826</td>\n",
       "      <td>0.753500</td>\n",
       "      <td>-0.233661</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>1.048066</td>\n",
       "      <td>1.002485</td>\n",
       "      <td>1.001775</td>\n",
       "      <td>4.196737</td>\n",
       "      <td>0.164686</td>\n",
       "      <td>1.795438</td>\n",
       "      <td>0.153116</td>\n",
       "      <td>0.592058</td>\n",
       "      <td>0.615632</td>\n",
       "      <td>0.754450</td>\n",
       "      <td>-0.257918</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>1.001259</td>\n",
       "      <td>0.949645</td>\n",
       "      <td>0.947911</td>\n",
       "      <td>4.162156</td>\n",
       "      <td>0.114286</td>\n",
       "      <td>1.576030</td>\n",
       "      <td>0.134317</td>\n",
       "      <td>0.567260</td>\n",
       "      <td>0.570621</td>\n",
       "      <td>0.747603</td>\n",
       "      <td>-0.288980</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.959371</td>\n",
       "      <td>0.925667</td>\n",
       "      <td>0.923090</td>\n",
       "      <td>4.134613</td>\n",
       "      <td>0.104707</td>\n",
       "      <td>1.467341</td>\n",
       "      <td>0.091999</td>\n",
       "      <td>0.572832</td>\n",
       "      <td>0.544392</td>\n",
       "      <td>0.734682</td>\n",
       "      <td>-0.265848</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.926525</td>\n",
       "      <td>0.890107</td>\n",
       "      <td>0.886332</td>\n",
       "      <td>4.110307</td>\n",
       "      <td>0.062047</td>\n",
       "      <td>1.346549</td>\n",
       "      <td>0.043494</td>\n",
       "      <td>0.550891</td>\n",
       "      <td>0.530822</td>\n",
       "      <td>0.732101</td>\n",
       "      <td>-0.285552</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>32</td>\n",
       "      <td>0.896205</td>\n",
       "      <td>0.851770</td>\n",
       "      <td>0.847468</td>\n",
       "      <td>4.079433</td>\n",
       "      <td>0.035504</td>\n",
       "      <td>1.233504</td>\n",
       "      <td>0.003040</td>\n",
       "      <td>0.513531</td>\n",
       "      <td>0.495442</td>\n",
       "      <td>0.720446</td>\n",
       "      <td>-0.301157</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>33</td>\n",
       "      <td>0.869929</td>\n",
       "      <td>0.840162</td>\n",
       "      <td>0.833912</td>\n",
       "      <td>4.060366</td>\n",
       "      <td>-0.011693</td>\n",
       "      <td>1.216180</td>\n",
       "      <td>-0.009054</td>\n",
       "      <td>0.513938</td>\n",
       "      <td>0.483147</td>\n",
       "      <td>0.718463</td>\n",
       "      <td>-0.300052</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>34</td>\n",
       "      <td>0.841553</td>\n",
       "      <td>0.818831</td>\n",
       "      <td>0.812987</td>\n",
       "      <td>4.026773</td>\n",
       "      <td>0.000408</td>\n",
       "      <td>1.150611</td>\n",
       "      <td>-0.044954</td>\n",
       "      <td>0.510355</td>\n",
       "      <td>0.484837</td>\n",
       "      <td>0.703743</td>\n",
       "      <td>-0.327877</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>35</td>\n",
       "      <td>0.813666</td>\n",
       "      <td>0.793607</td>\n",
       "      <td>0.788600</td>\n",
       "      <td>4.002845</td>\n",
       "      <td>-0.012094</td>\n",
       "      <td>1.090879</td>\n",
       "      <td>-0.066805</td>\n",
       "      <td>0.502641</td>\n",
       "      <td>0.465445</td>\n",
       "      <td>0.696949</td>\n",
       "      <td>-0.371058</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>36</td>\n",
       "      <td>0.794150</td>\n",
       "      <td>0.780424</td>\n",
       "      <td>0.777033</td>\n",
       "      <td>3.982197</td>\n",
       "      <td>-0.021729</td>\n",
       "      <td>1.073963</td>\n",
       "      <td>-0.063112</td>\n",
       "      <td>0.501435</td>\n",
       "      <td>0.435522</td>\n",
       "      <td>0.702869</td>\n",
       "      <td>-0.394880</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>37</td>\n",
       "      <td>0.777467</td>\n",
       "      <td>0.780895</td>\n",
       "      <td>0.776568</td>\n",
       "      <td>3.945455</td>\n",
       "      <td>0.026655</td>\n",
       "      <td>1.048036</td>\n",
       "      <td>-0.084732</td>\n",
       "      <td>0.504158</td>\n",
       "      <td>0.440998</td>\n",
       "      <td>0.693561</td>\n",
       "      <td>-0.361583</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>38</td>\n",
       "      <td>0.752661</td>\n",
       "      <td>0.741220</td>\n",
       "      <td>0.735421</td>\n",
       "      <td>3.918063</td>\n",
       "      <td>-0.073599</td>\n",
       "      <td>0.949692</td>\n",
       "      <td>-0.115942</td>\n",
       "      <td>0.481892</td>\n",
       "      <td>0.451345</td>\n",
       "      <td>0.683627</td>\n",
       "      <td>-0.411711</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>39</td>\n",
       "      <td>0.737610</td>\n",
       "      <td>0.730487</td>\n",
       "      <td>0.725321</td>\n",
       "      <td>3.890646</td>\n",
       "      <td>-0.073409</td>\n",
       "      <td>0.967242</td>\n",
       "      <td>-0.127962</td>\n",
       "      <td>0.459109</td>\n",
       "      <td>0.404667</td>\n",
       "      <td>0.684004</td>\n",
       "      <td>-0.401726</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>40</td>\n",
       "      <td>0.713667</td>\n",
       "      <td>0.705468</td>\n",
       "      <td>0.699440</td>\n",
       "      <td>3.868979</td>\n",
       "      <td>-0.108646</td>\n",
       "      <td>0.875201</td>\n",
       "      <td>-0.136216</td>\n",
       "      <td>0.437989</td>\n",
       "      <td>0.404204</td>\n",
       "      <td>0.678744</td>\n",
       "      <td>-0.424735</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>41</td>\n",
       "      <td>0.686174</td>\n",
