notebooks/3.Sage_integration_example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Example: Random KS model (we just take any with Picard Rank 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# some general packages\n",
    "import numpy as np\n",
    "import os\n",
    "import pickle\n",
    "\n",
    "# ML packages\n",
    "import tensorflow as tf\n",
    "\n",
    "# data generation\n",
    "from cymetric.pointgen.pointgen_mathematica import ToricPointGeneratorMathematica\n",
    "from cymetric.pointgen.nphelper import prepare_dataset, prepare_basis_pickle\n",
    "from cymetric.models.tfhelper import prepare_tf_basis, train_model\n",
    "\n",
    "# training\n",
    "from cymetric.models.tfmodels import PhiFSModelToric\n",
    "from cymetric.models.callbacks import SigmaCallback, TransitionCallback, RicciCallback, VolkCallback\n",
    "from cymetric.models.metrics import SigmaLoss, TransitionLoss, RicciLoss, VolkLoss\n",
    "from cymetric.models.measures import sigma_measure, transition_measure_loss, ricci_measure"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compute toric information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['dim_cy',\n",
       " 'vol_j_norm',\n",
       " 'coeff_aK',\n",
       " 'exp_aK',\n",
       " 'exps_sections',\n",
       " 'patch_masks',\n",
       " 'glsm_charges',\n",
       " 'non_ci_coeffs',\n",
       " 'non_ci_exps',\n",
       " 'int_nums']"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from cymetric.sage.sagelib import prepare_toric_cy_data\n",
    "\n",
    "work_dir = \"./ToricModel\"\n",
    "vertices = [[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], [0, 0, 0, 1], [-1, -1, -1, 0], [2, 0, 0, -1]]  # P2 fibered over P2\n",
    "polytope = LatticePolytope(vertices)\n",
    "pConfig = PointConfiguration(polytope.points(), star=[0 for _ in range(len(vertices[0]))])\n",
    "triangs = pConfig.restrict_to_connected_triangulations().restrict_to_fine_triangulations().restrict_to_regular_triangulations().restrict_to_star_triangulations([0 for _ in range(len(vertices[0]))]).triangulations_list()\n",
    "triang = triangs[0]\n",
    "tv_fan = triang.fan()\n",
    "tv = ToricVariety(tv_fan)\n",
    "\n",
    "toric_data = prepare_toric_cy_data(tv, os.path.join(work_dir, \"toric_data.pickle\"))\n",
    "list(toric_data.keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generate points"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "work_dir = \"./ToricModel\"\n",
    "with open(os.path.join(work_dir, 'toric_data.pickle'), 'rb') as f:\n",
    "    toric_data = pickle.load(f)\n",
    "num_pts       = int(50000)\n",
    "precision     = int(10)\n",
    "verbose       = int(1)\n",
    "kmoduli       = np.ones(len(toric_data['exps_sections']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "mathematica_pointgen = ToricPointGeneratorMathematica(toric_data, kmoduli, precision=precision, verbose=verbose)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Connected to logging socket: tcp://127.0.0.1:51037\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:DEBUG:Evaluating a new expression.\n",
      "pointgenMathematica:DEBUG:Running with 10 Mathematica kernels.\n",
      "pointgenMathematica:DEBUG:Initializing generation of 50000 points...\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Number of points on CY from one ambient space intersection: 3\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Now generating 50000 points...\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 0% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 5% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 10% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 15% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 20% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 25% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 30% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 35% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 40% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 45% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 50% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 55% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 60% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 65% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 70% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 75% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 80% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 85% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 90% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:Generated 95% of points\n",
      "WolframKernel-<tcp://127.0.0.1:51037>:INFO:done.\n",
      "/Users/ruehle/venv-ml/lib/python3.8/site-packages/numpy/linalg/linalg.py:2146: RuntimeWarning: divide by zero encountered in det\n",
      "  r = _umath_linalg.det(a, signature=signature)\n",
      "/Users/ruehle/venv-ml/lib/python3.8/site-packages/numpy/linalg/linalg.py:2146: RuntimeWarning: invalid value encountered in det\n",
      "  r = _umath_linalg.det(a, signature=signature)\n"
     ]
    }
   ],
   "source": [
    "kappa = mathematica_pointgen.prepare_dataset(num_pts, work_dir)\n",
