An_Odd_Music_Generator.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/amirkasaei/An-Odd-Music-Generator/blob/main/An_Odd_Music_Generator.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rG540utTBHVi"
      },
      "source": [
        "# **Libraries & Constants**\n",
        "\n",
        "---\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Q1qXvuLDsHy-"
      },
      "source": [
        "## **imports**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PkpWZM1-y5D6",
        "outputId": "1526c8ff-a7f4-474e-f451-df3712bc45dc"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Requirement already satisfied: librosa in /usr/local/lib/python3.7/dist-packages (0.8.1)\n",
            "Requirement already satisfied: scipy>=1.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (1.4.1)\n",
            "Requirement already satisfied: numpy>=1.15.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (1.21.6)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (21.3)\n",
            "Requirement already satisfied: joblib>=0.14 in /usr/local/lib/python3.7/dist-packages (from librosa) (1.1.0)\n",
            "Requirement already satisfied: decorator>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (4.4.2)\n",
            "Requirement already satisfied: numba>=0.43.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (0.51.2)\n",
            "Requirement already satisfied: resampy>=0.2.2 in /usr/local/lib/python3.7/dist-packages (from librosa) (0.2.2)\n",
            "Requirement already satisfied: soundfile>=0.10.2 in /usr/local/lib/python3.7/dist-packages (from librosa) (0.10.3.post1)\n",
            "Requirement already satisfied: audioread>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (2.1.9)\n",
            "Requirement already satisfied: scikit-learn!=0.19.0,>=0.14.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (1.0.2)\n",
            "Requirement already satisfied: pooch>=1.0 in /usr/local/lib/python3.7/dist-packages (from librosa) (1.6.0)\n",
            "Requirement already satisfied: llvmlite<0.35,>=0.34.0.dev0 in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa) (0.34.0)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa) (57.4.0)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=20.0->librosa) (3.0.9)\n",
            "Requirement already satisfied: requests>=2.19.0 in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa) (2.23.0)\n",
            "Requirement already satisfied: appdirs>=1.3.0 in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa) (1.4.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa) (2022.6.15)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa) (2.10)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa) (3.0.4)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.19.0->pooch>=1.0->librosa) (1.24.3)\n",
            "Requirement already satisfied: six>=1.3 in /usr/local/lib/python3.7/dist-packages (from resampy>=0.2.2->librosa) (1.15.0)\n",
            "Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn!=0.19.0,>=0.14.0->librosa) (3.1.0)\n",
            "Requirement already satisfied: cffi>=1.0 in /usr/local/lib/python3.7/dist-packages (from soundfile>=0.10.2->librosa) (1.15.0)\n",
            "Requirement already satisfied: pycparser in /usr/local/lib/python3.7/dist-packages (from cffi>=1.0->soundfile>=0.10.2->librosa) (2.21)\n"
          ]
        }
      ],
      "source": [
        "pip install librosa"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "U9giafB6qqAv"
      },
      "outputs": [],
      "source": [
        "import pandas as pd\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "from keras.models import Sequential, load_model\n",
        "from keras.layers import Dense, Input, Softmax, BatchNormalization, Dropout, Flatten, LSTM, GRU, SimpleRNN, Concatenate, concatenate, MaxPooling1D, GlobalMaxPool1D, Bidirectional, GlobalMaxPooling1D, GlobalAveragePooling1D, Embedding\n",
        "# from keras.layers.merge import concatenate\n",
        "from keras.losses import CategoricalCrossentropy\n",
        "from keras.metrics import CategoricalAccuracy\n",
        "from keras.preprocessing.text import Tokenizer\n",
        "from tensorflow.keras.optimizers import RMSprop, Adam, Adamax, Nadam, SGD\n",
        "from tensorflow.keras import Model\n",
        "from tensorflow.keras.optimizers.schedules import InverseTimeDecay, ExponentialDecay\n",
        "from tensorflow.keras.utils import to_categorical, plot_model\n",
        "from keras.callbacks import ModelCheckpoint\n",
        "from keras.preprocessing.sequence import pad_sequences\n",
        "from sklearn.preprocessing import LabelEncoder\n",
        "from sklearn.model_selection import train_test_split, KFold\n",
        "from sklearn import metrics\n",
        "from matplotlib import pyplot as plt\n",
        "from sklearn.model_selection import KFold\n",
        "import os\n",
        "import string\n",
        "import cv2 as cv\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from copy import deepcopy\n",
        "from __future__ import unicode_literals\n",
        "from scipy.io import wavfile\n",
        "import scipy.signal as sps\n",
        "import librosa"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mHrW-d9w4Css"
      },
      "source": [
        "## **Google drive mount**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5cL9P0mc4NBD",
        "outputId": "631aabfb-da30-4c92-d4cf-ba0cfa9df2f6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LXvHNESIZ_BV"
      },
      "source": [
        "# **Helper functions**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "UTPdekKNZ9o9"
      },
      "outputs": [],
      "source": [
        "def argmaxKeepDimensions(arr):\n",
        "  tmp = np.zeros_like(arr)\n",
        "  tmp[np.arange(len(arr)), arr.argmax(1)] = 1\n",
        "  return tmp\n",
        "\n",
        "def outputResult(model, history, X_train, y_train, X_test, y_test, noteDict):\n",
        "  # print(\"result in training set:\", \"\\ntrain acc= \",\n",
        "  #       self.history['accuracy'], \"\\ntrain loss= \", self.history['loss'])\n",
        "  # print(\"\\nresult in test set:\", \"\\ntest acc= \",\n",
        "  #       self.history['val_accuracy'], \"\\ntest loss= \", self.history['val_loss'], '\\n')\n",
        "\n",
        "  # best models based on acc or loss in tarin set or test set\n",
        "  trainHistory = list(\n",
        "      map(lambda x, y: [x, y], history['accuracy'], history['loss']))\n",
        "  testHistory = list(\n",
        "      map(lambda x, y: [x, y], history['val_accuracy'], history['val_loss']))\n",
        "  print(\n",
        "      f\"\\nbest model based on min training set loss:  acc= {min(trainHistory, key = lambda k: k[1])[0]}  loss= {min(trainHistory, key = lambda k: k[1])[1]}\")\n",
        "  print(\n",
        "      f\"best model based on min test set loss:  acc= {min(testHistory, key = lambda k: k[1])[0]}  loss= {min(testHistory, key = lambda k: k[1])[1]}\")\n",
        "  print(\n",
        "      f\"best model based on max training set accuracy:  acc= {max(trainHistory, key = lambda k: k[0])[0]}  loss= {max(trainHistory, key = lambda k: k[0])[1]}\")\n",
        "  print(\n",
        "      f\"best model based on max test set accuracy:  acc= {max(testHistory, key = lambda k: k[0])[0]}  loss= {max(testHistory, key = lambda k: k[0])[1]}\")\n",
        "\n",
        "  print(\"\\nevaluate dataset with best model based on maximum test set accuracy\")\n",
        "  print(\"evaluate train set= \", model.evaluate(\n",
        "      X_train, y_train, verbose=0))\n",
        "  print(\"evaluate test set= \", model.evaluate(X_test, y_test, verbose=0))\n",
        "\n",
        "  y_train_pred = model.predict(X_train)\n",
        "  # convert probablities to 0s and 1s\n",
        "  y_train_pred = argmaxKeepDimensions(y_train_pred)\n",
        "\n",
        "  y_test_pred = model.predict(X_test)\n",
        "  # convert probablities to 0s and 1s\n",
        "  y_test_pred = argmaxKeepDimensions(y_test_pred)\n",
        "\n",
        "  # confusion matrix and precision, recall and f1 report\n",
        "  print('\\n', '-'*30, 'metrics for traning set', '-'*30)\n",
        "\n",
        "  print(\"confusion matrix: \\n\", metrics.confusion_matrix(\n",
        "      np.argmax(y_train, axis=1), np.argmax(y_train_pred, axis=1)))\n",
        "  print(metrics.classification_report(y_train,\n",
        "        y_train_pred, digits=3, target_names=list(noteDict.keys())))\n",
        "\n",
        "  # confusion matrix and precision, recall and f1 report\n",
        "  print('-'*30, 'metrics for test set', '-'*30)\n",
        "\n",
        "  print(\"confusion matrix: \\n\", metrics.confusion_matrix(\n",
        "      np.argmax(y_test, axis=1), np.argmax(y_test_pred, axis=1)))\n",
        "  print(metrics.classification_report(y_test,\n",
        "        y_test_pred, digits=3, target_names=list(noteDict.keys())))\n",
        "\n",
        "  # self.model.summary()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "kfL7Q0Tqn2-z"
      },
      "outputs": [],
      "source": [
        "def showPlots(model, history):\n",
        "    plt.plot(history['accuracy'],\n",
        "              label='training accuracy', marker='.', color='green')\n",
        "    plt.plot(history['val_accuracy'],\n",
        "              label='test accuracy', marker='.', color='blue')\n",
        "    plt.xlabel('Epochs')\n",
        "    plt.ylabel('Accuracy')\n",
        "    plt.legend()\n",
        "    plt.show()\n",
        "\n",
        "    plt.plot(history['loss'],\n",
        "              label='training loss', marker='.', color='green')\n",
        "    plt.plot(history['val_loss'],\n",
        "              label='test loss', marker='.', color='blue')\n",
        "    plt.xlabel('Epochs')\n",
        "    plt.ylabel('Loss')\n",
        "    plt.legend()\n",
        "    plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9g2bT1KDsTU1"
      },
      "source": [
        "# **Denoising Network**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WcHmHbcQsYPY"
      },
      "source": [
        "## **Preprocessing**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "uyTcURre1olb"
      },
      "outputs": [],
      "source": [
        "sample_rate = 4000  # sample rate"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LMMvqFfpEgVv"
      },
      "source": [
        "**Paths**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "U0gbIM9TEdrg"
      },
      "outputs": [],
      "source": [
        "notesPath = '/content/drive/MyDrive/Deep Learning/Music Generator/piano_triads'\n",
        "noisesPath = '/content/drive/MyDrive/Deep Learning/Music Generator/artifacts'\n",
        "songsPath = '/content/drive/MyDrive/Deep Learning/Music Generator/piano_pieces'"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7vAxP9i7tAyg"
      },
      "source": [
        "**Helper Functions**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "MnKTrgO1siMx"