       "      <td>0.688385</td>\n",
       "      <td>0.684387</td>\n",
       "      <td>3.831050</td>\n",
       "      <td>-0.141274</td>\n",
       "      <td>0.936754</td>\n",
       "      <td>-0.180390</td>\n",
       "      <td>0.427133</td>\n",
       "      <td>0.377154</td>\n",
       "      <td>0.666324</td>\n",
       "      <td>-0.441655</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>42</td>\n",
       "      <td>0.658717</td>\n",
       "      <td>0.668715</td>\n",
       "      <td>0.662789</td>\n",
       "      <td>3.792890</td>\n",
       "      <td>-0.135644</td>\n",
       "      <td>0.882554</td>\n",
       "      <td>-0.199588</td>\n",
       "      <td>0.418912</td>\n",
       "      <td>0.356150</td>\n",
       "      <td>0.661509</td>\n",
       "      <td>-0.474473</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>43</td>\n",
       "      <td>0.639043</td>\n",
       "      <td>0.635865</td>\n",
       "      <td>0.629174</td>\n",
       "      <td>3.754600</td>\n",
       "      <td>-0.178921</td>\n",
       "      <td>0.786243</td>\n",
       "      <td>-0.239193</td>\n",
       "      <td>0.397063</td>\n",
       "      <td>0.341288</td>\n",
       "      <td>0.648758</td>\n",
       "      <td>-0.476448</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>44</td>\n",
       "      <td>0.618091</td>\n",
       "      <td>0.649731</td>\n",
       "      <td>0.644373</td>\n",
       "      <td>3.715542</td>\n",
       "      <td>-0.137483</td>\n",
       "      <td>0.835741</td>\n",
       "      <td>-0.187603</td>\n",
       "      <td>0.403854</td>\n",
       "      <td>0.358125</td>\n",
       "      <td>0.654557</td>\n",
       "      <td>-0.487752</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>45</td>\n",
       "      <td>0.599943</td>\n",
       "      <td>0.608331</td>\n",
       "      <td>0.601685</td>\n",
       "      <td>3.675800</td>\n",
       "      <td>-0.195145</td>\n",
       "      <td>0.757495</td>\n",
       "      <td>-0.269838</td>\n",
       "      <td>0.372967</td>\n",
       "      <td>0.326002</td>\n",
       "      <td>0.638743</td>\n",
       "      <td>-0.492547</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>46</td>\n",
       "      <td>0.578111</td>\n",
       "      <td>0.599526</td>\n",
       "      <td>0.592837</td>\n",
       "      <td>3.638317</td>\n",
       "      <td>-0.218748</td>\n",
       "      <td>0.723505</td>\n",
       "      <td>-0.229977</td>\n",
       "      <td>0.389943</td>\n",
       "      <td>0.317727</td>\n",
       "      <td>0.630548</td>\n",
       "      <td>-0.508618</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>47</td>\n",
       "      <td>0.551933</td>\n",
       "      <td>0.583466</td>\n",
       "      <td>0.577352</td>\n",
       "      <td>3.577370</td>\n",
       "      <td>-0.171922</td>\n",
       "      <td>0.704550</td>\n",
       "      <td>-0.295828</td>\n",
       "      <td>0.385299</td>\n",
       "      <td>0.310126</td>\n",
       "      <td>0.634760</td>\n",
       "      <td>-0.525538</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>48</td>\n",
       "      <td>0.531069</td>\n",
       "      <td>0.582183</td>\n",
       "      <td>0.574924</td>\n",
       "      <td>3.529681</td>\n",
       "      <td>-0.199786</td>\n",
       "      <td>0.722204</td>\n",
       "      <td>-0.273836</td>\n",
       "      <td>0.382596</td>\n",
       "      <td>0.304137</td>\n",
       "      <td>0.623811</td>\n",
       "      <td>-0.489416</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>49</td>\n",
       "      <td>0.513899</td>\n",
       "      <td>0.542340</td>\n",
       "      <td>0.537162</td>\n",
       "      <td>3.441041</td>\n",
       "      <td>-0.257738</td>\n",
       "      <td>0.720748</td>\n",
       "      <td>-0.330953</td>\n",
       "      <td>0.353444</td>\n",
       "      <td>0.290167</td>\n",
       "      <td>0.616206</td>\n",
       "      <td>-0.535620</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>50</td>\n",
       "      <td>0.494980</td>\n",
       "      <td>0.520818</td>\n",
       "      <td>0.513895</td>\n",
       "      <td>3.377208</td>\n",
       "      <td>-0.263081</td>\n",
       "      <td>0.640757</td>\n",
       "      <td>-0.336842</td>\n",
       "      <td>0.357402</td>\n",
       "      <td>0.277069</td>\n",
       "      <td>0.610570</td>\n",
       "      <td>-0.551923</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>51</td>\n",
       "      <td>0.475942</td>\n",
       "      <td>0.513250</td>\n",
       "      <td>0.506271</td>\n",
       "      <td>3.282820</td>\n",
       "      <td>-0.231316</td>\n",
       "      <td>0.623126</td>\n",
       "      <td>-0.338715</td>\n",
       "      <td>0.357980</td>\n",
       "      <td>0.295211</td>\n",
       "      <td>0.617698</td>\n",
       "      <td>-0.556633</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>52</td>\n",
       "      <td>0.443097</td>\n",
       "      <td>0.476312</td>\n",
       "      <td>0.468950</td>\n",
       "      <td>3.190165</td>\n",
       "      <td>-0.273897</td>\n",
       "      <td>0.640706</td>\n",
       "      <td>-0.396251</td>\n",
       "      <td>0.329230</td>\n",
       "      <td>0.253318</td>\n",
       "      <td>0.600126</td>\n",
       "      <td>-0.591797</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>53</td>\n",
       "      <td>0.416419</td>\n",
       "      <td>0.455069</td>\n",
       "      <td>0.447073</td>\n",
       "      <td>3.094522</td>\n",
       "      <td>-0.263357</td>\n",
       "      <td>0.567210</td>\n",
       "      <td>-0.408348</td>\n",
       "      <td>0.327265</td>\n",
       "      <td>0.240562</td>\n",
       "      <td>0.591047</td>\n",
       "      <td>-0.572317</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>54</td>\n",
       "      <td>0.381079</td>\n",
       "      <td>0.427065</td>\n",