    "prepare_basis_pickle(mathematica_pointgen, work_dir, kappa);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## set up NN"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "load in data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf.get_logger().setLevel('ERROR')\n",
    "data = np.load(os.path.join(work_dir, 'dataset.npz'))\n",
    "toric_data = np.load(os.path.join(work_dir, 'toric_data.pickle'), allow_pickle=True)\n",
    "BASIS = prepare_tf_basis(np.load(os.path.join(work_dir, 'basis.pickle'), allow_pickle=True))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "set up NN callbacks (info on training process)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "scb = SigmaCallback((data['X_val'], data['y_val']))\n",
    "rcb = RicciCallback((data['X_val'], data['y_val']), data['val_pullbacks'])\n",
    "volkck = VolkCallback((data['X_val'], data['y_val']))\n",
    "cb_list = [scb, volkck, rcb]\n",
    "\n",
    "cmetrics = [SigmaLoss(), VolkLoss()]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "define the NN hyperparameters (wrapping everything in int() and float() is necessary in sage)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "nfold       = int(BASIS['NFOLD'].numpy().real)\n",
    "amb         = [int(6)]\n",
    "n_in        = int(12)\n",
    "n_out       = int(1)\n",
    "n_hiddens   = [int(128), int(128), int(128)]\n",
    "acts        = ['gelu', 'gelu', 'gelu']\n",
    "n_epochs    = int(30)\n",
    "batch_sizes = [int(128), int(50000)]\n",
    "alpha       = [float(1.), float(1.), float(1.), float(1.), float(1.)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "initialize the NN and the Toric Phi Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.Sequential()\n",
    "model.add(tf.keras.Input(shape=(int(n_in))))\n",
    "for n_hidden, act in zip(n_hiddens, acts):\n",
    "    model.add(tf.keras.layers.Dense(n_hidden, activation=act))\n",
    "model.add(tf.keras.layers.Dense(n_out, use_bias=False))\n",
    "\n",
    "fs_model = PhiFSModelToric(model, BASIS, alpha=alpha, toric_data=toric_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## train NN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch  1/30\n",
      " - Sigma measure val:      0.4424\n",
      " - Volk val:               22.3904\n",
      " - Ricci measure val:      7.3883\n",
      "352/352 [==============================] - ETA: 0s - sigma_loss: 0.3633 - volk_loss: 0.0000e+00 - Sigma measure val:      0.3901\n",
      " - Volk val:               14.7942\n",
      " - Ricci measure val:      7.2322\n",
      "352/352 [==============================] - 31s 68ms/step - sigma_loss: 0.3633 - volk_loss: 0.0000e+00 - sigma_val: 0.3901 - volk_val: 14.7942 - ricci_val: 7.2322\n",
      " - Sigma measure val:      0.3901\n",
      " - Volk val:               14.7942\n",
      " - Ricci measure val:      7.2322\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.3237 - volk_loss: 6.4312 - Sigma measure val:      0.3927\n",
      " - Volk val:               14.2859\n",
      " - Ricci measure val:      5.6926\n",
      "1/1 [==============================] - 22s 22s/step - sigma_loss: 0.3237 - volk_loss: 6.4312 - sigma_val: 0.3927 - volk_val: 14.2859 - ricci_val: 5.6926\n",
      "\n",
      "Epoch  2/30\n",
      " - Sigma measure val:      0.3927\n",
      " - Volk val:               14.2859\n",
      " - Ricci measure val:      5.6926\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.3126 - volk_loss: 0.0000e+00 - Sigma measure val:      0.3694\n",
      " - Volk val:               15.2084\n",
      " - Ricci measure val:      7.9090\n",
      "352/352 [==============================] - 19s 49ms/step - sigma_loss: 0.3125 - volk_loss: 0.0000e+00 - sigma_val: 0.3694 - volk_val: 15.2084 - ricci_val: 7.9090\n",
      " - Sigma measure val:      0.3694\n",
      " - Volk val:               15.2084\n",
      " - Ricci measure val:      7.9090\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2994 - volk_loss: 6.6597 - Sigma measure val:      0.3702\n",
      " - Volk val:               15.1914\n",
      " - Ricci measure val:      7.3781\n",
      "1/1 [==============================] - 22s 22s/step - sigma_loss: 0.2994 - volk_loss: 6.6597 - sigma_val: 0.3702 - volk_val: 15.1914 - ricci_val: 7.3781\n",
      "\n",
      "Epoch  3/30\n",
      " - Sigma measure val:      0.3702\n",
      " - Volk val:               15.1914\n",
      " - Ricci measure val:      7.3781\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.2946 - volk_loss: 0.0000e+00 - Sigma measure val:      0.3560\n",
      " - Volk val:               15.1078\n",
      " - Ricci measure val:      6.2463\n",
      "352/352 [==============================] - 20s 50ms/step - sigma_loss: 0.2946 - volk_loss: 0.0000e+00 - sigma_val: 0.3560 - volk_val: 15.1078 - ricci_val: 6.2463\n",
      " - Sigma measure val:      0.3560\n",
      " - Volk val:               15.1078\n",
      " - Ricci measure val:      6.2463\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2865 - volk_loss: 6.5646 - Sigma measure val:      0.3575\n",
      " - Volk val:               14.6676\n",
      " - Ricci measure val:      6.2145\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.2865 - volk_loss: 6.5646 - sigma_val: 0.3575 - volk_val: 14.6676 - ricci_val: 6.2145\n",