      },
      "outputs": [],
      "source": [
        "# mix piece and noise together\n",
        "def AddNoiseToPiece(piece, noise, noise_factor):\n",
        "  noisy_piece = piece + (noise_factor*noise)\n",
        "  return noisy_piece\n",
        "\n",
        "# append wav np.array of notes by having note names in a piece\n",
        "def appendNotes(piece, notes, notesLabel):\n",
        "  appendedNotes = []\n",
        "  for piece_note in piece:\n",
        "    index = notesLabel.index(piece_note)\n",
        "    appendedNotes.extend(notes[index])\n",
        "  return np.array(appendedNotes)\n",
        "\n",
        "# add padding to wav piece to grow up its length to max_length\n",
        "def addPadding(piece_wav):\n",
        "  max_length = sample_rate * 9\n",
        "  return np.pad(piece_wav, (0,max_length-piece_wav.shape[0]))\n",
        "\n",
        "# split np.array to multiple specific length array\n",
        "def splitToSpecificLength(arr, length):\n",
        "  arr = np.array(arr)\n",
        "  numberOfPiece = arr.shape[0]//length + 1 if arr.shape[0]%length != 0 else arr.shape[0]//length\n",
        "  return np.split(arr, [length*i for i in range(1, numberOfPiece)])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "alM-xLiJve5K"
      },
      "source": [
        "**Read all notes and noise**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LAauC64WvdSC",
        "outputId": "45fe8b42-91ed-49bd-d646-69e271d39adb"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(432, 12000)\n",
            "(42, 36000)\n"
          ]
        }
      ],
      "source": [
        "notes, notesLabel, noises = [], [], []\n",
        "for note in os.listdir(notesPath):\n",
        "  data  = librosa.load(notesPath+'/'+note, sr=sample_rate, dtype=np.float32)[0]\n",
        "  notes.append(data)\n",
        "  notesLabel.append(note[:-4])\n",
        "\n",
        "for noise in os.listdir(noisesPath):\n",
        "  # noises must have 1/3 note sample rate because their length are 3X\n",
        "  noises.append(librosa.load(noisesPath + '/'+ noise, sr=sample_rate, dtype=np.float32)[0]) # sample rate for a 9s song: 1 noises per song\n",
        "\n",
        "notes = np.array(notes)\n",
        "noises = np.array(noises)\n",
        "# notesLabel = np.array(notesLabel)\n",
        "print(notes.shape)\n",
        "print(noises.shape)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cFsnuCFIvzu0"
      },
      "source": [
        "**Read all song notes**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "yjWsCrQkmKxg"
      },
      "outputs": [],
      "source": [
        "songs = []\n",
        "for song in os.listdir(songsPath):\n",
        "  data  = open(songsPath + '/' + song, 'r')\n",
        "  data = data.readlines()\n",
        "  if data:\n",
        "    songs.append(data[0].split(' / '))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9uN7Lpgv_oje"
      },
      "source": [
        "**Estimate number of dataset**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pDEY55_S_nnt",
        "outputId": "e14c46e0-f0ec-4d12-d84f-8ca2bc2d3e3d"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1240.0"
            ]
          },
          "metadata": {},
          "execution_count": 60
        }
      ],
      "source": [
        "def estimateNumberOfSamples(songs, i, j):\n",
        "  numberOfNoises = 40\n",
        "\n",
        "  songsLength = [len(x) for x in songs]\n",
        "  songsLength = np.array(songsLength)\n",
        "  numberOfPiecesPerSong = np.ceil(songsLength/3)\n",
        "\n",
        "  #print(numberOfPiecesPerSong[0])\n",
        "  numberOfPieces = np.sum(numberOfPiecesPerSong[i : j])\n",
        "  numberOfSamples = numberOfPieces * numberOfNoises\n",
        "\n",
        "  return numberOfSamples\n",
        "\n",
        "estimateNumberOfSamples(songs, 0, 1)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OGWG_4YXMPQV",
        "outputId": "5937c6c7-48b2-4355-8fcd-3221f44901fa"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(36000,)"
            ]
          },
          "metadata": {},
          "execution_count": 22
        }
      ],
      "source": [
        "appendNotes(songs[0][:3], notes, notesLabel).shape"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U8ll2ZHIz8Jm"
      },
      "source": [
        "**DAE Data Generator**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "QMeGF9grwTKI"
      },
      "outputs": [],
      "source": [
        "def DAE_data_generator(songs, batchsize):\n",
        "    X_DAE_batch, y_DAE_batch = [], []\n",
        "    batchcount = 0\n",
        "    while True:\n",
        "        for song in songs:\n",
        "          # separate the notes of song with 30 length\n",
        "          song_pieces = splitToSpecificLength(song, 10) + splitToSpecificLength(song, 20)# + splitToSpecificLength(song, 4) + splitToSpecificLength(song, 5) + splitToSpecificLength(song, 20) + splitToSpecificLength(song, 30)\n",
        "\n",
        "          # create all pieces in the song\n",
        "          for piece in song_pieces:\n",
        "              piece_wav = appendNotes(piece, notes, notesLabel)\n",
        "              piece_wav_pad = addPadding(piece_wav)\n",
        "              for noise in noises:\n",
        "                noisy_piece_wav = AddNoiseToPiece(piece_wav_pad, noise, 0.25)\n",
        "\n",
        "                X_DAE_batch.append(noisy_piece_wav)\n",
        "                # X_DAE_batch.append(piece_wav_pad)\n",
        "                y_DAE_batch.append(piece_wav_pad)\n",
        "\n",
        "                batchcount += 1\n",
        "                if batchcount == batchsize:\n",
        "                  X_DAE_batch = np.array(X_DAE_batch)\n",
        "                  y_DAE_batch = np.array(y_DAE_batch)\n",
        "\n",
        "                  yield (X_DAE_batch, y_DAE_batch)\n",
        "                  X_DAE_batch = []\n",
        "                  y_DAE_batch = []\n",
        "                  batchcount = 0"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def DAE_data_generator1(songs, batchsize):\n",
        "    X_DAE_batch, y_DAE_batch = [], []\n",
        "    batchcount = 0\n",
        "    while True:\n",
        "        for song in songs[:5]:\n",
        "          # separate the notes of song with 3 length\n",
        "          song_pieces = splitToSpecificLength(song, 3)# + splitToSpecificLength(song, 4) + splitToSpecificLength(song, 5) + splitToSpecificLength(song, 20) + splitToSpecificLength(song, 30)\n",
        "\n",
        "          # create all pieces in the song\n",
        "          for piece in song_pieces:\n",
        "              piece_wav = appendNotes(piece, notes, notesLabel)\n",
        "              piece_wav_pad = addPadding(piece_wav)\n",
        "              for noise in noises:\n",
        "                noisy_piece_wav = AddNoiseToPiece(piece_wav_pad, noise, 0.25)\n",
        "\n",
        "                X_DAE_batch.append(noisy_piece_wav)\n",
        "                # X_DAE_batch.append(piece_wav_pad)\n",
        "                y_DAE_batch.append(piece_wav_pad)\n",
        "\n",
        "                batchcount += 1\n",
        "                if batchcount == batchsize:\n",
        "                  X_DAE_batch = np.array(X_DAE_batch)\n",
        "                  y_DAE_batch = np.array(y_DAE_batch)\n",
        "\n",
        "                  yield (X_DAE_batch, y_DAE_batch)\n",
        "                  X_DAE_batch = []\n",
        "                  y_DAE_batch = []\n",
        "                  batchcount = 0"
      ],
      "metadata": {
        "id": "IQYqHPrx0rAZ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6YZs2_NCO_2K"
      },
      "source": [
        "## **Implementation**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 533
        },
        "id": "QdDxczVJPJ7J",
        "outputId": "7ec74182-f989-4cc2-f6d9-cc5f789fefa8"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<IPython.core.display.Image object>"
            ],
            "image/png": "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\n"
          },
          "metadata": {},
          "execution_count": 89
        }
      ],
      "source": [
        "batchsize = 200 #720\n",
        "\n",
        "inputShape = ouputShape = 36000\n",
        "\n",
        "input = Input(shape=(inputShape, ), name=\"input_noisy_note\")\n",
        "\n",
        "# x = tf.keras.layers.Conv1D(1, 500, strides=100, activation='linear')(input)\n",
        "# x = tf.keras.layers.Flatten()(x)\n",
        "x = Dense(512, activation='relu', name=\"encoder_1\")(input)\n",
        "# x = Dense(128, activation='linear', name=\"encoder_2\")(x)\n",
        "# x = Dense(32, activation='tanh', name=\"encoder_3\")(x)\n",
        "x = Dense(64, activation='linear', name=\"encoded\")(x)\n",
        "# x = Dense(32, activation='tanh', name=\"decoder_1\")(x)\n",
        "# x = Dense(128, activation='linear', name=\"decoder_1\")(x)\n",
        "x = Dense(512, activation='linear', name=\"decoder_2\")(x)\n",
        "out = Dense(ouputShape, activation='linear', name=\"reconstructed_note\")(x)\n",
        "\n",
        "DAE_model = Model(inputs=input, outputs=out)\n",
        "\n",
        "# Compile the Model\n",
        "metrics = ['mean_absolute_error', 'mean_squared_logarithmic_error', 'cosine_similarity']\n",
        "DAE_model.compile(loss=\"mean_squared_error\", optimizer=Adam(0.01), metrics=metrics)\n",
        "\n",
        "filepath = '/content/drive/MyDrive/Deep Learning/Music Generator/DAE_best_model_test.hdf5'\n",
        "checkpoint = ModelCheckpoint(filepath=filepath,\n",
        "                              monitor='loss',\n",
        "                              verbose=0,\n",
        "                              save_best_only=True,\n",
        "                              mode='min')\n",
        "\n",
        "plot_model(DAE_model, to_file=\"DAE_model.png\", show_shapes=True)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "NoLzS3xUQ-DI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "798835d0-2963-4c37-89c5-79cc9631015c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/1000\n",
            "27/27 [==============================] - 4s 145ms/step - loss: 0.0010 - mean_absolute_error: 0.0171 - mean_squared_logarithmic_error: 4.0314e-04 - cosine_similarity: 0.1474\n",
            "Epoch 2/1000\n",
            "27/27 [==============================] - 4s 159ms/step - loss: 9.7897e-04 - mean_absolute_error: 0.0177 - mean_squared_logarithmic_error: 3.7769e-04 - cosine_similarity: 0.2177\n",
            "Epoch 3/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.0030 - mean_absolute_error: 0.0289 - mean_squared_logarithmic_error: 0.0010 - cosine_similarity: 0.1476\n",
            "Epoch 4/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 20.0144 - mean_absolute_error: 1.8811 - mean_squared_logarithmic_error: 0.4962 - cosine_similarity: 0.0052\n",