       "      <td>0.418513</td>\n",
       "      <td>2.969312</td>\n",
       "      <td>-0.293837</td>\n",
       "      <td>0.532459</td>\n",
       "      <td>-0.398403</td>\n",
       "      <td>0.318014</td>\n",
       "      <td>0.231471</td>\n",
       "      <td>0.581199</td>\n",
       "      <td>-0.592115</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>55</td>\n",
       "      <td>0.355929</td>\n",
       "      <td>0.410520</td>\n",
       "      <td>0.404347</td>\n",
       "      <td>2.867131</td>\n",
       "      <td>-0.314790</td>\n",
       "      <td>0.581885</td>\n",
       "      <td>-0.418207</td>\n",
       "      <td>0.319087</td>\n",
       "      <td>0.232009</td>\n",
       "      <td>0.577342</td>\n",
       "      <td>-0.609682</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>56</td>\n",
       "      <td>0.319924</td>\n",
       "      <td>0.370838</td>\n",
       "      <td>0.362753</td>\n",
       "      <td>2.770278</td>\n",
       "      <td>-0.324259</td>\n",
       "      <td>0.500918</td>\n",
       "      <td>-0.450474</td>\n",
       "      <td>0.287149</td>\n",
       "      <td>0.180541</td>\n",
       "      <td>0.567415</td>\n",
       "      <td>-0.629545</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>57</td>\n",
       "      <td>0.305417</td>\n",
       "      <td>0.352242</td>\n",
       "      <td>0.343249</td>\n",
       "      <td>2.674159</td>\n",
       "      <td>-0.346094</td>\n",
       "      <td>0.499649</td>\n",
       "      <td>-0.461490</td>\n",
       "      <td>0.271018</td>\n",
       "      <td>0.182382</td>\n",
       "      <td>0.566523</td>\n",
       "      <td>-0.640153</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>58</td>\n",
       "      <td>0.274738</td>\n",
       "      <td>0.333500</td>\n",
       "      <td>0.325628</td>\n",
       "      <td>2.596433</td>\n",
       "      <td>-0.330931</td>\n",
       "      <td>0.466013</td>\n",
       "      <td>-0.487758</td>\n",
       "      <td>0.271886</td>\n",
       "      <td>0.177050</td>\n",
       "      <td>0.566962</td>\n",
       "      <td>-0.654633</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59</td>\n",
       "      <td>0.246081</td>\n",
       "      <td>0.318508</td>\n",
       "      <td>0.311064</td>\n",
       "      <td>2.525644</td>\n",
       "      <td>-0.334234</td>\n",
       "      <td>0.453944</td>\n",
       "      <td>-0.479581</td>\n",
       "      <td>0.282558</td>\n",
       "      <td>0.165140</td>\n",
       "      <td>0.554410</td>\n",
       "      <td>-0.679372</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>60</td>\n",
       "      <td>0.221007</td>\n",
       "      <td>0.293738</td>\n",
       "      <td>0.286029</td>\n",
       "      <td>2.457074</td>\n",
       "      <td>-0.405571</td>\n",
       "      <td>0.452056</td>\n",
       "      <td>-0.496111</td>\n",
       "      <td>0.254108</td>\n",
       "      <td>0.150738</td>\n",
       "      <td>0.547757</td>\n",
       "      <td>-0.671820</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>61</td>\n",
       "      <td>0.196667</td>\n",
       "      <td>0.270557</td>\n",
       "      <td>0.262940</td>\n",
       "      <td>2.412505</td>\n",
       "      <td>-0.400581</td>\n",
       "      <td>0.390070</td>\n",
       "      <td>-0.522853</td>\n",
       "      <td>0.245712</td>\n",
       "      <td>0.132310</td>\n",
       "      <td>0.537998</td>\n",
       "      <td>-0.691642</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>62</td>\n",
       "      <td>0.171061</td>\n",
       "      <td>0.267690</td>\n",
       "      <td>0.259408</td>\n",
       "      <td>2.371992</td>\n",
       "      <td>-0.393722</td>\n",
       "      <td>0.429463</td>\n",
       "      <td>-0.528606</td>\n",
       "      <td>0.244286</td>\n",
       "      <td>0.119027</td>\n",
       "      <td>0.534008</td>\n",
       "      <td>-0.701181</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>63</td>\n",
       "      <td>0.150992</td>\n",
       "      <td>0.251073</td>\n",
       "      <td>0.244136</td>\n",
       "      <td>2.314713</td>\n",
       "      <td>-0.431453</td>\n",
       "      <td>0.396318</td>\n",
       "      <td>-0.525832</td>\n",
       "      <td>0.241684</td>\n",
       "      <td>0.121364</td>\n",
       "      <td>0.537210</td>\n",
       "      <td>-0.700913</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>64</td>\n",
       "      <td>0.122002</td>\n",
       "      <td>0.221025</td>\n",
       "      <td>0.212638</td>\n",
       "      <td>2.265645</td>\n",
       "      <td>-0.463512</td>\n",
       "      <td>0.349820</td>\n",
       "      <td>-0.568093</td>\n",
       "      <td>0.226168</td>\n",
       "      <td>0.089076</td>\n",
       "      <td>0.524796</td>\n",
       "      <td>-0.722793</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>65</td>\n",
       "      <td>0.110052</td>\n",
       "      <td>0.224859</td>\n",
       "      <td>0.217427</td>\n",
       "      <td>2.229676</td>\n",
       "      <td>-0.444257</td>\n",
       "      <td>0.371289</td>\n",
       "      <td>-0.558361</td>\n",
       "      <td>0.233818</td>\n",
       "      <td>0.103509</td>\n",
       "      <td>0.525515</td>\n",
       "      <td>-0.721775</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>66</td>\n",
       "      <td>0.084297</td>\n",
       "      <td>0.195381</td>\n",
       "      <td>0.187612</td>\n",
       "      <td>2.178961</td>\n",
       "      <td>-0.492575</td>\n",
       "      <td>0.358591</td>\n",
       "      <td>-0.595087</td>\n",
       "      <td>0.200598</td>\n",
       "      <td>0.072052</td>\n",
       "      <td>0.515727</td>\n",
       "      <td>-0.737371</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>67</td>\n",
       "      <td>0.067775</td>\n",
       "      <td>0.180590</td>\n",
       "      <td>0.172981</td>\n",