      "\n",
      "Epoch  4/30\n",
      " - Sigma measure val:      0.3575\n",
      " - Volk val:               14.6676\n",
      " - Ricci measure val:      6.2145\n",
      "352/352 [==============================] - ETA: 0s - sigma_loss: 0.2790 - volk_loss: 0.0000e+00 - Sigma measure val:      0.3367\n",
      " - Volk val:               15.6396\n",
      " - Ricci measure val:      99.9598\n",
      "352/352 [==============================] - 24s 62ms/step - sigma_loss: 0.2790 - volk_loss: 0.0000e+00 - sigma_val: 0.3367 - volk_val: 15.6396 - ricci_val: 99.9598\n",
      " - Sigma measure val:      0.3367\n",
      " - Volk val:               15.6396\n",
      " - Ricci measure val:      99.9598\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2691 - volk_loss: 6.9708 - Sigma measure val:      0.3377\n",
      " - Volk val:               15.4413\n",
      " - Ricci measure val:      2.8698\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.2691 - volk_loss: 6.9708 - sigma_val: 0.3377 - volk_val: 15.4413 - ricci_val: 2.8698\n",
      "\n",
      "Epoch  5/30\n",
      " - Sigma measure val:      0.3377\n",
      " - Volk val:               15.4413\n",
      " - Ricci measure val:      2.8698\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.2648 - volk_loss: 0.0000e+00 - Sigma measure val:      0.3180\n",
      " - Volk val:               15.3294\n",
      " - Ricci measure val:      8.1073\n",
      "352/352 [==============================] - 18s 45ms/step - sigma_loss: 0.2649 - volk_loss: 0.0000e+00 - sigma_val: 0.3180 - volk_val: 15.3294 - ricci_val: 8.1073\n",
      " - Sigma measure val:      0.3180\n",
      " - Volk val:               15.3294\n",
      " - Ricci measure val:      8.1073\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2563 - volk_loss: 6.7116 - Sigma measure val:      0.3188\n",
      " - Volk val:               14.7498\n",
      " - Ricci measure val:      14.4950\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.2563 - volk_loss: 6.7116 - sigma_val: 0.3188 - volk_val: 14.7498 - ricci_val: 14.4950\n",
      "\n",
      "Epoch  6/30\n",
      " - Sigma measure val:      0.3188\n",
      " - Volk val:               14.7498\n",
      " - Ricci measure val:      14.4950\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.2527 - volk_loss: 0.0000e+00 - Sigma measure val:      0.3102\n",
      " - Volk val:               16.3841\n",
      " - Ricci measure val:      6.3094\n",
      "352/352 [==============================] - 17s 43ms/step - sigma_loss: 0.2527 - volk_loss: 0.0000e+00 - sigma_val: 0.3102 - volk_val: 16.3841 - ricci_val: 6.3094\n",
      " - Sigma measure val:      0.3102\n",
      " - Volk val:               16.3841\n",
      " - Ricci measure val:      6.3094\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2468 - volk_loss: 7.4838 - Sigma measure val:      0.3114\n",
      " - Volk val:               16.1693\n",
      " - Ricci measure val:      6.6321\n",
      "1/1 [==============================] - 19s 19s/step - sigma_loss: 0.2468 - volk_loss: 7.4838 - sigma_val: 0.3114 - volk_val: 16.1693 - ricci_val: 6.6321\n",
      "\n",
      "Epoch  7/30\n",
      " - Sigma measure val:      0.3114\n",
      " - Volk val:               16.1693\n",
      " - Ricci measure val:      6.6321\n",
      "352/352 [==============================] - ETA: 0s - sigma_loss: 0.2404 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2960\n",
      " - Volk val:               16.7423\n",
      " - Ricci measure val:      7.4289\n",
      "352/352 [==============================] - 17s 43ms/step - sigma_loss: 0.2404 - volk_loss: 0.0000e+00 - sigma_val: 0.2960 - volk_val: 16.7423 - ricci_val: 7.4289\n",
      " - Sigma measure val:      0.2960\n",
      " - Volk val:               16.7423\n",
      " - Ricci measure val:      7.4289\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2330 - volk_loss: 7.6786 - Sigma measure val:      0.2954\n",
      " - Volk val:               16.5991\n",
      " - Ricci measure val:      7.4069\n",
      "1/1 [==============================] - 19s 19s/step - sigma_loss: 0.2330 - volk_loss: 7.6786 - sigma_val: 0.2954 - volk_val: 16.5991 - ricci_val: 7.4069\n",
      "\n",
      "Epoch  8/30\n",
      " - Sigma measure val:      0.2954\n",
      " - Volk val:               16.5991\n",
      " - Ricci measure val:      7.4069\n",
      "352/352 [==============================] - ETA: 0s - sigma_loss: 0.2304 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2823\n",
      " - Volk val:               16.7467\n",
      " - Ricci measure val:      6.8023\n",
      "352/352 [==============================] - 17s 42ms/step - sigma_loss: 0.2304 - volk_loss: 0.0000e+00 - sigma_val: 0.2823 - volk_val: 16.7467 - ricci_val: 6.8023\n",
      " - Sigma measure val:      0.2823\n",
      " - Volk val:               16.7467\n",
      " - Ricci measure val:      6.8023\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2244 - volk_loss: 7.7007 - Sigma measure val:      0.2826\n",
      " - Volk val:               16.5198\n",
      " - Ricci measure val:      6.8259\n",
      "1/1 [==============================] - 26s 26s/step - sigma_loss: 0.2244 - volk_loss: 7.7007 - sigma_val: 0.2826 - volk_val: 16.5198 - ricci_val: 6.8259\n",