            "Epoch 5/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 609.1896 - mean_absolute_error: 12.1340 - mean_squared_logarithmic_error: 2.3794 - cosine_similarity: 0.0012\n",
            "Epoch 6/1000\n",
            "27/27 [==============================] - 3s 97ms/step - loss: 3080.3154 - mean_absolute_error: 28.5784 - mean_squared_logarithmic_error: 4.0164 - cosine_similarity: -2.3870e-05\n",
            "Epoch 7/1000\n",
            "27/27 [==============================] - 3s 95ms/step - loss: 18872.2500 - mean_absolute_error: 57.7318 - mean_squared_logarithmic_error: 5.5063 - cosine_similarity: -0.0034\n",
            "Epoch 8/1000\n",
            "27/27 [==============================] - 2s 86ms/step - loss: 43327.1758 - mean_absolute_error: 90.9908 - mean_squared_logarithmic_error: 6.3257 - cosine_similarity: -0.0022\n",
            "Epoch 9/1000\n",
            "27/27 [==============================] - 2s 85ms/step - loss: 48836.7812 - mean_absolute_error: 81.4507 - mean_squared_logarithmic_error: 5.6903 - cosine_similarity: 2.6185e-04\n",
            "Epoch 10/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 86375.1875 - mean_absolute_error: 91.0965 - mean_squared_logarithmic_error: 5.6579 - cosine_similarity: -8.3901e-04\n",
            "Epoch 11/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 81432.2344 - mean_absolute_error: 86.4898 - mean_squared_logarithmic_error: 5.2667 - cosine_similarity: 4.7687e-04\n",
            "Epoch 12/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 13997.9482 - mean_absolute_error: 35.4292 - mean_squared_logarithmic_error: 3.4788 - cosine_similarity: 0.0053\n",
            "Epoch 13/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 3425.2556 - mean_absolute_error: 16.3096 - mean_squared_logarithmic_error: 2.0778 - cosine_similarity: 0.0010\n",
            "Epoch 14/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 454.1240 - mean_absolute_error: 7.5654 - mean_squared_logarithmic_error: 1.4219 - cosine_similarity: 0.0036\n",
            "Epoch 15/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 248.9237 - mean_absolute_error: 5.8681 - mean_squared_logarithmic_error: 1.2139 - cosine_similarity: 0.0027\n",
            "Epoch 16/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 147.5485 - mean_absolute_error: 4.5690 - mean_squared_logarithmic_error: 1.0230 - cosine_similarity: 0.0015\n",
            "Epoch 17/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 100.7124 - mean_absolute_error: 4.1651 - mean_squared_logarithmic_error: 0.9631 - cosine_similarity: 4.8623e-04\n",
            "Epoch 18/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 86.1454 - mean_absolute_error: 3.7407 - mean_squared_logarithmic_error: 0.8742 - cosine_similarity: 0.0019\n",
            "Epoch 19/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 72.4998 - mean_absolute_error: 3.4876 - mean_squared_logarithmic_error: 0.8211 - cosine_similarity: 0.0018\n",
            "Epoch 20/1000\n",
            "27/27 [==============================] - 2s 88ms/step - loss: 57.4808 - mean_absolute_error: 3.0952 - mean_squared_logarithmic_error: 0.7410 - cosine_similarity: 0.0012\n",
            "Epoch 21/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 51.2109 - mean_absolute_error: 2.9216 - mean_squared_logarithmic_error: 0.7017 - cosine_similarity: 9.6311e-04\n",
            "Epoch 22/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 44.3745 - mean_absolute_error: 2.6854 - mean_squared_logarithmic_error: 0.6478 - cosine_similarity: -0.0015\n",
            "Epoch 23/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 39.8040 - mean_absolute_error: 2.5313 - mean_squared_logarithmic_error: 0.6128 - cosine_similarity: 0.0013\n",
            "Epoch 24/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 36.1325 - mean_absolute_error: 2.3971 - mean_squared_logarithmic_error: 0.5826 - cosine_similarity: 0.0010\n",
            "Epoch 25/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 32.1992 - mean_absolute_error: 2.2519 - mean_squared_logarithmic_error: 0.5501 - cosine_similarity: 2.8787e-05\n",
            "Epoch 26/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 28.2959 - mean_absolute_error: 2.0769 - mean_squared_logarithmic_error: 0.5089 - cosine_similarity: 5.9015e-04\n",
            "Epoch 27/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 25.5108 - mean_absolute_error: 1.9498 - mean_squared_logarithmic_error: 0.4785 - cosine_similarity: -8.7575e-05\n",
            "Epoch 28/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 24.1946 - mean_absolute_error: 1.8876 - mean_squared_logarithmic_error: 0.4639 - cosine_similarity: -4.8979e-04\n",
            "Epoch 29/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 22.2503 - mean_absolute_error: 1.8196 - mean_squared_logarithmic_error: 0.4505 - cosine_similarity: 2.8972e-04\n",
            "Epoch 30/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 20.3946 - mean_absolute_error: 1.7196 - mean_squared_logarithmic_error: 0.4262 - cosine_similarity: 5.7350e-04\n",
            "Epoch 31/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 18.1135 - mean_absolute_error: 1.6022 - mean_squared_logarithmic_error: 0.3980 - cosine_similarity: 0.0021\n",
            "Epoch 32/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 17.1182 - mean_absolute_error: 1.5305 - mean_squared_logarithmic_error: 0.3790 - cosine_similarity: 0.0017\n",
            "Epoch 33/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 14.6843 - mean_absolute_error: 1.4383 - mean_squared_logarithmic_error: 0.3601 - cosine_similarity: 0.0023\n",
            "Epoch 34/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 15.3093 - mean_absolute_error: 1.4509 - mean_squared_logarithmic_error: 0.3627 - cosine_similarity: 0.0032\n",
            "Epoch 35/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 12.2817 - mean_absolute_error: 1.2981 - mean_squared_logarithmic_error: 0.3256 - cosine_similarity: 0.0022\n",
            "Epoch 36/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 12.8443 - mean_absolute_error: 1.3207 - mean_squared_logarithmic_error: 0.3327 - cosine_similarity: 0.0026\n",
            "Epoch 37/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 12.9007 - mean_absolute_error: 1.3310 - mean_squared_logarithmic_error: 0.3361 - cosine_similarity: 0.0032\n",
            "Epoch 38/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 11.7665 - mean_absolute_error: 1.2518 - mean_squared_logarithmic_error: 0.3164 - cosine_similarity: 0.0043\n",
            "Epoch 39/1000\n",
            "27/27 [==============================] - 2s 78ms/step - loss: 11.6164 - mean_absolute_error: 1.2383 - mean_squared_logarithmic_error: 0.3126 - cosine_similarity: 0.0046\n",
            "Epoch 40/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 9.8313 - mean_absolute_error: 1.1075 - mean_squared_logarithmic_error: 0.2788 - cosine_similarity: 0.0043\n",
            "Epoch 41/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 9.9831 - mean_absolute_error: 1.1097 - mean_squared_logarithmic_error: 0.2785 - cosine_similarity: 0.0038\n",
            "Epoch 42/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 10.5248 - mean_absolute_error: 1.1412 - mean_squared_logarithmic_error: 0.2892 - cosine_similarity: 0.0052\n",
            "Epoch 43/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 11.6865 - mean_absolute_error: 1.1479 - mean_squared_logarithmic_error: 0.2861 - cosine_similarity: 0.0052\n",
            "Epoch 44/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 12.0373 - mean_absolute_error: 1.1551 - mean_squared_logarithmic_error: 0.2888 - cosine_similarity: 0.0046\n",
            "Epoch 45/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 14.8624 - mean_absolute_error: 1.1356 - mean_squared_logarithmic_error: 0.2734 - cosine_similarity: 0.0089\n",
            "Epoch 46/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 19.4737 - mean_absolute_error: 1.2912 - mean_squared_logarithmic_error: 0.3074 - cosine_similarity: 0.0029\n",
            "Epoch 47/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 31.7419 - mean_absolute_error: 1.3847 - mean_squared_logarithmic_error: 0.3094 - cosine_similarity: 0.0094\n",
            "Epoch 48/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 44.5673 - mean_absolute_error: 1.6146 - mean_squared_logarithmic_error: 0.3429 - cosine_similarity: 0.0017\n",
            "Epoch 49/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 71.4189 - mean_absolute_error: 1.8251 - mean_squared_logarithmic_error: 0.3637 - cosine_similarity: 0.0124\n",
            "Epoch 50/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 91.0270 - mean_absolute_error: 2.0564 - mean_squared_logarithmic_error: 0.3822 - cosine_similarity: -0.0018\n",
            "Epoch 51/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 114.7649 - mean_absolute_error: 2.2155 - mean_squared_logarithmic_error: 0.4029 - cosine_similarity: 0.0127\n",
            "Epoch 52/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 89.2558 - mean_absolute_error: 2.0546 - mean_squared_logarithmic_error: 0.3792 - cosine_similarity: 0.0014\n",
            "Epoch 53/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 50.6442 - mean_absolute_error: 1.6389 - mean_squared_logarithmic_error: 0.3384 - cosine_similarity: 0.0095\n",
            "Epoch 54/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 20.4014 - mean_absolute_error: 1.2637 - mean_squared_logarithmic_error: 0.2879 - cosine_similarity: 0.0037\n",
            "Epoch 55/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 9.8379 - mean_absolute_error: 0.9788 - mean_squared_logarithmic_error: 0.2428 - cosine_similarity: 0.0078\n",
            "Epoch 56/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 5.4965 - mean_absolute_error: 0.7776 - mean_squared_logarithmic_error: 0.1938 - cosine_similarity: 0.0065\n",
            "Epoch 57/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 5.1385 - mean_absolute_error: 0.7781 - mean_squared_logarithmic_error: 0.1977 - cosine_similarity: 0.0074\n",
            "Epoch 58/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 4.2849 - mean_absolute_error: 0.6941 - mean_squared_logarithmic_error: 0.1746 - cosine_similarity: 0.0089\n",
            "Epoch 59/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 3.6804 - mean_absolute_error: 0.6745 - mean_squared_logarithmic_error: 0.1720 - cosine_similarity: 0.0090\n",
            "Epoch 60/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 3.5195 - mean_absolute_error: 0.6514 - mean_squared_logarithmic_error: 0.1662 - cosine_similarity: 0.0092\n",
            "Epoch 61/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 3.6041 - mean_absolute_error: 0.6595 - mean_squared_logarithmic_error: 0.1689 - cosine_similarity: 0.0094\n",