       "      <td>2.150645</td>\n",
       "      <td>-0.509207</td>\n",
       "      <td>0.306095</td>\n",
       "      <td>-0.607633</td>\n",
       "      <td>0.202909</td>\n",
       "      <td>0.068772</td>\n",
       "      <td>0.516757</td>\n",
       "      <td>-0.744492</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>68</td>\n",
       "      <td>0.040582</td>\n",
       "      <td>0.164882</td>\n",
       "      <td>0.156736</td>\n",
       "      <td>2.106309</td>\n",
       "      <td>-0.505606</td>\n",
       "      <td>0.290759</td>\n",
       "      <td>-0.631990</td>\n",
       "      <td>0.195551</td>\n",
       "      <td>0.060496</td>\n",
       "      <td>0.509099</td>\n",
       "      <td>-0.770730</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>69</td>\n",
       "      <td>0.018287</td>\n",
       "      <td>0.151402</td>\n",
       "      <td>0.142792</td>\n",
       "      <td>2.073548</td>\n",
       "      <td>-0.528272</td>\n",
       "      <td>0.255634</td>\n",
       "      <td>-0.625706</td>\n",
       "      <td>0.191737</td>\n",
       "      <td>0.044957</td>\n",
       "      <td>0.503400</td>\n",
       "      <td>-0.772964</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>70</td>\n",
       "      <td>-0.001501</td>\n",
       "      <td>0.140778</td>\n",
       "      <td>0.132635</td>\n",
       "      <td>2.039958</td>\n",
       "      <td>-0.523613</td>\n",
       "      <td>0.262072</td>\n",
       "      <td>-0.648512</td>\n",
       "      <td>0.178252</td>\n",
       "      <td>0.032700</td>\n",
       "      <td>0.499754</td>\n",
       "      <td>-0.779527</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>71</td>\n",
       "      <td>-0.024027</td>\n",
       "      <td>0.122707</td>\n",
       "      <td>0.114453</td>\n",
       "      <td>2.006583</td>\n",
       "      <td>-0.563375</td>\n",
       "      <td>0.248571</td>\n",
       "      <td>-0.671108</td>\n",
       "      <td>0.167446</td>\n",
       "      <td>0.030208</td>\n",
       "      <td>0.492569</td>\n",
       "      <td>-0.795266</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>72</td>\n",
       "      <td>-0.035263</td>\n",
       "      <td>0.118675</td>\n",
       "      <td>0.111007</td>\n",
       "      <td>1.981663</td>\n",
       "      <td>-0.551369</td>\n",
       "      <td>0.265290</td>\n",
       "      <td>-0.679104</td>\n",
       "      <td>0.167948</td>\n",
       "      <td>0.015241</td>\n",
       "      <td>0.491593</td>\n",
       "      <td>-0.803202</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>73</td>\n",
       "      <td>-0.058930</td>\n",
       "      <td>0.099653</td>\n",
       "      <td>0.091156</td>\n",
       "      <td>1.951330</td>\n",
       "      <td>-0.578246</td>\n",
       "      <td>0.204564</td>\n",
       "      <td>-0.694910</td>\n",
       "      <td>0.160098</td>\n",
       "      <td>0.010551</td>\n",
       "      <td>0.485414</td>\n",
       "      <td>-0.809549</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>74</td>\n",
       "      <td>-0.081367</td>\n",
       "      <td>0.087367</td>\n",
       "      <td>0.078679</td>\n",
       "      <td>1.925260</td>\n",
       "      <td>-0.585081</td>\n",
       "      <td>0.195700</td>\n",
       "      <td>-0.713296</td>\n",
       "      <td>0.145154</td>\n",
       "      <td>-0.002489</td>\n",
       "      <td>0.480436</td>\n",
       "      <td>-0.816251</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>75</td>\n",
       "      <td>-0.098248</td>\n",
       "      <td>0.075447</td>\n",
       "      <td>0.066680</td>\n",
       "      <td>1.914016</td>\n",
       "      <td>-0.600206</td>\n",
       "      <td>0.158778</td>\n",
       "      <td>-0.724476</td>\n",
       "      <td>0.146019</td>\n",
       "      <td>-0.014538</td>\n",
       "      <td>0.478109</td>\n",
       "      <td>-0.824263</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>76</td>\n",
       "      <td>-0.115454</td>\n",
       "      <td>0.065487</td>\n",
       "      <td>0.056536</td>\n",
       "      <td>1.882789</td>\n",
       "      <td>-0.603070</td>\n",
       "      <td>0.150646</td>\n",
       "      <td>-0.726228</td>\n",
       "      <td>0.140199</td>\n",
       "      <td>-0.025163</td>\n",
       "      <td>0.475472</td>\n",
       "      <td>-0.842357</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>77</td>\n",
       "      <td>-0.134750</td>\n",
       "      <td>0.065997</td>\n",
       "      <td>0.056922</td>\n",
       "      <td>1.878222</td>\n",
       "      <td>-0.603782</td>\n",
       "      <td>0.148198</td>\n",
       "      <td>-0.728316</td>\n",
       "      <td>0.141955</td>\n",
       "      <td>-0.018916</td>\n",
       "      <td>0.475619</td>\n",
       "      <td>-0.837601</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>78</td>\n",
       "      <td>-0.152247</td>\n",
       "      <td>0.048527</td>\n",
       "      <td>0.039753</td>\n",
       "      <td>1.850924</td>\n",
       "      <td>-0.621950</td>\n",
       "      <td>0.139474</td>\n",
       "      <td>-0.759329</td>\n",
       "      <td>0.128361</td>\n",
       "      <td>-0.034021</td>\n",
       "      <td>0.469027</td>\n",
       "      <td>-0.854461</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>79</td>\n",
       "      <td>-0.171768</td>\n",
       "      <td>0.043979</td>\n",
       "      <td>0.035069</td>\n",
       "      <td>1.836700</td>\n",
       "      <td>-0.628541</td>\n",
       "      <td>0.129589</td>\n",
       "      <td>-0.750970</td>\n",
       "      <td>0.128198</td>\n",
       "      <td>-0.039440</td>\n",
       "      <td>0.467033</td>\n",
       "      <td>-0.862017</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>80</td>\n",
       "      <td>-0.186480</td>\n",
       "      <td>0.038166</td>\n",
       "      <td>0.029049</td>\n",
       "      <td>1.830149</td>\n",