      "\n",
      "Epoch  9/30\n",
      " - Sigma measure val:      0.2826\n",
      " - Volk val:               16.5198\n",
      " - Ricci measure val:      6.8259\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.2210 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2722\n",
      " - Volk val:               16.2356\n",
      " - Ricci measure val:      6.6605\n",
      "352/352 [==============================] - 17s 42ms/step - sigma_loss: 0.2210 - volk_loss: 0.0000e+00 - sigma_val: 0.2722 - volk_val: 16.2356 - ricci_val: 6.6605\n",
      " - Sigma measure val:      0.2722\n",
      " - Volk val:               16.2356\n",
      " - Ricci measure val:      6.6605\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2164 - volk_loss: 7.3182 - Sigma measure val:      0.2724\n",
      " - Volk val:               16.1252\n",
      " - Ricci measure val:      6.7019\n",
      "1/1 [==============================] - 19s 19s/step - sigma_loss: 0.2164 - volk_loss: 7.3182 - sigma_val: 0.2724 - volk_val: 16.1252 - ricci_val: 6.7019\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 10/30\n",
      " - Sigma measure val:      0.2724\n",
      " - Volk val:               16.1252\n",
      " - Ricci measure val:      6.7019\n",
      "352/352 [==============================] - ETA: 0s - sigma_loss: 0.2137 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2658\n",
      " - Volk val:               17.3512\n",
      " - Ricci measure val:      6.9787\n",
      "352/352 [==============================] - 17s 42ms/step - sigma_loss: 0.2137 - volk_loss: 0.0000e+00 - sigma_val: 0.2658 - volk_val: 17.3512 - ricci_val: 6.9787\n",
      " - Sigma measure val:      0.2658\n",
      " - Volk val:               17.3512\n",
      " - Ricci measure val:      6.9787\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.2061 - volk_loss: 8.0918 - Sigma measure val:      0.2661\n",
      " - Volk val:               17.1969\n",
      " - Ricci measure val:      6.9601\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.2061 - volk_loss: 8.0918 - sigma_val: 0.2661 - volk_val: 17.1969 - ricci_val: 6.9601\n",
      "\n",
      "Epoch 11/30\n",
      " - Sigma measure val:      0.2661\n",
      " - Volk val:               17.1969\n",
      " - Ricci measure val:      6.9601\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.2057 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2604\n",
      " - Volk val:               16.7528\n",
      " - Ricci measure val:      6.4993\n",
      "352/352 [==============================] - 17s 43ms/step - sigma_loss: 0.2057 - volk_loss: 0.0000e+00 - sigma_val: 0.2604 - volk_val: 16.7528 - ricci_val: 6.4993\n",
      " - Sigma measure val:      0.2604\n",
      " - Volk val:               16.7528\n",
      " - Ricci measure val:      6.4993\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1987 - volk_loss: 7.6944 - Sigma measure val:      0.2603\n",
      " - Volk val:               16.5468\n",
      " - Ricci measure val:      6.6196\n",
      "1/1 [==============================] - 18s 18s/step - sigma_loss: 0.1987 - volk_loss: 7.6944 - sigma_val: 0.2603 - volk_val: 16.5468 - ricci_val: 6.6196\n",
      "\n",
      "Epoch 12/30\n",
      " - Sigma measure val:      0.2603\n",
      " - Volk val:               16.5468\n",
      " - Ricci measure val:      6.6196\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1994 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2498\n",
      " - Volk val:               16.9573\n",
      " - Ricci measure val:      6.9443\n",
      "352/352 [==============================] - 17s 42ms/step - sigma_loss: 0.1992 - volk_loss: 0.0000e+00 - sigma_val: 0.2498 - volk_val: 16.9573 - ricci_val: 6.9443\n",
      " - Sigma measure val:      0.2498\n",
      " - Volk val:               16.9573\n",
      " - Ricci measure val:      6.9443\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1947 - volk_loss: 7.8091 - Sigma measure val:      0.2496\n",
      " - Volk val:               16.7007\n",
      " - Ricci measure val:      7.1671\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1947 - volk_loss: 7.8091 - sigma_val: 0.2496 - volk_val: 16.7007 - ricci_val: 7.1671\n",
      "\n",
      "Epoch 13/30\n",
      " - Sigma measure val:      0.2496\n",
      " - Volk val:               16.7007\n",
      " - Ricci measure val:      7.1671\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1921 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2441\n",
      " - Volk val:               18.1308\n",
      " - Ricci measure val:      6.3658\n",
      "352/352 [==============================] - 25s 66ms/step - sigma_loss: 0.1922 - volk_loss: 0.0000e+00 - sigma_val: 0.2441 - volk_val: 18.1308 - ricci_val: 6.3658\n",
      " - Sigma measure val:      0.2441\n",
      " - Volk val:               18.1308\n",
      " - Ricci measure val:      6.3658\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1955 - volk_loss: 8.5979 - Sigma measure val:      0.2447\n",
      " - Volk val:               18.0843\n",
      " - Ricci measure val:      6.2560\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1955 - volk_loss: 8.5979 - sigma_val: 0.2447 - volk_val: 18.0843 - ricci_val: 6.2560\n",