            "Epoch 62/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 3.5108 - mean_absolute_error: 0.6578 - mean_squared_logarithmic_error: 0.1692 - cosine_similarity: 0.0104\n",
            "Epoch 63/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 3.4880 - mean_absolute_error: 0.6478 - mean_squared_logarithmic_error: 0.1666 - cosine_similarity: 0.0116\n",
            "Epoch 64/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 3.2129 - mean_absolute_error: 0.6165 - mean_squared_logarithmic_error: 0.1575 - cosine_similarity: 0.0118\n",
            "Epoch 65/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 3.1928 - mean_absolute_error: 0.5961 - mean_squared_logarithmic_error: 0.1513 - cosine_similarity: 0.0097\n",
            "Epoch 66/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 3.0374 - mean_absolute_error: 0.5968 - mean_squared_logarithmic_error: 0.1521 - cosine_similarity: 0.0117\n",
            "Epoch 67/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 3.3896 - mean_absolute_error: 0.6147 - mean_squared_logarithmic_error: 0.1566 - cosine_similarity: 0.0107\n",
            "Epoch 68/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 3.4687 - mean_absolute_error: 0.6298 - mean_squared_logarithmic_error: 0.1611 - cosine_similarity: 0.0137\n",
            "Epoch 69/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 3.8057 - mean_absolute_error: 0.6215 - mean_squared_logarithmic_error: 0.1570 - cosine_similarity: 0.0095\n",
            "Epoch 70/1000\n",
            "27/27 [==============================] - 2s 87ms/step - loss: 4.0510 - mean_absolute_error: 0.6369 - mean_squared_logarithmic_error: 0.1606 - cosine_similarity: 0.0157\n",
            "Epoch 71/1000\n",
            "27/27 [==============================] - 3s 104ms/step - loss: 4.8674 - mean_absolute_error: 0.6516 - mean_squared_logarithmic_error: 0.1622 - cosine_similarity: 0.0079\n",
            "Epoch 72/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 5.7583 - mean_absolute_error: 0.7001 - mean_squared_logarithmic_error: 0.1738 - cosine_similarity: 0.0152\n",
            "Epoch 73/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 7.5602 - mean_absolute_error: 0.7348 - mean_squared_logarithmic_error: 0.1786 - cosine_similarity: 0.0075\n",
            "Epoch 74/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 9.5576 - mean_absolute_error: 0.8142 - mean_squared_logarithmic_error: 0.1953 - cosine_similarity: 0.0156\n",
            "Epoch 75/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 12.5987 - mean_absolute_error: 0.8431 - mean_squared_logarithmic_error: 0.1951 - cosine_similarity: 0.0071\n",
            "Epoch 76/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 16.8819 - mean_absolute_error: 0.9711 - mean_squared_logarithmic_error: 0.2193 - cosine_similarity: 0.0127\n",
            "Epoch 77/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 22.8659 - mean_absolute_error: 1.0653 - mean_squared_logarithmic_error: 0.2326 - cosine_similarity: 0.0085\n",
            "Epoch 78/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 27.9792 - mean_absolute_error: 1.1974 - mean_squared_logarithmic_error: 0.2557 - cosine_similarity: 0.0059\n",
            "Epoch 79/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 31.1955 - mean_absolute_error: 1.2195 - mean_squared_logarithmic_error: 0.2542 - cosine_similarity: 0.0126\n",
            "Epoch 80/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 18.5359 - mean_absolute_error: 1.0212 - mean_squared_logarithmic_error: 0.2263 - cosine_similarity: 0.0065\n",
            "Epoch 81/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 31.3762 - mean_absolute_error: 1.1338 - mean_squared_logarithmic_error: 0.2307 - cosine_similarity: 0.0136\n",
            "Epoch 82/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 16.3538 - mean_absolute_error: 0.9073 - mean_squared_logarithmic_error: 0.2004 - cosine_similarity: 0.0102\n",
            "Epoch 83/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 15.2669 - mean_absolute_error: 0.8486 - mean_squared_logarithmic_error: 0.1898 - cosine_similarity: 0.0149\n",
            "Epoch 84/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 17.4462 - mean_absolute_error: 0.8206 - mean_squared_logarithmic_error: 0.1753 - cosine_similarity: 0.0114\n",
            "Epoch 85/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 9.5158 - mean_absolute_error: 0.7056 - mean_squared_logarithmic_error: 0.1636 - cosine_similarity: 0.0153\n",
            "Epoch 86/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 8.5543 - mean_absolute_error: 0.6663 - mean_squared_logarithmic_error: 0.1555 - cosine_similarity: 0.0130\n",
            "Epoch 87/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 3.8662 - mean_absolute_error: 0.5453 - mean_squared_logarithmic_error: 0.1340 - cosine_similarity: 0.0139\n",
            "Epoch 88/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 2.9370 - mean_absolute_error: 0.4913 - mean_squared_logarithmic_error: 0.1217 - cosine_similarity: 0.0148\n",
            "Epoch 89/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.8033 - mean_absolute_error: 0.4219 - mean_squared_logarithmic_error: 0.1040 - cosine_similarity: 0.0146\n",
            "Epoch 90/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.5346 - mean_absolute_error: 0.3961 - mean_squared_logarithmic_error: 0.0977 - cosine_similarity: 0.0155\n",
            "Epoch 91/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.3495 - mean_absolute_error: 0.3920 - mean_squared_logarithmic_error: 0.0967 - cosine_similarity: 0.0158\n",
            "Epoch 92/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.2391 - mean_absolute_error: 0.3754 - mean_squared_logarithmic_error: 0.0921 - cosine_similarity: 0.0166\n",
            "Epoch 93/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 1.1434 - mean_absolute_error: 0.3689 - mean_squared_logarithmic_error: 0.0903 - cosine_similarity: 0.0162\n",
            "Epoch 94/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.0410 - mean_absolute_error: 0.3444 - mean_squared_logarithmic_error: 0.0834 - cosine_similarity: 0.0183\n",
            "Epoch 95/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 1.0412 - mean_absolute_error: 0.3523 - mean_squared_logarithmic_error: 0.0857 - cosine_similarity: 0.0167\n",
            "Epoch 96/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.9707 - mean_absolute_error: 0.3351 - mean_squared_logarithmic_error: 0.0808 - cosine_similarity: 0.0183\n",
            "Epoch 97/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.9552 - mean_absolute_error: 0.3330 - mean_squared_logarithmic_error: 0.0801 - cosine_similarity: 0.0170\n",
            "Epoch 98/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.9580 - mean_absolute_error: 0.3317 - mean_squared_logarithmic_error: 0.0802 - cosine_similarity: 0.0197\n",
            "Epoch 99/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.8526 - mean_absolute_error: 0.3100 - mean_squared_logarithmic_error: 0.0733 - cosine_similarity: 0.0173\n",
            "Epoch 100/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 0.9597 - mean_absolute_error: 0.3292 - mean_squared_logarithmic_error: 0.0799 - cosine_similarity: 0.0197\n",
            "Epoch 101/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 0.9217 - mean_absolute_error: 0.3106 - mean_squared_logarithmic_error: 0.0740 - cosine_similarity: 0.0173\n",
            "Epoch 102/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.0258 - mean_absolute_error: 0.3257 - mean_squared_logarithmic_error: 0.0792 - cosine_similarity: 0.0202\n",
            "Epoch 103/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 1.0791 - mean_absolute_error: 0.3258 - mean_squared_logarithmic_error: 0.0780 - cosine_similarity: 0.0169\n",
            "Epoch 104/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.2444 - mean_absolute_error: 0.3329 - mean_squared_logarithmic_error: 0.0804 - cosine_similarity: 0.0212\n",
            "Epoch 105/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.3347 - mean_absolute_error: 0.3241 - mean_squared_logarithmic_error: 0.0765 - cosine_similarity: 0.0176\n",
            "Epoch 106/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.9136 - mean_absolute_error: 0.3495 - mean_squared_logarithmic_error: 0.0829 - cosine_similarity: 0.0238\n",
            "Epoch 107/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.7245 - mean_absolute_error: 0.2859 - mean_squared_logarithmic_error: 0.0661 - cosine_similarity: 0.0179\n",
            "Epoch 108/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 2.2819 - mean_absolute_error: 0.3504 - mean_squared_logarithmic_error: 0.0802 - cosine_similarity: 0.0217\n",
            "Epoch 109/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 4.0617 - mean_absolute_error: 0.4179 - mean_squared_logarithmic_error: 0.0938 - cosine_similarity: 0.0177\n",
            "Epoch 110/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 4.1897 - mean_absolute_error: 0.4213 - mean_squared_logarithmic_error: 0.0958 - cosine_similarity: 0.0212\n",
            "Epoch 111/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 6.9699 - mean_absolute_error: 0.5020 - mean_squared_logarithmic_error: 0.1110 - cosine_similarity: 0.0185\n",
            "Epoch 112/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 5.2799 - mean_absolute_error: 0.4851 - mean_squared_logarithmic_error: 0.1135 - cosine_similarity: 0.0215\n",
            "Epoch 113/1000\n",
            "27/27 [==============================] - 2s 78ms/step - loss: 7.0437 - mean_absolute_error: 0.5135 - mean_squared_logarithmic_error: 0.1170 - cosine_similarity: 0.0201\n",
            "Epoch 114/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 3.9134 - mean_absolute_error: 0.4837 - mean_squared_logarithmic_error: 0.1182 - cosine_similarity: 0.0211\n",
            "Epoch 115/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 4.2845 - mean_absolute_error: 0.5122 - mean_squared_logarithmic_error: 0.1241 - cosine_similarity: 0.0142\n",
            "Epoch 116/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 4.0132 - mean_absolute_error: 0.5291 - mean_squared_logarithmic_error: 0.1315 - cosine_similarity: 0.0235\n",
            "Epoch 117/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 4.9375 - mean_absolute_error: 0.5746 - mean_squared_logarithmic_error: 0.1370 - cosine_similarity: 0.0076\n",
            "Epoch 118/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 5.4193 - mean_absolute_error: 0.5576 - mean_squared_logarithmic_error: 0.1313 - cosine_similarity: 0.0279\n",
            "Epoch 119/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 6.9321 - mean_absolute_error: 0.5892 - mean_squared_logarithmic_error: 0.1327 - cosine_similarity: 0.0028\n",
            "Epoch 120/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 10.2146 - mean_absolute_error: 0.6830 - mean_squared_logarithmic_error: 0.1511 - cosine_similarity: 0.0237\n",