       "      <td>-0.621257</td>\n",
       "      <td>0.116480</td>\n",
       "      <td>-0.769787</td>\n",
       "      <td>0.120776</td>\n",
       "      <td>-0.045041</td>\n",
       "      <td>0.464130</td>\n",
       "      <td>-0.863055</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>81</td>\n",
       "      <td>-0.202474</td>\n",
       "      <td>0.031894</td>\n",
       "      <td>0.022716</td>\n",
       "      <td>1.821135</td>\n",
       "      <td>-0.638060</td>\n",
       "      <td>0.111184</td>\n",
       "      <td>-0.776156</td>\n",
       "      <td>0.119827</td>\n",
       "      <td>-0.050493</td>\n",
       "      <td>0.463925</td>\n",
       "      <td>-0.869637</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>82</td>\n",
       "      <td>-0.219985</td>\n",
       "      <td>0.025162</td>\n",
       "      <td>0.016591</td>\n",
       "      <td>1.806260</td>\n",
       "      <td>-0.640977</td>\n",
       "      <td>0.103495</td>\n",
       "      <td>-0.786721</td>\n",
       "      <td>0.116717</td>\n",
       "      <td>-0.050746</td>\n",
       "      <td>0.460647</td>\n",
       "      <td>-0.875948</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>83</td>\n",
       "      <td>-0.237313</td>\n",
       "      <td>0.020592</td>\n",
       "      <td>0.011626</td>\n",
       "      <td>1.801409</td>\n",
       "      <td>-0.643287</td>\n",
       "      <td>0.090415</td>\n",
       "      <td>-0.792121</td>\n",
       "      <td>0.114190</td>\n",
       "      <td>-0.058001</td>\n",
       "      <td>0.459079</td>\n",
       "      <td>-0.878678</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>84</td>\n",
       "      <td>-0.250109</td>\n",
       "      <td>0.013043</td>\n",
       "      <td>0.004042</td>\n",
       "      <td>1.795240</td>\n",
       "      <td>-0.656521</td>\n",
       "      <td>0.078188</td>\n",
       "      <td>-0.803820</td>\n",
       "      <td>0.112858</td>\n",
       "      <td>-0.064970</td>\n",
       "      <td>0.456417</td>\n",
       "      <td>-0.885054</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>85</td>\n",
       "      <td>-0.263943</td>\n",
       "      <td>0.007856</td>\n",
       "      <td>-0.000985</td>\n",
       "      <td>1.785501</td>\n",
       "      <td>-0.659377</td>\n",
       "      <td>0.072531</td>\n",
       "      <td>-0.812175</td>\n",
       "      <td>0.107008</td>\n",
       "      <td>-0.069746</td>\n",
       "      <td>0.455314</td>\n",
       "      <td>-0.886934</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>86</td>\n",
       "      <td>-0.280294</td>\n",
       "      <td>0.003565</td>\n",
       "      <td>-0.005538</td>\n",
       "      <td>1.778569</td>\n",
       "      <td>-0.666055</td>\n",
       "      <td>0.064058</td>\n",
       "      <td>-0.816177</td>\n",
       "      <td>0.107192</td>\n",
       "      <td>-0.072119</td>\n",
       "      <td>0.453961</td>\n",
       "      <td>-0.893732</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>87</td>\n",
       "      <td>-0.288129</td>\n",
       "      <td>0.002483</td>\n",
       "      <td>-0.006649</td>\n",
       "      <td>1.778045</td>\n",
       "      <td>-0.666879</td>\n",
       "      <td>0.057104</td>\n",
       "      <td>-0.815343</td>\n",
       "      <td>0.106715</td>\n",
       "      <td>-0.070123</td>\n",
       "      <td>0.452821</td>\n",
       "      <td>-0.895534</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>88</td>\n",
       "      <td>-0.303255</td>\n",
       "      <td>-0.000776</td>\n",
       "      <td>-0.010089</td>\n",
       "      <td>1.772699</td>\n",
       "      <td>-0.672119</td>\n",
       "      <td>0.061925</td>\n",
       "      <td>-0.825670</td>\n",
       "      <td>0.103414</td>\n",
       "      <td>-0.073457</td>\n",
       "      <td>0.452931</td>\n",
       "      <td>-0.900434</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>89</td>\n",
       "      <td>-0.312906</td>\n",
       "      <td>-0.003248</td>\n",
       "      <td>-0.012425</td>\n",
       "      <td>1.769217</td>\n",
       "      <td>-0.671893</td>\n",
       "      <td>0.054544</td>\n",
       "      <td>-0.831305</td>\n",
       "      <td>0.105047</td>\n",
       "      <td>-0.076261</td>\n",
       "      <td>0.452558</td>\n",
       "      <td>-0.901304</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>90</td>\n",
       "      <td>-0.325885</td>\n",
       "      <td>-0.005243</td>\n",
       "      <td>-0.014808</td>\n",
       "      <td>1.766008</td>\n",
       "      <td>-0.675104</td>\n",
       "      <td>0.049917</td>\n",
       "      <td>-0.832984</td>\n",
       "      <td>0.103304</td>\n",
       "      <td>-0.078345</td>\n",
       "      <td>0.451546</td>\n",
       "      <td>-0.902804</td>\n",
       "      <td>00:35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>91</td>\n",
       "      <td>-0.335751</td>\n",
       "      <td>-0.007712</td>\n",
       "      <td>-0.017172</td>\n",
       "      <td>1.765584</td>\n",
       "      <td>-0.676647</td>\n",
       "      <td>0.044709</td>\n",
       "      <td>-0.836528</td>\n",
       "      <td>0.100550</td>\n",
       "      <td>-0.079750</td>\n",
       "      <td>0.450184</td>\n",
       "      <td>-0.905476</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>92</td>\n",
       "      <td>-0.343069</td>\n",
       "      <td>-0.010378</td>\n",
       "      <td>-0.019740</td>\n",
       "      <td>1.761020</td>\n",
       "      <td>-0.681546</td>\n",
       "      <td>0.043413</td>\n",
       "      <td>-0.837005</td>\n",
       "      <td>0.098233</td>\n",
       "      <td>-0.083635</td>\n",
       "      <td>0.449718</td>\n",
       "      <td>-0.908120</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>93</td>\n",
       "      <td>-0.351836</td>\n",
       "      <td>-0.010092</td>\n",
       "      <td>-0.019520</td>\n",