      "\n",
      "Epoch 14/30\n",
      " - Sigma measure val:      0.2447\n",
      " - Volk val:               18.0843\n",
      " - Ricci measure val:      6.2560\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1871 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2378\n",
      " - Volk val:               17.4779\n",
      " - Ricci measure val:      7.3191\n",
      "352/352 [==============================] - 18s 45ms/step - sigma_loss: 0.1871 - volk_loss: 0.0000e+00 - sigma_val: 0.2378 - volk_val: 17.4779 - ricci_val: 7.3191\n",
      " - Sigma measure val:      0.2378\n",
      " - Volk val:               17.4779\n",
      " - Ricci measure val:      7.3191\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1817 - volk_loss: 8.1902 - Sigma measure val:      0.2377\n",
      " - Volk val:               17.2560\n",
      " - Ricci measure val:      7.2445\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1817 - volk_loss: 8.1902 - sigma_val: 0.2377 - volk_val: 17.2560 - ricci_val: 7.2445\n",
      "\n",
      "Epoch 15/30\n",
      " - Sigma measure val:      0.2377\n",
      " - Volk val:               17.2560\n",
      " - Ricci measure val:      7.2445\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1806 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2307\n",
      " - Volk val:               17.8575\n",
      " - Ricci measure val:      10.1130\n",
      "352/352 [==============================] - 18s 46ms/step - sigma_loss: 0.1805 - volk_loss: 0.0000e+00 - sigma_val: 0.2307 - volk_val: 17.8575 - ricci_val: 10.1130\n",
      " - Sigma measure val:      0.2307\n",
      " - Volk val:               17.8575\n",
      " - Ricci measure val:      10.1130\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1773 - volk_loss: 8.4169 - Sigma measure val:      0.2311\n",
      " - Volk val:               17.6343\n",
      " - Ricci measure val:      19.1542\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1773 - volk_loss: 8.4169 - sigma_val: 0.2311 - volk_val: 17.6343 - ricci_val: 19.1542\n",
      "\n",
      "Epoch 16/30\n",
      " - Sigma measure val:      0.2311\n",
      " - Volk val:               17.6343\n",
      " - Ricci measure val:      19.1542\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1752 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2250\n",
      " - Volk val:               17.7260\n",
      " - Ricci measure val:      4.3149\n",
      "352/352 [==============================] - 19s 47ms/step - sigma_loss: 0.1752 - volk_loss: 0.0000e+00 - sigma_val: 0.2250 - volk_val: 17.7260 - ricci_val: 4.3149\n",
      " - Sigma measure val:      0.2250\n",
      " - Volk val:               17.7260\n",
      " - Ricci measure val:      4.3149\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1666 - volk_loss: 8.3513 - Sigma measure val:      0.2260\n",
      " - Volk val:               17.5230\n",
      " - Ricci measure val:      5.2529\n",
      "1/1 [==============================] - 19s 19s/step - sigma_loss: 0.1666 - volk_loss: 8.3513 - sigma_val: 0.2260 - volk_val: 17.5230 - ricci_val: 5.2529\n",
      "\n",
      "Epoch 17/30\n",
      " - Sigma measure val:      0.2260\n",
      " - Volk val:               17.5230\n",
      " - Ricci measure val:      5.2529\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1707 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2241\n",
      " - Volk val:               18.0473\n",
      " - Ricci measure val:      6.6356\n",
      "352/352 [==============================] - 18s 44ms/step - sigma_loss: 0.1708 - volk_loss: 0.0000e+00 - sigma_val: 0.2241 - volk_val: 18.0473 - ricci_val: 6.6356\n",
      " - Sigma measure val:      0.2241\n",
      " - Volk val:               18.0473\n",
      " - Ricci measure val:      6.6356\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1666 - volk_loss: 8.5487 - Sigma measure val:      0.2250\n",
      " - Volk val:               17.8996\n",
      " - Ricci measure val:      6.5859\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1666 - volk_loss: 8.5487 - sigma_val: 0.2250 - volk_val: 17.8996 - ricci_val: 6.5859\n",
      "\n",
      "Epoch 18/30\n",
      " - Sigma measure val:      0.2250\n",
      " - Volk val:               17.8996\n",
      " - Ricci measure val:      6.5859\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1655 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2156\n",
      " - Volk val:               18.1368\n",
      " - Ricci measure val:      -2.2898e+01\n",
      "352/352 [==============================] - 17s 43ms/step - sigma_loss: 0.1656 - volk_loss: 0.0000e+00 - sigma_val: 0.2156 - volk_val: 18.1368 - ricci_val: -22.8981\n",
      " - Sigma measure val:      0.2156\n",
      " - Volk val:               18.1368\n",
      " - Ricci measure val:      -2.2898e+01\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1638 - volk_loss: 8.5720 - Sigma measure val:      0.2167\n",
      " - Volk val:               17.9434\n",
      " - Ricci measure val:      8.5782\n",
      "1/1 [==============================] - 27s 27s/step - sigma_loss: 0.1638 - volk_loss: 8.5720 - sigma_val: 0.2167 - volk_val: 17.9434 - ricci_val: 8.5782\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 19/30\n",