            "Epoch 121/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 13.6994 - mean_absolute_error: 0.7622 - mean_squared_logarithmic_error: 0.1627 - cosine_similarity: 0.0050\n",
            "Epoch 122/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 16.7995 - mean_absolute_error: 0.8140 - mean_squared_logarithmic_error: 0.1695 - cosine_similarity: 0.0186\n",
            "Epoch 123/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 15.7221 - mean_absolute_error: 0.8031 - mean_squared_logarithmic_error: 0.1686 - cosine_similarity: 0.0053\n",
            "Epoch 124/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 11.7878 - mean_absolute_error: 0.7032 - mean_squared_logarithmic_error: 0.1491 - cosine_similarity: 0.0166\n",
            "Epoch 125/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 6.9679 - mean_absolute_error: 0.5898 - mean_squared_logarithmic_error: 0.1348 - cosine_similarity: 0.0129\n",
            "Epoch 126/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 3.7974 - mean_absolute_error: 0.4843 - mean_squared_logarithmic_error: 0.1127 - cosine_similarity: 0.0109\n",
            "Epoch 127/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 2.2467 - mean_absolute_error: 0.3920 - mean_squared_logarithmic_error: 0.0976 - cosine_similarity: 0.0278\n",
            "Epoch 128/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 1.4751 - mean_absolute_error: 0.3412 - mean_squared_logarithmic_error: 0.0813 - cosine_similarity: 0.0113\n",
            "Epoch 129/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.2449 - mean_absolute_error: 0.2838 - mean_squared_logarithmic_error: 0.0689 - cosine_similarity: 0.0327\n",
            "Epoch 130/1000\n",
            "27/27 [==============================] - 2s 78ms/step - loss: 1.0616 - mean_absolute_error: 0.2735 - mean_squared_logarithmic_error: 0.0619 - cosine_similarity: 0.0185\n",
            "Epoch 131/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.9704 - mean_absolute_error: 0.2582 - mean_squared_logarithmic_error: 0.0596 - cosine_similarity: 0.0268\n",
            "Epoch 132/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.7117 - mean_absolute_error: 0.2410 - mean_squared_logarithmic_error: 0.0522 - cosine_similarity: 0.0218\n",
            "Epoch 133/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.1284 - mean_absolute_error: 0.2710 - mean_squared_logarithmic_error: 0.0616 - cosine_similarity: 0.0235\n",
            "Epoch 134/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 1.3143 - mean_absolute_error: 0.2830 - mean_squared_logarithmic_error: 0.0628 - cosine_similarity: 0.0232\n",
            "Epoch 135/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.2392 - mean_absolute_error: 0.2881 - mean_squared_logarithmic_error: 0.0648 - cosine_similarity: 0.0211\n",
            "Epoch 136/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.4986 - mean_absolute_error: 0.2992 - mean_squared_logarithmic_error: 0.0666 - cosine_similarity: 0.0253\n",
            "Epoch 137/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.4045 - mean_absolute_error: 0.2998 - mean_squared_logarithmic_error: 0.0668 - cosine_similarity: 0.0159\n",
            "Epoch 138/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.7255 - mean_absolute_error: 0.2990 - mean_squared_logarithmic_error: 0.0653 - cosine_similarity: 0.0274\n",
            "Epoch 139/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.5304 - mean_absolute_error: 0.3033 - mean_squared_logarithmic_error: 0.0670 - cosine_similarity: 0.0156\n",
            "Epoch 140/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 1.8296 - mean_absolute_error: 0.3058 - mean_squared_logarithmic_error: 0.0670 - cosine_similarity: 0.0267\n",
            "Epoch 141/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.4723 - mean_absolute_error: 0.3023 - mean_squared_logarithmic_error: 0.0667 - cosine_similarity: 0.0159\n",
            "Epoch 142/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 1.7397 - mean_absolute_error: 0.3041 - mean_squared_logarithmic_error: 0.0673 - cosine_similarity: 0.0301\n",
            "Epoch 143/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.2743 - mean_absolute_error: 0.2812 - mean_squared_logarithmic_error: 0.0618 - cosine_similarity: 0.0138\n",
            "Epoch 144/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 1.3676 - mean_absolute_error: 0.2822 - mean_squared_logarithmic_error: 0.0632 - cosine_similarity: 0.0322\n",
            "Epoch 145/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.0536 - mean_absolute_error: 0.2677 - mean_squared_logarithmic_error: 0.0595 - cosine_similarity: 0.0176\n",
            "Epoch 146/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 1.1139 - mean_absolute_error: 0.2577 - mean_squared_logarithmic_error: 0.0576 - cosine_similarity: 0.0331\n",
            "Epoch 147/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.9054 - mean_absolute_error: 0.2548 - mean_squared_logarithmic_error: 0.0566 - cosine_similarity: 0.0180\n",
            "Epoch 148/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.9104 - mean_absolute_error: 0.2333 - mean_squared_logarithmic_error: 0.0515 - cosine_similarity: 0.0343\n",
            "Epoch 149/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.7566 - mean_absolute_error: 0.2394 - mean_squared_logarithmic_error: 0.0530 - cosine_similarity: 0.0193\n",
            "Epoch 150/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.8006 - mean_absolute_error: 0.2235 - mean_squared_logarithmic_error: 0.0503 - cosine_similarity: 0.0363\n",
            "Epoch 151/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.6956 - mean_absolute_error: 0.2248 - mean_squared_logarithmic_error: 0.0495 - cosine_similarity: 0.0214\n",
            "Epoch 152/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.7728 - mean_absolute_error: 0.2227 - mean_squared_logarithmic_error: 0.0500 - cosine_similarity: 0.0325\n",
            "Epoch 153/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.6901 - mean_absolute_error: 0.2237 - mean_squared_logarithmic_error: 0.0495 - cosine_similarity: 0.0253\n",
            "Epoch 154/1000\n",
            "27/27 [==============================] - 3s 104ms/step - loss: 0.7541 - mean_absolute_error: 0.2131 - mean_squared_logarithmic_error: 0.0483 - cosine_similarity: 0.0340\n",
            "Epoch 155/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.7423 - mean_absolute_error: 0.2270 - mean_squared_logarithmic_error: 0.0513 - cosine_similarity: 0.0242\n",
            "Epoch 156/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.3833 - mean_absolute_error: 0.1862 - mean_squared_logarithmic_error: 0.0412 - cosine_similarity: 0.0387\n",
            "Epoch 157/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.9721 - mean_absolute_error: 0.2527 - mean_squared_logarithmic_error: 0.0582 - cosine_similarity: 0.0174\n",
            "Epoch 158/1000\n",
            "27/27 [==============================] - 2s 79ms/step - loss: 1.0759 - mean_absolute_error: 0.2732 - mean_squared_logarithmic_error: 0.0646 - cosine_similarity: 0.0380\n",
            "Epoch 159/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.4315 - mean_absolute_error: 0.3102 - mean_squared_logarithmic_error: 0.0710 - cosine_similarity: 0.0093\n",
            "Epoch 160/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.4705 - mean_absolute_error: 0.3328 - mean_squared_logarithmic_error: 0.0781 - cosine_similarity: 0.0289\n",
            "Epoch 161/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.9387 - mean_absolute_error: 0.3587 - mean_squared_logarithmic_error: 0.0809 - cosine_similarity: 0.0127\n",
            "Epoch 162/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 2.4903 - mean_absolute_error: 0.3872 - mean_squared_logarithmic_error: 0.0909 - cosine_similarity: 0.0136\n",
            "Epoch 163/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 4.7666 - mean_absolute_error: 0.5009 - mean_squared_logarithmic_error: 0.1134 - cosine_similarity: 0.0193\n",
            "Epoch 164/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 7.8596 - mean_absolute_error: 0.6100 - mean_squared_logarithmic_error: 0.1252 - cosine_similarity: 0.0030\n",
            "Epoch 165/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 9.6755 - mean_absolute_error: 0.7445 - mean_squared_logarithmic_error: 0.1542 - cosine_similarity: 0.0146\n",
            "Epoch 166/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 12.7422 - mean_absolute_error: 0.8322 - mean_squared_logarithmic_error: 0.1756 - cosine_similarity: 0.0057\n",
            "Epoch 167/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 28.6013 - mean_absolute_error: 0.9298 - mean_squared_logarithmic_error: 0.1689 - cosine_similarity: -0.0092\n",
            "Epoch 168/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 24.1984 - mean_absolute_error: 0.8767 - mean_squared_logarithmic_error: 0.1679 - cosine_similarity: 0.0268\n",
            "Epoch 169/1000\n",
            "27/27 [==============================] - 2s 78ms/step - loss: 9.9781 - mean_absolute_error: 0.7247 - mean_squared_logarithmic_error: 0.1475 - cosine_similarity: 3.3003e-04\n",
            "Epoch 170/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 5.3055 - mean_absolute_error: 0.5338 - mean_squared_logarithmic_error: 0.1069 - cosine_similarity: 0.0113\n",
            "Epoch 171/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.2130 - mean_absolute_error: 0.3042 - mean_squared_logarithmic_error: 0.0623 - cosine_similarity: 0.0101\n",
            "Epoch 172/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.4345 - mean_absolute_error: 0.1998 - mean_squared_logarithmic_error: 0.0407 - cosine_similarity: 0.0244\n",
            "Epoch 173/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2192 - mean_absolute_error: 0.1441 - mean_squared_logarithmic_error: 0.0280 - cosine_similarity: 0.0349\n",
            "Epoch 174/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.1890 - mean_absolute_error: 0.1332 - mean_squared_logarithmic_error: 0.0261 - cosine_similarity: 0.0417\n",
            "Epoch 175/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1762 - mean_absolute_error: 0.1210 - mean_squared_logarithmic_error: 0.0239 - cosine_similarity: 0.0529\n",
            "Epoch 176/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1741 - mean_absolute_error: 0.1231 - mean_squared_logarithmic_error: 0.0237 - cosine_similarity: 0.0417\n",
            "Epoch 177/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1658 - mean_absolute_error: 0.1145 - mean_squared_logarithmic_error: 0.0222 - cosine_similarity: 0.0557\n",
            "Epoch 178/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.1628 - mean_absolute_error: 0.1166 - mean_squared_logarithmic_error: 0.0220 - cosine_similarity: 0.0448\n",
            "Epoch 179/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.1496 - mean_absolute_error: 0.1062 - mean_squared_logarithmic_error: 0.0200 - cosine_similarity: 0.0573\n",