       "      <td>1.762408</td>\n",
       "      <td>-0.679949</td>\n",
       "      <td>0.038372</td>\n",
       "      <td>-0.837684</td>\n",
       "      <td>0.100658</td>\n",
       "      <td>-0.081365</td>\n",
       "      <td>0.449550</td>\n",
       "      <td>-0.908149</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>94</td>\n",
       "      <td>-0.356374</td>\n",
       "      <td>-0.011895</td>\n",
       "      <td>-0.021378</td>\n",
       "      <td>1.759331</td>\n",
       "      <td>-0.681498</td>\n",
       "      <td>0.035656</td>\n",
       "      <td>-0.840271</td>\n",
       "      <td>0.099077</td>\n",
       "      <td>-0.083380</td>\n",
       "      <td>0.449616</td>\n",
       "      <td>-0.909555</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>95</td>\n",
       "      <td>-0.362251</td>\n",
       "      <td>-0.013264</td>\n",
       "      <td>-0.022725</td>\n",
       "      <td>1.757666</td>\n",
       "      <td>-0.683705</td>\n",
       "      <td>0.034682</td>\n",
       "      <td>-0.842917</td>\n",
       "      <td>0.098075</td>\n",
       "      <td>-0.084634</td>\n",
       "      <td>0.449207</td>\n",
       "      <td>-0.910177</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>96</td>\n",
       "      <td>-0.367783</td>\n",
       "      <td>-0.013222</td>\n",
       "      <td>-0.022709</td>\n",
       "      <td>1.757321</td>\n",
       "      <td>-0.683970</td>\n",
       "      <td>0.035219</td>\n",
       "      <td>-0.842692</td>\n",
       "      <td>0.098445</td>\n",
       "      <td>-0.084522</td>\n",
       "      <td>0.449064</td>\n",
       "      <td>-0.910536</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>97</td>\n",
       "      <td>-0.368744</td>\n",
       "      <td>-0.013842</td>\n",
       "      <td>-0.023317</td>\n",
       "      <td>1.757113</td>\n",
       "      <td>-0.684457</td>\n",
       "      <td>0.033837</td>\n",
       "      <td>-0.843935</td>\n",
       "      <td>0.097806</td>\n",
       "      <td>-0.084790</td>\n",
       "      <td>0.448868</td>\n",
       "      <td>-0.910982</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>98</td>\n",
       "      <td>-0.367732</td>\n",
       "      <td>-0.013866</td>\n",
       "      <td>-0.023357</td>\n",
       "      <td>1.757200</td>\n",
       "      <td>-0.684259</td>\n",
       "      <td>0.033624</td>\n",
       "      <td>-0.844176</td>\n",
       "      <td>0.097858</td>\n",
       "      <td>-0.084822</td>\n",
       "      <td>0.448898</td>\n",
       "      <td>-0.911178</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>99</td>\n",
       "      <td>-0.370888</td>\n",
       "      <td>-0.013845</td>\n",
       "      <td>-0.023328</td>\n",
       "      <td>1.757342</td>\n",
       "      <td>-0.684152</td>\n",
       "      <td>0.033363</td>\n",
       "      <td>-0.844163</td>\n",
       "      <td>0.097954</td>\n",
       "      <td>-0.084692</td>\n",
       "      <td>0.448900</td>\n",
       "      <td>-0.911175</td>\n",
       "      <td>00:36</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Better model found at epoch 0 with 👉🏻LMAE👈🏻 value: 1.9120843410491943.\n",
      "Better model found at epoch 1 with 👉🏻LMAE👈🏻 value: 1.8619025945663452.\n",
      "Better model found at epoch 2 with 👉🏻LMAE👈🏻 value: 1.807070255279541.\n",
      "Better model found at epoch 3 with 👉🏻LMAE👈🏻 value: 1.759743571281433.\n",
      "Better model found at epoch 4 with 👉🏻LMAE👈🏻 value: 1.7186700105667114.\n",
      "Better model found at epoch 5 with 👉🏻LMAE👈🏻 value: 1.6862719058990479.\n",
      "Better model found at epoch 6 with 👉🏻LMAE👈🏻 value: 1.6590485572814941.\n",
      "Better model found at epoch 7 with 👉🏻LMAE👈🏻 value: 1.643379807472229.\n",
      "Better model found at epoch 8 with 👉🏻LMAE👈🏻 value: 1.6195101737976074.\n",
      "Better model found at epoch 9 with 👉🏻LMAE👈🏻 value: 1.5883119106292725.\n",
      "Better model found at epoch 10 with 👉🏻LMAE👈🏻 value: 1.5807496309280396.\n",
      "Better model found at epoch 11 with 👉🏻LMAE👈🏻 value: 1.54558527469635.\n",
      "Better model found at epoch 12 with 👉🏻LMAE👈🏻 value: 1.525373935699463.\n",
      "Better model found at epoch 13 with 👉🏻LMAE👈🏻 value: 1.4936565160751343.\n",
      "Better model found at epoch 14 with 👉🏻LMAE👈🏻 value: 1.4701259136199951.\n",
      "Better model found at epoch 15 with 👉🏻LMAE👈🏻 value: 1.4514720439910889.\n",
      "Better model found at epoch 16 with 👉🏻LMAE👈🏻 value: 1.432856559753418.\n",
      "Better model found at epoch 17 with 👉🏻LMAE👈🏻 value: 1.400344967842102.\n",
      "Better model found at epoch 18 with 👉🏻LMAE👈🏻 value: 1.3792815208435059.\n",
      "Better model found at epoch 19 with 👉🏻LMAE👈🏻 value: 1.3619393110275269.\n",
      "Better model found at epoch 20 with 👉🏻LMAE👈🏻 value: 1.3380300998687744.\n",
      "Better model found at epoch 21 with 👉🏻LMAE👈🏻 value: 1.2932649850845337.\n",
      "Better model found at epoch 22 with 👉🏻LMAE👈🏻 value: 1.2692570686340332.\n",
      "Better model found at epoch 23 with 👉🏻LMAE👈🏻 value: 1.21198308467865.\n",
      "Better model found at epoch 24 with 👉🏻LMAE👈🏻 value: 1.1715271472930908.\n",
      "Better model found at epoch 25 with 👉🏻LMAE👈🏻 value: 1.118619441986084.\n",
      "Better model found at epoch 26 with 👉🏻LMAE👈🏻 value: 1.0714688301086426.\n",
      "Better model found at epoch 27 with 👉🏻LMAE👈🏻 value: 1.0351771116256714.\n",
      "Better model found at epoch 28 with 👉🏻LMAE👈🏻 value: 1.001774787902832.\n",
      "Better model found at epoch 29 with 👉🏻LMAE👈🏻 value: 0.9479114413261414.\n",
      "Better model found at epoch 30 with 👉🏻LMAE👈🏻 value: 0.9230896830558777.\n",
      "Better model found at epoch 31 with 👉🏻LMAE👈🏻 value: 0.8863322734832764.\n",
      "Better model found at epoch 32 with 👉🏻LMAE👈🏻 value: 0.8474680185317993.\n",
      "Better model found at epoch 33 with 👉🏻LMAE👈🏻 value: 0.8339117169380188.\n",
      "Better model found at epoch 34 with 👉🏻LMAE👈🏻 value: 0.8129868507385254.\n",