      " - Sigma measure val:      0.2167\n",
      " - Volk val:               17.9434\n",
      " - Ricci measure val:      8.5782\n",
      "352/352 [==============================] - ETA: 0s - sigma_loss: 0.1621 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2129\n",
      " - Volk val:               17.8786\n",
      " - Ricci measure val:      14.9100\n",
      "352/352 [==============================] - 18s 44ms/step - sigma_loss: 0.1621 - volk_loss: 0.0000e+00 - sigma_val: 0.2129 - volk_val: 17.8786 - ricci_val: 14.9100\n",
      " - Sigma measure val:      0.2129\n",
      " - Volk val:               17.8786\n",
      " - Ricci measure val:      14.9100\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1557 - volk_loss: 8.4488 - Sigma measure val:      0.2135\n",
      " - Volk val:               17.7072\n",
      " - Ricci measure val:      20.9418\n",
      "1/1 [==============================] - 19s 19s/step - sigma_loss: 0.1557 - volk_loss: 8.4488 - sigma_val: 0.2135 - volk_val: 17.7072 - ricci_val: 20.9418\n",
      "\n",
      "Epoch 20/30\n",
      " - Sigma measure val:      0.2135\n",
      " - Volk val:               17.7072\n",
      " - Ricci measure val:      20.9418\n",
      "352/352 [==============================] - ETA: 0s - sigma_loss: 0.1573 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2117\n",
      " - Volk val:               17.6022\n",
      " - Ricci measure val:      -1.9760e+00\n",
      "352/352 [==============================] - 17s 43ms/step - sigma_loss: 0.1573 - volk_loss: 0.0000e+00 - sigma_val: 0.2117 - volk_val: 17.6022 - ricci_val: -1.9760\n",
      " - Sigma measure val:      0.2117\n",
      " - Volk val:               17.6022\n",
      " - Ricci measure val:      -1.9760e+00\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1549 - volk_loss: 8.2800 - Sigma measure val:      0.2119\n",
      " - Volk val:               17.4942\n",
      " - Ricci measure val:      -2.6636e+01\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1549 - volk_loss: 8.2800 - sigma_val: 0.2119 - volk_val: 17.4942 - ricci_val: -26.6361\n",
      "\n",
      "Epoch 21/30\n",
      " - Sigma measure val:      0.2119\n",
      " - Volk val:               17.4942\n",
      " - Ricci measure val:      -2.6636e+01\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1537 - volk_loss: 0.0000e+00 - Sigma measure val:      0.2062\n",
      " - Volk val:               17.7078\n",
      " - Ricci measure val:      6.6027\n",
      "352/352 [==============================] - 18s 46ms/step - sigma_loss: 0.1537 - volk_loss: 0.0000e+00 - sigma_val: 0.2062 - volk_val: 17.7078 - ricci_val: 6.6027\n",
      " - Sigma measure val:      0.2062\n",
      " - Volk val:               17.7078\n",
      " - Ricci measure val:      6.6027\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1528 - volk_loss: 8.3181 - Sigma measure val:      0.2074\n",
      " - Volk val:               17.5619\n",
      " - Ricci measure val:      6.6567\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1528 - volk_loss: 8.3181 - sigma_val: 0.2074 - volk_val: 17.5619 - ricci_val: 6.6567\n",
      "\n",
      "Epoch 22/30\n",
      " - Sigma measure val:      0.2074\n",
      " - Volk val:               17.5619\n",
      " - Ricci measure val:      6.6567\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1509 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1993\n",
      " - Volk val:               17.7589\n",
      " - Ricci measure val:      7.5112\n",
      "352/352 [==============================] - 18s 44ms/step - sigma_loss: 0.1508 - volk_loss: 0.0000e+00 - sigma_val: 0.1993 - volk_val: 17.7589 - ricci_val: 7.5112\n",
      " - Sigma measure val:      0.1993\n",
      " - Volk val:               17.7589\n",
      " - Ricci measure val:      7.5112\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1462 - volk_loss: 8.3591 - Sigma measure val:      0.2000\n",
      " - Volk val:               17.6295\n",
      " - Ricci measure val:      7.7451\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1462 - volk_loss: 8.3591 - sigma_val: 0.2000 - volk_val: 17.6295 - ricci_val: 7.7451\n",
      "\n",
      "Epoch 23/30\n",
      " - Sigma measure val:      0.2000\n",
      " - Volk val:               17.6295\n",
      " - Ricci measure val:      7.7451\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1475 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1965\n",
      " - Volk val:               17.8926\n",
      " - Ricci measure val:      7.4428\n",
      "352/352 [==============================] - 18s 46ms/step - sigma_loss: 0.1475 - volk_loss: 0.0000e+00 - sigma_val: 0.1965 - volk_val: 17.8926 - ricci_val: 7.4428\n",
      " - Sigma measure val:      0.1965\n",
      " - Volk val:               17.8926\n",
      " - Ricci measure val:      7.4428\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1431 - volk_loss: 8.4214 - Sigma measure val:      0.1971\n",
      " - Volk val:               17.7315\n",
      " - Ricci measure val:      8.3540\n",
      "1/1 [==============================] - 19s 19s/step - sigma_loss: 0.1431 - volk_loss: 8.4214 - sigma_val: 0.1971 - volk_val: 17.7315 - ricci_val: 8.3540\n",
      "\n",
      "Epoch 24/30\n",