            "Epoch 180/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.1169 - mean_absolute_error: 0.1060 - mean_squared_logarithmic_error: 0.0187 - cosine_similarity: 0.0470\n",
            "Epoch 181/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.1538 - mean_absolute_error: 0.1074 - mean_squared_logarithmic_error: 0.0200 - cosine_similarity: 0.0548\n",
            "Epoch 182/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.1810 - mean_absolute_error: 0.1179 - mean_squared_logarithmic_error: 0.0223 - cosine_similarity: 0.0454\n",
            "Epoch 183/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1813 - mean_absolute_error: 0.1146 - mean_squared_logarithmic_error: 0.0216 - cosine_similarity: 0.0529\n",
            "Epoch 184/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2015 - mean_absolute_error: 0.1213 - mean_squared_logarithmic_error: 0.0230 - cosine_similarity: 0.0452\n",
            "Epoch 185/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2041 - mean_absolute_error: 0.1214 - mean_squared_logarithmic_error: 0.0231 - cosine_similarity: 0.0489\n",
            "Epoch 186/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.2251 - mean_absolute_error: 0.1253 - mean_squared_logarithmic_error: 0.0239 - cosine_similarity: 0.0428\n",
            "Epoch 187/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2358 - mean_absolute_error: 0.1220 - mean_squared_logarithmic_error: 0.0229 - cosine_similarity: 0.0456\n",
            "Epoch 188/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2551 - mean_absolute_error: 0.1266 - mean_squared_logarithmic_error: 0.0241 - cosine_similarity: 0.0414\n",
            "Epoch 189/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2394 - mean_absolute_error: 0.1249 - mean_squared_logarithmic_error: 0.0235 - cosine_similarity: 0.0453\n",
            "Epoch 190/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.2233 - mean_absolute_error: 0.1225 - mean_squared_logarithmic_error: 0.0232 - cosine_similarity: 0.0445\n",
            "Epoch 191/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.2000 - mean_absolute_error: 0.1213 - mean_squared_logarithmic_error: 0.0226 - cosine_similarity: 0.0475\n",
            "Epoch 192/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.1849 - mean_absolute_error: 0.1100 - mean_squared_logarithmic_error: 0.0202 - cosine_similarity: 0.0467\n",
            "Epoch 193/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.1799 - mean_absolute_error: 0.1164 - mean_squared_logarithmic_error: 0.0215 - cosine_similarity: 0.0516\n",
            "Epoch 194/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.1605 - mean_absolute_error: 0.1072 - mean_squared_logarithmic_error: 0.0200 - cosine_similarity: 0.0533\n",
            "Epoch 195/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1384 - mean_absolute_error: 0.1048 - mean_squared_logarithmic_error: 0.0191 - cosine_similarity: 0.0594\n",
            "Epoch 196/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.1393 - mean_absolute_error: 0.1019 - mean_squared_logarithmic_error: 0.0194 - cosine_similarity: 0.0666\n",
            "Epoch 197/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1364 - mean_absolute_error: 0.0983 - mean_squared_logarithmic_error: 0.0181 - cosine_similarity: 0.0676\n",
            "Epoch 198/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1721 - mean_absolute_error: 0.1035 - mean_squared_logarithmic_error: 0.0205 - cosine_similarity: 0.0746\n",
            "Epoch 199/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2377 - mean_absolute_error: 0.1110 - mean_squared_logarithmic_error: 0.0220 - cosine_similarity: 0.0657\n",
            "Epoch 200/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.3362 - mean_absolute_error: 0.1192 - mean_squared_logarithmic_error: 0.0238 - cosine_similarity: 0.0626\n",
            "Epoch 201/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.6440 - mean_absolute_error: 0.1428 - mean_squared_logarithmic_error: 0.0286 - cosine_similarity: 0.0475\n",
            "Epoch 202/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.7547 - mean_absolute_error: 0.1646 - mean_squared_logarithmic_error: 0.0312 - cosine_similarity: 0.0381\n",
            "Epoch 203/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.3197 - mean_absolute_error: 0.1888 - mean_squared_logarithmic_error: 0.0372 - cosine_similarity: 0.0285\n",
            "Epoch 204/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.6103 - mean_absolute_error: 0.1973 - mean_squared_logarithmic_error: 0.0372 - cosine_similarity: 0.0284\n",
            "Epoch 205/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2119 - mean_absolute_error: 0.1423 - mean_squared_logarithmic_error: 0.0262 - cosine_similarity: 0.0232\n",
            "Epoch 206/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.6981 - mean_absolute_error: 0.1948 - mean_squared_logarithmic_error: 0.0402 - cosine_similarity: 0.0351\n",
            "Epoch 207/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.4453 - mean_absolute_error: 0.1827 - mean_squared_logarithmic_error: 0.0380 - cosine_similarity: 0.0198\n",
            "Epoch 208/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.4055 - mean_absolute_error: 0.1716 - mean_squared_logarithmic_error: 0.0357 - cosine_similarity: 0.0436\n",
            "Epoch 209/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.3891 - mean_absolute_error: 0.1811 - mean_squared_logarithmic_error: 0.0381 - cosine_similarity: 0.0201\n",
            "Epoch 210/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.3920 - mean_absolute_error: 0.1775 - mean_squared_logarithmic_error: 0.0376 - cosine_similarity: 0.0491\n",
            "Epoch 211/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.3945 - mean_absolute_error: 0.1822 - mean_squared_logarithmic_error: 0.0377 - cosine_similarity: 0.0277\n",
            "Epoch 212/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.3888 - mean_absolute_error: 0.1812 - mean_squared_logarithmic_error: 0.0365 - cosine_similarity: 0.0380\n",
            "Epoch 213/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.4067 - mean_absolute_error: 0.1875 - mean_squared_logarithmic_error: 0.0376 - cosine_similarity: 0.0289\n",
            "Epoch 214/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.4456 - mean_absolute_error: 0.1940 - mean_squared_logarithmic_error: 0.0384 - cosine_similarity: 0.0299\n",
            "Epoch 215/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.5217 - mean_absolute_error: 0.2086 - mean_squared_logarithmic_error: 0.0413 - cosine_similarity: 0.0339\n",
            "Epoch 216/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.4968 - mean_absolute_error: 0.2037 - mean_squared_logarithmic_error: 0.0374 - cosine_similarity: 0.0132\n",
            "Epoch 217/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.5303 - mean_absolute_error: 0.2138 - mean_squared_logarithmic_error: 0.0389 - cosine_similarity: 0.0394\n",
            "Epoch 218/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.6662 - mean_absolute_error: 0.2404 - mean_squared_logarithmic_error: 0.0459 - cosine_similarity: 0.0025\n",
            "Epoch 219/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.7186 - mean_absolute_error: 0.2460 - mean_squared_logarithmic_error: 0.0472 - cosine_similarity: 0.0348\n",
            "Epoch 220/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.8092 - mean_absolute_error: 0.2637 - mean_squared_logarithmic_error: 0.0515 - cosine_similarity: -0.0024\n",
            "Epoch 221/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.0054 - mean_absolute_error: 0.2949 - mean_squared_logarithmic_error: 0.0590 - cosine_similarity: 0.0303\n",
            "Epoch 222/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.1240 - mean_absolute_error: 0.3098 - mean_squared_logarithmic_error: 0.0601 - cosine_similarity: -0.0049\n",
            "Epoch 223/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.4426 - mean_absolute_error: 0.3539 - mean_squared_logarithmic_error: 0.0718 - cosine_similarity: 0.0227\n",
            "Epoch 224/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.9734 - mean_absolute_error: 0.4006 - mean_squared_logarithmic_error: 0.0807 - cosine_similarity: 0.0028\n",
            "Epoch 225/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 1.8796 - mean_absolute_error: 0.4072 - mean_squared_logarithmic_error: 0.0843 - cosine_similarity: 0.0085\n",
            "Epoch 226/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.7587 - mean_absolute_error: 0.3906 - mean_squared_logarithmic_error: 0.0779 - cosine_similarity: 0.0132\n",
            "Epoch 227/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.8576 - mean_absolute_error: 0.2969 - mean_squared_logarithmic_error: 0.0583 - cosine_similarity: 0.0095\n",
            "Epoch 228/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.4466 - mean_absolute_error: 0.2259 - mean_squared_logarithmic_error: 0.0413 - cosine_similarity: 0.0149\n",
            "Epoch 229/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.6440 - mean_absolute_error: 0.2505 - mean_squared_logarithmic_error: 0.0474 - cosine_similarity: 0.0246\n",
            "Epoch 230/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.6535 - mean_absolute_error: 0.2548 - mean_squared_logarithmic_error: 0.0476 - cosine_similarity: 0.0059\n",
            "Epoch 231/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.9624 - mean_absolute_error: 0.2945 - mean_squared_logarithmic_error: 0.0553 - cosine_similarity: 0.0194\n",
            "Epoch 232/1000\n",
            "27/27 [==============================] - 3s 95ms/step - loss: 1.5073 - mean_absolute_error: 0.3558 - mean_squared_logarithmic_error: 0.0694 - cosine_similarity: 0.0105\n",
            "Epoch 233/1000\n",
            "27/27 [==============================] - 3s 101ms/step - loss: 2.0128 - mean_absolute_error: 0.4052 - mean_squared_logarithmic_error: 0.0789 - cosine_similarity: 0.0097\n",
            "Epoch 234/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 2.3582 - mean_absolute_error: 0.4432 - mean_squared_logarithmic_error: 0.0876 - cosine_similarity: 0.0193\n",
            "Epoch 235/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 2.1652 - mean_absolute_error: 0.4228 - mean_squared_logarithmic_error: 0.0810 - cosine_similarity: 1.3675e-04\n",
            "Epoch 236/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 1.9382 - mean_absolute_error: 0.3931 - mean_squared_logarithmic_error: 0.0749 - cosine_similarity: 0.0235\n",
            "Epoch 237/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.3004 - mean_absolute_error: 0.3386 - mean_squared_logarithmic_error: 0.0625 - cosine_similarity: 2.9162e-04\n",
            "Epoch 238/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.8167 - mean_absolute_error: 0.2803 - mean_squared_logarithmic_error: 0.0505 - cosine_similarity: 0.0230\n",