      "Better model found at epoch 35 with 👉🏻LMAE👈🏻 value: 0.7886001467704773.\n",
      "Better model found at epoch 36 with 👉🏻LMAE👈🏻 value: 0.7770330309867859.\n",
      "Better model found at epoch 37 with 👉🏻LMAE👈🏻 value: 0.7765684127807617.\n",
      "Better model found at epoch 38 with 👉🏻LMAE👈🏻 value: 0.7354208827018738.\n",
      "Better model found at epoch 39 with 👉🏻LMAE👈🏻 value: 0.7253214120864868.\n",
      "Better model found at epoch 40 with 👉🏻LMAE👈🏻 value: 0.699440062046051.\n",
      "Better model found at epoch 41 with 👉🏻LMAE👈🏻 value: 0.6843871474266052.\n",
      "Better model found at epoch 42 with 👉🏻LMAE👈🏻 value: 0.6627888679504395.\n",
      "Better model found at epoch 43 with 👉🏻LMAE👈🏻 value: 0.629173755645752.\n",
      "Better model found at epoch 45 with 👉🏻LMAE👈🏻 value: 0.6016845107078552.\n",
      "Better model found at epoch 46 with 👉🏻LMAE👈🏻 value: 0.5928371548652649.\n",
      "Better model found at epoch 47 with 👉🏻LMAE👈🏻 value: 0.5773522853851318.\n",
      "Better model found at epoch 48 with 👉🏻LMAE👈🏻 value: 0.5749238729476929.\n",
      "Better model found at epoch 49 with 👉🏻LMAE👈🏻 value: 0.5371617078781128.\n",
      "Better model found at epoch 50 with 👉🏻LMAE👈🏻 value: 0.5138949155807495.\n",
      "Better model found at epoch 51 with 👉🏻LMAE👈🏻 value: 0.506271481513977.\n",
      "Better model found at epoch 52 with 👉🏻LMAE👈🏻 value: 0.4689498543739319.\n",
      "Better model found at epoch 53 with 👉🏻LMAE👈🏻 value: 0.44707292318344116.\n",
      "Better model found at epoch 54 with 👉🏻LMAE👈🏻 value: 0.41851258277893066.\n",
      "Better model found at epoch 55 with 👉🏻LMAE👈🏻 value: 0.40434688329696655.\n",
      "Better model found at epoch 56 with 👉🏻LMAE👈🏻 value: 0.3627530634403229.\n",
      "Better model found at epoch 57 with 👉🏻LMAE👈🏻 value: 0.34324926137924194.\n",
      "Better model found at epoch 58 with 👉🏻LMAE👈🏻 value: 0.3256276249885559.\n",
      "Better model found at epoch 59 with 👉🏻LMAE👈🏻 value: 0.3110637962818146.\n",
      "Better model found at epoch 60 with 👉🏻LMAE👈🏻 value: 0.28602898120880127.\n",
      "Better model found at epoch 61 with 👉🏻LMAE👈🏻 value: 0.2629399299621582.\n",
      "Better model found at epoch 62 with 👉🏻LMAE👈🏻 value: 0.25940844416618347.\n",
      "Better model found at epoch 63 with 👉🏻LMAE👈🏻 value: 0.24413636326789856.\n",
      "Better model found at epoch 64 with 👉🏻LMAE👈🏻 value: 0.21263840794563293.\n",
      "Better model found at epoch 66 with 👉🏻LMAE👈🏻 value: 0.18761210143566132.\n",
      "Better model found at epoch 67 with 👉🏻LMAE👈🏻 value: 0.17298081517219543.\n",
      "Better model found at epoch 68 with 👉🏻LMAE👈🏻 value: 0.1567360758781433.\n",
      "Better model found at epoch 69 with 👉🏻LMAE👈🏻 value: 0.14279183745384216.\n",
      "Better model found at epoch 70 with 👉🏻LMAE👈🏻 value: 0.1326354742050171.\n",
      "Better model found at epoch 71 with 👉🏻LMAE👈🏻 value: 0.11445330828428268.\n",
      "Better model found at epoch 72 with 👉🏻LMAE👈🏻 value: 0.11100741475820541.\n",
      "Better model found at epoch 73 with 👉🏻LMAE👈🏻 value: 0.0911564826965332.\n",
      "Better model found at epoch 74 with 👉🏻LMAE👈🏻 value: 0.07867909967899323.\n",
      "Better model found at epoch 75 with 👉🏻LMAE👈🏻 value: 0.0666799247264862.\n",
      "Better model found at epoch 76 with 👉🏻LMAE👈🏻 value: 0.05653608590364456.\n",
      "Better model found at epoch 78 with 👉🏻LMAE👈🏻 value: 0.03975300490856171.\n",
      "Better model found at epoch 79 with 👉🏻LMAE👈🏻 value: 0.03506893664598465.\n",
      "Better model found at epoch 80 with 👉🏻LMAE👈🏻 value: 0.029049329459667206.\n",
      "Better model found at epoch 81 with 👉🏻LMAE👈🏻 value: 0.022715650498867035.\n",
      "Better model found at epoch 82 with 👉🏻LMAE👈🏻 value: 0.01659085601568222.\n",
      "Better model found at epoch 83 with 👉🏻LMAE👈🏻 value: 0.01162576675415039.\n",
      "Better model found at epoch 84 with 👉🏻LMAE👈🏻 value: 0.004042387008666992.\n",
      "Better model found at epoch 85 with 👉🏻LMAE👈🏻 value: -0.0009848251938819885.\n",
      "Better model found at epoch 86 with 👉🏻LMAE👈🏻 value: -0.005537807941436768.\n",
      "Better model found at epoch 87 with 👉🏻LMAE👈🏻 value: -0.0066489651799201965.\n",
      "Better model found at epoch 88 with 👉🏻LMAE👈🏻 value: -0.010088667273521423.\n",
      "Better model found at epoch 89 with 👉🏻LMAE👈🏻 value: -0.012424618005752563.\n",
      "Better model found at epoch 90 with 👉🏻LMAE👈🏻 value: -0.014807827770709991.\n",
      "Better model found at epoch 91 with 👉🏻LMAE👈🏻 value: -0.01717177778482437.\n",
      "Better model found at epoch 92 with 👉🏻LMAE👈🏻 value: -0.01974017173051834.\n",
      "Better model found at epoch 94 with 👉🏻LMAE👈🏻 value: -0.02137802541255951.\n",
      "Better model found at epoch 95 with 👉🏻LMAE👈🏻 value: -0.022725358605384827.\n",
      "Better model found at epoch 97 with 👉🏻LMAE👈🏻 value: -0.023317456245422363.\n",
      "Better model found at epoch 98 with 👉🏻LMAE👈🏻 value: -0.023357026278972626.\n"
     ]
    }
   ],
   "source": [
    "learner.fit_one_cycle(100,1.5e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learner.recorder.plot_losses()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/antor/miniconda3/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:211: RuntimeWarning: Glyph 128073 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/home/antor/miniconda3/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:211: RuntimeWarning: Glyph 127995 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/home/antor/miniconda3/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:211: RuntimeWarning: Glyph 128072 missing from current font.\n",
      "  font.set_text(s, 0.0, flags=flags)\n",
      "/home/antor/miniconda3/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:180: RuntimeWarning: Glyph 128073 missing from current font.\n",
      "  font.set_text(s, 0, flags=flags)\n",
      "/home/antor/miniconda3/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:180: RuntimeWarning: Glyph 127995 missing from current font.\n",