      " - Sigma measure val:      0.1971\n",
      " - Volk val:               17.7315\n",
      " - Ricci measure val:      8.3540\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1445 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1939\n",
      " - Volk val:               18.1585\n",
      " - Ricci measure val:      -4.7787e-01\n",
      "352/352 [==============================] - 19s 48ms/step - sigma_loss: 0.1445 - volk_loss: 0.0000e+00 - sigma_val: 0.1939 - volk_val: 18.1585 - ricci_val: -0.4779\n",
      " - Sigma measure val:      0.1939\n",
      " - Volk val:               18.1585\n",
      " - Ricci measure val:      -4.7787e-01\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1375 - volk_loss: 8.5814 - Sigma measure val:      0.1947\n",
      " - Volk val:               18.0281\n",
      " - Ricci measure val:      2.8888\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1375 - volk_loss: 8.5814 - sigma_val: 0.1947 - volk_val: 18.0281 - ricci_val: 2.8888\n",
      "\n",
      "Epoch 25/30\n",
      " - Sigma measure val:      0.1947\n",
      " - Volk val:               18.0281\n",
      " - Ricci measure val:      2.8888\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1416 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1988\n",
      " - Volk val:               18.1550\n",
      " - Ricci measure val:      7.2949\n",
      "352/352 [==============================] - 20s 50ms/step - sigma_loss: 0.1416 - volk_loss: 0.0000e+00 - sigma_val: 0.1988 - volk_val: 18.1550 - ricci_val: 7.2949\n",
      " - Sigma measure val:      0.1988\n",
      " - Volk val:               18.1550\n",
      " - Ricci measure val:      7.2949\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1398 - volk_loss: 8.6553 - Sigma measure val:      0.1994\n",
      " - Volk val:               17.9889\n",
      " - Ricci measure val:      8.4446\n",
      "1/1 [==============================] - 21s 21s/step - sigma_loss: 0.1398 - volk_loss: 8.6553 - sigma_val: 0.1994 - volk_val: 17.9889 - ricci_val: 8.4446\n",
      "\n",
      "Epoch 26/30\n",
      " - Sigma measure val:      0.1994\n",
      " - Volk val:               17.9889\n",
      " - Ricci measure val:      8.4446\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1386 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1888\n",
      " - Volk val:               18.2076\n",
      " - Ricci measure val:      7.9849\n",
      "352/352 [==============================] - 18s 45ms/step - sigma_loss: 0.1386 - volk_loss: 0.0000e+00 - sigma_val: 0.1888 - volk_val: 18.2076 - ricci_val: 7.9849\n",
      " - Sigma measure val:      0.1888\n",
      " - Volk val:               18.2076\n",
      " - Ricci measure val:      7.9849\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1346 - volk_loss: 8.6434 - Sigma measure val:      0.1895\n",
      " - Volk val:               18.0978\n",
      " - Ricci measure val:      7.3953\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1346 - volk_loss: 8.6434 - sigma_val: 0.1895 - volk_val: 18.0978 - ricci_val: 7.3953\n",
      "\n",
      "Epoch 27/30\n",
      " - Sigma measure val:      0.1895\n",
      " - Volk val:               18.0978\n",
      " - Ricci measure val:      7.3953\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1364 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1897\n",
      " - Volk val:               18.1575\n",
      " - Ricci measure val:      20.4711\n",
      "352/352 [==============================] - 21s 53ms/step - sigma_loss: 0.1365 - volk_loss: 0.0000e+00 - sigma_val: 0.1897 - volk_val: 18.1575 - ricci_val: 20.4711\n",
      " - Sigma measure val:      0.1897\n",
      " - Volk val:               18.1575\n",
      " - Ricci measure val:      20.4711\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1326 - volk_loss: 8.6115 - Sigma measure val:      0.1902\n",
      " - Volk val:               18.0284\n",
      " - Ricci measure val:      40.7374\n",
      "1/1 [==============================] - 21s 21s/step - sigma_loss: 0.1326 - volk_loss: 8.6115 - sigma_val: 0.1902 - volk_val: 18.0284 - ricci_val: 40.7374\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 28/30\n",
      " - Sigma measure val:      0.1902\n",
      " - Volk val:               18.0284\n",
      " - Ricci measure val:      40.7374\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1343 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1873\n",
      " - Volk val:               18.3949\n",
      " - Ricci measure val:      6.2894\n",
      "352/352 [==============================] - 18s 45ms/step - sigma_loss: 0.1343 - volk_loss: 0.0000e+00 - sigma_val: 0.1873 - volk_val: 18.3949 - ricci_val: 6.2894\n",
      " - Sigma measure val:      0.1873\n",
      " - Volk val:               18.3949\n",
      " - Ricci measure val:      6.2894\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1329 - volk_loss: 8.7474 - Sigma measure val:      0.1877\n",
      " - Volk val:               18.3103\n",
      " - Ricci measure val:      6.2977\n",
      "1/1 [==============================] - 21s 21s/step - sigma_loss: 0.1329 - volk_loss: 8.7474 - sigma_val: 0.1877 - volk_val: 18.3103 - ricci_val: 6.2977\n",
      "\n",
      "Epoch 29/30\n",