            "Epoch 239/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.3923 - mean_absolute_error: 0.2165 - mean_squared_logarithmic_error: 0.0366 - cosine_similarity: 0.0071\n",
            "Epoch 240/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2116 - mean_absolute_error: 0.1626 - mean_squared_logarithmic_error: 0.0263 - cosine_similarity: 0.0396\n",
            "Epoch 241/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.0948 - mean_absolute_error: 0.1137 - mean_squared_logarithmic_error: 0.0153 - cosine_similarity: 0.0139\n",
            "Epoch 242/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1279 - mean_absolute_error: 0.1118 - mean_squared_logarithmic_error: 0.0181 - cosine_similarity: 0.0594\n",
            "Epoch 243/1000\n",
            "27/27 [==============================] - 2s 85ms/step - loss: 0.0837 - mean_absolute_error: 0.0961 - mean_squared_logarithmic_error: 0.0142 - cosine_similarity: 0.0274\n",
            "Epoch 244/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.0743 - mean_absolute_error: 0.0850 - mean_squared_logarithmic_error: 0.0124 - cosine_similarity: 0.0648\n",
            "Epoch 245/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.0712 - mean_absolute_error: 0.0824 - mean_squared_logarithmic_error: 0.0119 - cosine_similarity: 0.0486\n",
            "Epoch 246/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.0703 - mean_absolute_error: 0.0751 - mean_squared_logarithmic_error: 0.0106 - cosine_similarity: 0.0683\n",
            "Epoch 247/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.0813 - mean_absolute_error: 0.0805 - mean_squared_logarithmic_error: 0.0116 - cosine_similarity: 0.0574\n",
            "Epoch 248/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.0947 - mean_absolute_error: 0.0809 - mean_squared_logarithmic_error: 0.0121 - cosine_similarity: 0.0609\n",
            "Epoch 249/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1082 - mean_absolute_error: 0.0844 - mean_squared_logarithmic_error: 0.0126 - cosine_similarity: 0.0621\n",
            "Epoch 250/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.1343 - mean_absolute_error: 0.0899 - mean_squared_logarithmic_error: 0.0139 - cosine_similarity: 0.0474\n",
            "Epoch 251/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.1413 - mean_absolute_error: 0.0917 - mean_squared_logarithmic_error: 0.0137 - cosine_similarity: 0.0558\n",
            "Epoch 252/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.1797 - mean_absolute_error: 0.0986 - mean_squared_logarithmic_error: 0.0155 - cosine_similarity: 0.0384\n",
            "Epoch 253/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.1815 - mean_absolute_error: 0.1032 - mean_squared_logarithmic_error: 0.0161 - cosine_similarity: 0.0471\n",
            "Epoch 254/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.1354 - mean_absolute_error: 0.0906 - mean_squared_logarithmic_error: 0.0122 - cosine_similarity: 0.0349\n",
            "Epoch 255/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1940 - mean_absolute_error: 0.1108 - mean_squared_logarithmic_error: 0.0180 - cosine_similarity: 0.0423\n",
            "Epoch 256/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2008 - mean_absolute_error: 0.1141 - mean_squared_logarithmic_error: 0.0188 - cosine_similarity: 0.0323\n",
            "Epoch 257/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.1727 - mean_absolute_error: 0.1106 - mean_squared_logarithmic_error: 0.0173 - cosine_similarity: 0.0430\n",
            "Epoch 258/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2307 - mean_absolute_error: 0.1272 - mean_squared_logarithmic_error: 0.0223 - cosine_similarity: 0.0301\n",
            "Epoch 259/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2785 - mean_absolute_error: 0.1342 - mean_squared_logarithmic_error: 0.0247 - cosine_similarity: 0.0365\n",
            "Epoch 260/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2566 - mean_absolute_error: 0.1401 - mean_squared_logarithmic_error: 0.0252 - cosine_similarity: 0.0250\n",
            "Epoch 261/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.3003 - mean_absolute_error: 0.1447 - mean_squared_logarithmic_error: 0.0264 - cosine_similarity: 0.0311\n",
            "Epoch 262/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2488 - mean_absolute_error: 0.1454 - mean_squared_logarithmic_error: 0.0259 - cosine_similarity: 0.0216\n",
            "Epoch 263/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.2771 - mean_absolute_error: 0.1495 - mean_squared_logarithmic_error: 0.0276 - cosine_similarity: 0.0342\n",
            "Epoch 264/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2672 - mean_absolute_error: 0.1537 - mean_squared_logarithmic_error: 0.0285 - cosine_similarity: 0.0254\n",
            "Epoch 265/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2995 - mean_absolute_error: 0.1517 - mean_squared_logarithmic_error: 0.0284 - cosine_similarity: 0.0300\n",
            "Epoch 266/1000\n",
            "27/27 [==============================] - 2s 84ms/step - loss: 0.2621 - mean_absolute_error: 0.1458 - mean_squared_logarithmic_error: 0.0251 - cosine_similarity: 0.0454\n",
            "Epoch 267/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.4616 - mean_absolute_error: 0.1805 - mean_squared_logarithmic_error: 0.0347 - cosine_similarity: 0.0125\n",
            "Epoch 268/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.5920 - mean_absolute_error: 0.1947 - mean_squared_logarithmic_error: 0.0374 - cosine_similarity: 0.0418\n",
            "Epoch 269/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.7801 - mean_absolute_error: 0.2203 - mean_squared_logarithmic_error: 0.0428 - cosine_similarity: 0.0030\n",
            "Epoch 270/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.9989 - mean_absolute_error: 0.2489 - mean_squared_logarithmic_error: 0.0482 - cosine_similarity: 0.0355\n",
            "Epoch 271/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 1.0811 - mean_absolute_error: 0.2601 - mean_squared_logarithmic_error: 0.0490 - cosine_similarity: -0.0020\n",
            "Epoch 272/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 1.0594 - mean_absolute_error: 0.2674 - mean_squared_logarithmic_error: 0.0511 - cosine_similarity: 0.0332\n",
            "Epoch 273/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.8911 - mean_absolute_error: 0.2579 - mean_squared_logarithmic_error: 0.0484 - cosine_similarity: 0.0106\n",
            "Epoch 274/1000\n",
            "27/27 [==============================] - 2s 80ms/step - loss: 0.7144 - mean_absolute_error: 0.2346 - mean_squared_logarithmic_error: 0.0425 - cosine_similarity: 0.0219\n",
            "Epoch 275/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.5803 - mean_absolute_error: 0.2136 - mean_squared_logarithmic_error: 0.0375 - cosine_similarity: 0.0205\n",
            "Epoch 276/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.3931 - mean_absolute_error: 0.1819 - mean_squared_logarithmic_error: 0.0302 - cosine_similarity: 0.0186\n",
            "Epoch 277/1000\n",
            "27/27 [==============================] - 2s 82ms/step - loss: 0.2076 - mean_absolute_error: 0.1322 - mean_squared_logarithmic_error: 0.0199 - cosine_similarity: 0.0266\n",
            "Epoch 278/1000\n",
            "27/27 [==============================] - 2s 81ms/step - loss: 0.1168 - mean_absolute_error: 0.1021 - mean_squared_logarithmic_error: 0.0151 - cosine_similarity: 0.0335\n",
            "Epoch 279/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.0617 - mean_absolute_error: 0.0721 - mean_squared_logarithmic_error: 0.0097 - cosine_similarity: 0.0507\n",
            "Epoch 280/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.0450 - mean_absolute_error: 0.0618 - mean_squared_logarithmic_error: 0.0081 - cosine_similarity: 0.0654\n",
            "Epoch 281/1000\n",
            "27/27 [==============================] - 2s 83ms/step - loss: 0.0316 - mean_absolute_error: 0.0521 - mean_squared_logarithmic_error: 0.0062 - cosine_similarity: 0.0851\n",
            "Epoch 282/1000\n",
            "16/27 [================>.............] - ETA: 0s - loss: 0.0308 - mean_absolute_error: 0.0535 - mean_squared_logarithmic_error: 0.0064 - cosine_similarity: 0.0873"
          ]
        }
      ],
      "source": [
        "filepath_old = '/content/drive/MyDrive/Deep Learning/Music Generator/DAE_best_model_3.hdf5'\n",
        "\n",
        "with tf.device('/GPU:0'):\n",
        "  # DAE_model = load_model(filepath_old)\n",
        "  DAE_history = DAE_model.fit(DAE_data_generator1(songs, batchsize), steps_per_epoch=estimateNumberOfSamples(songs, 0, 5)/batchsize, epochs=1000, callbacks=[checkpoint]).history\n",
        "  DAE_model = load_model(filepath)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "aV3853-11djG"
      },
      "outputs": [],
      "source": [
        "DAE_model = load_model(filepath)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nm_WiatfZuff",
        "outputId": "6793275a-637d-448b-cf8f-597285a29d34"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "X_test=  [-0.00090682 -0.02104415  0.02712481 ...  0.00014189 -0.00020377\n",
            " -0.00030586]\n",
            "y_test_true=  [ 0.00558981 -0.01113711  0.02105595 ...  0.00012396  0.00011809\n",
            "  0.00010467]\n",
            "y_test_pred=  [-0.04421364  0.20197463 -0.2910446  ... -0.11576898 -0.0268769\n",
            " -0.27270103]\n",
            "differention=  [-0.04980345  0.21311174 -0.31210056 ... -0.11589294 -0.02699499\n",
            " -0.2728057 ]\n"
          ]
        }
      ],
      "source": [
        "X_test = songs[0]\n",
        "pieces = 3\n",
        "song_pieces = splitToSpecificLength(X_test, pieces)\n",
        "X_test = song_pieces[15]\n",
        "X_test_wav = appendNotes(X_test, notes, notesLabel)\n",
        "X_test_pad = addPadding(X_test_wav)\n",
        "X_test = AddNoiseToPiece(X_test_pad, noises[16], 0.25)\n",
        "#X_test = X_test_pad\n",
        "wavfile.write('input.wav', sample_rate, X_test)\n",
        "\n",
        "y_test_true = X_test_pad\n",
        "X_test = np.array([X_test])\n",
        "y_test_pred = DAE_model.predict(X_test)\n",
        "\n",
        "\n",
        "print('X_test= ', X_test[0])\n",
        "print('y_test_true= ', y_test_true)\n",
        "print('y_test_pred= ', y_test_pred[0])\n",
        "\n",
        "wavfile.write('output_true.wav', sample_rate, y_test_true)\n",
        "wavfile.write('output_pred.wav', sample_rate, y_test_pred[0])\n",
        "\n",
        "\n",
        "differention = np.subtract(y_test_pred[0], y_test_true)\n",
        "print('differention= ', differention)\n",
        "\n",
        "# y_pred = DAE_model.predict(X_test)\n",
        "# wavfile.write('pred.wav', 4200, y_pred[0])"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jHKA9UHazJkI"
      },
      "source": [
        "# **Note Recognition**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vIqIjSi41bS8"
      },
      "source": [