      "  font.set_text(s, 0, flags=flags)\n",
      "/home/antor/miniconda3/lib/python3.7/site-packages/matplotlib/backends/backend_agg.py:180: RuntimeWarning: Glyph 128072 missing from current font.\n",
      "  font.set_text(s, 0, flags=flags)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x2592 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learner.recorder.plot_metrics()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Optional Fine tune regular fit "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.callbacks.append(\n",
    "    ReduceLROnPlateauCallback(learner, monitor='train_loss', mode='min', factor=0.2, patience=3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.fit(100, 5e-5)# , moms=(0.75,0.70))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.fit_one_cycle(300, 5e-4)#, moms=(0.75,0.70))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.recorder.plot_lr(show_moms=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.recorder.plot_losses()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.recorder.plot_metrics()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.fit(10,1e-4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Validate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.to_fp32()\n",
    "val = learner.validate()[1]\n",
    "print(val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "try:\n",
    "    sub_fname = f'loss{learner.recorder.losses[-1]:.04f}val{val:.04f}'\n",
    "except:\n",
    "    sub_fname = f'val{val:.04f}'\n",
    "learner.save(sub_fname)\n",
    "print(sub_fname)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Inference\n",
    "\n",
    "Make sure `tranforms` are activated to test set otherwise TTA > 1 will be as TTA =1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "try:\n",
    "    xt_coulombmat = load_fn(f'xt_coulombmat32{ext}.npy')\n",
    "except:\n",
    "    xt_coulombmat = np.load(f'xt_coulombmat{ext}.npy', allow_pickle=True)\n",
    "    xt_coulombmat = np.array(xt_coulombmat.tolist()).astype(np.float32)\n",
    "    np.save(f'xt_coulombmat32{ext}.npy', xt_coulombmat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_fname = Path('test.npz')\n",
    "try:\n",
    "    npzfile  = np.load(fname_ext(test_fname, ext))\n",
    "    xt_xyz   = npzfile['x_xyz']\n",
    "    xt_type  = npzfile['x_type']\n",
    "    xt_ext   = npzfile['x_ext']\n",
    "    xt_atom  = npzfile['x_atom']\n",
    "    mt = npzfile['m']\n",
    "    xt_ids = npzfile['x_ids']\n",
    "except:\n",
    "    xt_xyz,xt_type,xt_ext,xt_atom,mt,xt_ids = \\\n",
    "        preprocess(test_fname.with_suffix('.csv'), type_index=types,ext=ext)\n",
    "    np.savez(fname_ext('_'+test_fname, ext), \n",
    "             x_xyz  = xt_xyz,\n",
    "             x_type = xt_type,\n",
    "             x_ext  = xt_ext,\n",
    "             x_atom = xt_atom,\n",
    "             m=mt,\n",
    "             x_ids=xt_ids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "xt_qm9_mulliken = load_fn(f'xt_qm9_mulliken{ext}.npy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "[v.shape for v in [xt_xyz,xt_type,xt_ext,xt_atom, xt_qm9_mulliken,xt_ids, mt]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "learner.data.add_test(ItemList(items=(MoleculeItem(i,*v) for i,v in \n",
    "                              enumerate(zip(xt_xyz,xt_type,xt_ext,xt_atom,xt_qm9_mulliken,xt_coulombmat)))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "TTA_N = 1\n",
    "learner.data.test_ds.tfms = tta_tfms if TTA_N > 1 else tfms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#learner.model = learner.model.module\n",
    "#data.batch_size = 4096*2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#learner.model = nn.DataParallel(learner.model)\n",
    "old_bs = data.batch_size\n",
    "data.batch_size *= 2\n",
    "\n",
    "sub = defaultdict(int)\n",
    "xt_ids_not_extended = (xt_ids!=0) & (xt_ids<=7163688) # TODO\n",
    "ids = xt_ids[xt_ids_not_extended]\n",
    "\n",
    "mb = master_bar(range(TTA_N))\n",
    "for tta in mb:\n",
    "    test_preds = np.zeros((0, 29), dtype=np.float32)\n",
    "\n",
    "    for batch_idx, batch in progress_bar(\n",
    "        enumerate(learner.dl(DatasetType.Test)), total=len(learner.dl(DatasetType.Test)), parent=mb):\n",
    "        _, _, preds_,_,_,_ = learner.pred_batch(ds_type=DatasetType.Test, batch=batch)\n",
    "        preds_ = preds_.sum(dim=1)\n",
    "        test_preds = np.concatenate([test_preds, preds_.data.cpu().numpy()], axis = 0)\n",
    "\n",
    "    preds = test_preds[xt_ids_not_extended]\n",
    "    for k in range(len(ids)):\n",
    "        sub[int(ids[k])] += preds[k]\n",
    "    \n",
    "for k in range(len(ids)):\n",
    "    sub[int(ids[k])] = sub[int(ids[k])]/TTA_N\n",
    "\n",
    "learner.model = learner.model.module\n",
    "data.batch_size = old_bs\n",
    "\n",
    "sub_df = pd.DataFrame(sub.items(), columns=['id', 'scalar_coupling_constant'])\n",
    "sub_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Submit to Kaggle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "csv_fname = f'{sub_fname}_tta{TTA_N}.csv'\n",
    "sub_df.to_csv(csv_fname, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "comp = 'champs-scalar-coupling'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "!kaggle competitions submit -c {comp} -f {csv_fname} -m 'QM9 tta {TTA_N} {ext}'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import time\n",
    "time.sleep(60)\n",
    "!kaggle competitions submissions -c {comp} -v > submissions-{comp}.csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "submissions = pd.read_csv(f'submissions-{comp}.csv')\n",
    "submissions.iloc[0].publicScore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "    "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (root)",
   "language": "python",
   "name": "root"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}