      " - Sigma measure val:      0.1877\n",
      " - Volk val:               18.3103\n",
      " - Ricci measure val:      6.2977\n",
      "350/352 [============================>.] - ETA: 0s - sigma_loss: 0.1320 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1858\n",
      " - Volk val:               18.6569\n",
      " - Ricci measure val:      6.6516\n",
      "352/352 [==============================] - 18s 44ms/step - sigma_loss: 0.1321 - volk_loss: 0.0000e+00 - sigma_val: 0.1858 - volk_val: 18.6569 - ricci_val: 6.6516\n",
      " - Sigma measure val:      0.1858\n",
      " - Volk val:               18.6569\n",
      " - Ricci measure val:      6.6516\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1300 - volk_loss: 8.9285 - Sigma measure val:      0.1866\n",
      " - Volk val:               18.4850\n",
      " - Ricci measure val:      6.6356\n",
      "1/1 [==============================] - 20s 20s/step - sigma_loss: 0.1300 - volk_loss: 8.9285 - sigma_val: 0.1866 - volk_val: 18.4850 - ricci_val: 6.6356\n",
      "\n",
      "Epoch 30/30\n",
      " - Sigma measure val:      0.1866\n",
      " - Volk val:               18.4850\n",
      " - Ricci measure val:      6.6356\n",
      "351/352 [============================>.] - ETA: 0s - sigma_loss: 0.1301 - volk_loss: 0.0000e+00 - Sigma measure val:      0.1786\n",
      " - Volk val:               18.5850\n",
      " - Ricci measure val:      6.3166\n",
      "352/352 [==============================] - 17s 42ms/step - sigma_loss: 0.1301 - volk_loss: 0.0000e+00 - sigma_val: 0.1786 - volk_val: 18.5850 - ricci_val: 6.3166\n",
      " - Sigma measure val:      0.1786\n",
      " - Volk val:               18.5850\n",
      " - Ricci measure val:      6.3166\n",
      "1/1 [==============================] - ETA: 0s - sigma_loss: 0.1241 - volk_loss: 8.8442 - Sigma measure val:      0.1793\n",
      " - Volk val:               18.4824\n",
      " - Ricci measure val:      6.2446\n",
      "1/1 [==============================] - 21s 21s/step - sigma_loss: 0.1241 - volk_loss: 8.8442 - sigma_val: 0.1793 - volk_val: 18.4824 - ricci_val: 6.2446\n"
     ]
    }
   ],
   "source": [
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '0'  # mostly silences TF logger\n",
    "fs_model, training_history = train_model(fs_model, data, epochs=n_epochs, batch_sizes=batch_sizes,\n",
    "                                         verbose=1, custom_metrics=cmetrics, callbacks=cb_list)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Check the volume normailzation of our data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vol_K from triple intersection numbers at t1=t2=1:                 20\n",
      "Vol_K from MC integration over FS metric in Kahler class t1=t2=1:  19.99\n",
      "Vol_K from MC integration over CY metric in Kahler class t1=t2=1:  18.39\n"
     ]
    }
   ],
   "source": [
    "dets_fs = np.real(tf.linalg.det(fs_model.fubini_study_pb(tf.cast(data['X_train'], tf.float32))).numpy())\n",
    "dets_cy = np.real(tf.linalg.det(fs_model(tf.cast(data['X_train'], tf.float32))).numpy())\n",
    "weights, omegas = data['y_train'][:,-2], data['y_train'][:,-1]\n",
    "vol_k_fs = np.mean(weights * dets_fs / omegas)\n",
    "vol_k_cy = np.mean(weights * dets_cy / omegas)\n",
    "print(\"Vol_K from triple intersection numbers at t1=t2=1:                \", toric_data['vol_j_norm'])\n",
    "print(\"Vol_K from MC integration over FS metric in Kahler class t1=t2=1: \", np.round(vol_k_fs, 2))\n",
    "print(\"Vol_K from MC integration over CY metric in Kahler class t1=t2=1: \", np.round(vol_k_cy, 2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "... and look at the value of the Kahler potential"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: shape=(10,), dtype=float32, numpy=\n",
       "array([0.25653175, 0.2446172 , 0.30648085, 0.28357574, 0.31832972,\n",
       "       0.47995326, 0.39768592, 0.44896987, 0.0806471 , 0.18551616],\n",
       "      dtype=float32)>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fs_model.get_kahler_potential(data['X_train'][:10])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plot the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import pandas as pd\n",
    "sns.set_style(\"darkgrid\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note that vol_k val should be constant (and equal to 20.0). So in this case, it is off by 15%)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "hist = training_history\n",
    "hist['epochs'] = list(range(n_epochs))\n",
    "hist_data = pd.DataFrame(hist)\n",
    "fig, [ax1, ax2] = plt.subplots(1, 2, figsize=(10,5));\n",
    "ax1.set_title(\"Monge-Ampere loss\")\n",
    "sns.scatterplot(x='epochs', y='sigma_loss', data=hist_data, alpha=0.75, ax=ax1, label=\"train\")\n",
    "sns.scatterplot(x='epochs', y='sigma_val', data=hist_data, alpha=0.75, ax=ax1, label=\"val\")\n",
    "\n",
    "ax2.set_title(\"Vol_K value of validation set\")\n",
    "sns.scatterplot(x='epochs', y='volk_val', data=hist_data, alpha=0.75, ax=ax2, label=\"val\");"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "SageMath 9.5",
   "language": "sage",
   "name": "sagemath"
  },
  "language": "python",
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}