        "## **Preprocessing**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "79U9x_Su2BnO"
      },
      "source": [
        "**Read Note name from Piano Triads folder**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Oumdc-C7zF2e"
      },
      "outputs": [],
      "source": [
        "notesPath = '/content/drive/MyDrive/Deep Learning/Music Generator/piano_triads'"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wHF-t_-v6qJg",
        "outputId": "e473683b-6efa-4b7c-b2ad-74ac05cfaf97"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((432, 12000), (432,))"
            ]
          },
          "metadata": {},
          "execution_count": 79
        }
      ],
      "source": [
        "note = []\n",
        "noteLabel = []\n",
        "noteDic = {}\n",
        "sample_rate = 4000\n",
        "i = 0\n",
        "for filename in os.listdir(notesPath):\n",
        "  # data = sps.resample(wavfile.read(path+'/'+filename)[1], note_sample_rate)\n",
        "  data = librosa.load(notesPath + '/'+ filename, sr=sample_rate, dtype=np.float32)[0]\n",
        "  note.append(data)\n",
        "  noteLabel.append(filename[:-4])\n",
        "  noteDic[filename[:-4]] = i\n",
        "  i += 1\n",
        "\n",
        "note = np.array(note)\n",
        "noteLabel = np.array(noteLabel)\n",
        "\n",
        "note.shape, noteLabel.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "dTTfTvp-pOvi",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7bed8aca-0571-4d28-9e57-78ccb77b0229"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(432, 432)"
            ]
          },
          "metadata": {},
          "execution_count": 80
        }
      ],
      "source": [
        "# create y true for each note\n",
        "# A_dim_2_0\n",
        "y_true_NR = []\n",
        "\n",
        "for label in noteLabel:\n",
        "  labelEncoded = noteDic[label]\n",
        "  labelEncoded_HOT = to_categorical(labelEncoded, num_classes=len(noteDic))\n",
        "  y_true_NR.append(labelEncoded_HOT)\n",
        "\n",
        "y_true_NR = np.array(y_true_NR)\n",
        "\n",
        "X_NR = note\n",
        "\n",
        "y_true_NR.shape"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "plg3bsOfjJYn"
      },
      "source": [
        "### **Data Augmentation**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WuHsZrgdmTgO"
      },
      "source": [
        "#### **Noise Injection**"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**medium noise**"
      ],
      "metadata": {
        "id": "utHF2z15vUua"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mrkgcZtdixXO",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "96f1c46c-df75-43b3-f19a-7895ec2c81be"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((864, 12000), (864, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 81
        }
      ],
      "source": [
        "y_true_NR_aug = y_true_NR\n",
        "for i in range(len(note)):\n",
        "  noise = np.random.randn(len(note[i]))*0.01\n",
        "  X_NR = np.append(X_NR, [noise+note[i]], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**high noise**"
      ],
      "metadata": {
        "id": "I2DCCZKPvrMW"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "for i in range(len(note)):\n",
        "  noise = np.random.randn(len(note[i]))*0.02\n",
        "  X_NR = np.append(X_NR, [noise+note[i]], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6S2jgcYCrX7b",
        "outputId": "6a88811d-1307-465d-de36-cef5457bb3bd"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((1296, 12000), (1296, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 82
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "for i in range(len(note)):\n",
        "  noise = np.random.randn(len(note[i]))*0.015\n",
        "  X_NR = np.append(X_NR, [noise+note[i]], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cEY4tfwxwsmf",
        "outputId": "3863faaa-8164-4079-a833-fc29dea2cca5"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((1728, 12000), (1728, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 83
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**low noise**"
      ],
      "metadata": {
        "id": "efjar5b4vvmE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "for i in range(len(note)):\n",
        "  noise = np.random.randn(len(note[i]))*0.005\n",
        "  X_NR = np.append(X_NR, [noise+note[i]], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iobNAp6Krzap",
        "outputId": "a8d35952-17d0-485f-bb7c-c9992be6aec7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((2160, 12000), (2160, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 84
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NmUGeK9MoHyt"
      },
      "source": [
        "####**Changing Pitch**"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**low pitch shift**"
      ],
      "metadata": {
        "id": "4DBDyJtpvysG"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "MMsP-nyyoHN6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a130d385-49cf-421b-b848-64d9927ae032"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((2592, 12000), (2592, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 85
        }
      ],
      "source": [
        "for i in range(len(note)):\n",
        "  data = librosa.effects.pitch_shift(note[i], sample_rate, n_steps=12)\n",
        "  X_NR = np.append(X_NR, [data], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**high pitch shift**"
      ],
      "metadata": {
        "id": "WuWm7k-8v2RS"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "for i in range(len(note)):\n",
        "  data = librosa.effects.pitch_shift(note[i], sample_rate, n_steps=20)\n",
        "  X_NR = np.append(X_NR, [data], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "a4rSdZC5sKS4",
        "outputId": "25dd104d-468a-4b81-90e9-627ae6e2e0bf"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((3024, 12000), (3024, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 86
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1t344qwo0F-o"
      },
      "source": [
        "#### **Change Volume**"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**increase volume**"
      ],
      "metadata": {
        "id": "aaJUGZZ0v6FE"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "d-0cva4i0Kcq",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4f51a986-4e7b-4bce-ae0a-854ce80658b0"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((3456, 12000), (3456, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 87
        }
      ],
      "source": [
        "for i in range(len(note)):\n",
        "  X_NR = np.append(X_NR, [note[i]*1.5], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**decrease volume**"
      ],
      "metadata": {
        "id": "K9ztf7T8v_kP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "for i in range(len(note)):\n",
        "  X_NR = np.append(X_NR, [note[i]*0.5], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2QxhQNU5sm_6",
        "outputId": "9aca21c6-6477-479a-badd-29177e147eb6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((3888, 12000), (3888, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 88
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "for i in range(len(note)):\n",
        "  X_NR = np.append(X_NR, [note[i]*0.75], axis=0)\n",
        "  y_true_NR_aug = np.append(y_true_NR_aug, [y_true_NR[i]], axis=0)\n",
        "\n",
        "X_NR.shape, y_true_NR_aug.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xPvfJ8zvs-rj",
        "outputId": "d3434496-d25e-42b7-8eb2-9ea5594ee6e6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((4320, 12000), (4320, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 89
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### **Split Dataset**"
      ],
      "metadata": {
        "id": "KGWOhtvqgfQI"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(X_NR, y_true_NR_aug, test_size=0.2)\n",
        "X_train.shape, y_train.shape, X_test.shape, y_test.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4TlHAyIpgmAi",
        "outputId": "84e3d7aa-0f48-44be-e2a5-17a1326f67aa"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((3456, 12000), (3456, 432), (864, 12000), (864, 432))"
            ]
          },
          "metadata": {},
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xzGuRMZc2UUT"
      },
      "source": [
        "## **Implementation**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 644
        },
        "id": "O8No160X1Nft",
        "outputId": "bdeef6fd-811c-4554-fafa-762e5f99d235"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<IPython.core.display.Image object>"
            ],
            "image/png": "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\n"
          },
          "metadata": {},
          "execution_count": 95
        }
      ],
      "source": [
        "input = Input(shape=(X_NR.shape[1],))\n",
        "x = Dense(128, activation='relu')(input)\n",
        "x = Dense(64, activation='relu')(x)\n",
        "x = Dense(64, activation='relu')(x)\n",
        "x = Dense(128, activation='sigmoid')(x)\n",
        "out = Dense(432, activation=\"softmax\")(x)\n",
        "\n",
        "NR_model = Model(inputs=input, outputs=out)\n",
        "\n",
        "# Compile the Model\n",
        "NR_model.compile(loss=\"categorical_crossentropy\", optimizer=Adam(0.001), metrics=['accuracy'])\n",
        "\n",
        "filepath = '/content/drive/MyDrive/Deep Learning/Music Generator/NR_best_model.hdf5'\n",
        "checkpoint = ModelCheckpoint(filepath=filepath,\n",
        "                              monitor='accuracy',\n",
        "                              verbose=0,\n",
        "                              save_best_only=True,\n",
        "                              mode='max')\n",
        "\n",
        "plot_model(NR_model, to_file=\"NR_model.png\", show_shapes=True)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## **NR Model**"
      ],
      "metadata": {
        "id": "US2y4cnYzL3I"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "MBiuqpJr8NhN",
        "outputId": "0db68b68-0c9e-49f1-b840-123722cd17f5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "54/54 - 2s - loss: 6.1425 - accuracy: 0.0052 - val_loss: 6.0642 - val_accuracy: 0.0104 - 2s/epoch - 36ms/step\n",
            "Epoch 2/100\n",
            "54/54 - 1s - loss: 5.8021 - accuracy: 0.0648 - val_loss: 5.7349 - val_accuracy: 0.0613 - 1s/epoch - 21ms/step\n",
            "Epoch 3/100\n",
            "54/54 - 1s - loss: 5.2617 - accuracy: 0.1875 - val_loss: 5.1700 - val_accuracy: 0.1586 - 1s/epoch - 22ms/step\n",
            "Epoch 4/100\n",
            "54/54 - 1s - loss: 4.5614 - accuracy: 0.3449 - val_loss: 4.5196 - val_accuracy: 0.2882 - 1s/epoch - 24ms/step\n",
            "Epoch 5/100\n",
            "54/54 - 2s - loss: 3.8418 - accuracy: 0.4890 - val_loss: 3.8698 - val_accuracy: 0.4051 - 2s/epoch - 35ms/step\n",
            "Epoch 6/100\n",
            "54/54 - 2s - loss: 3.2049 - accuracy: 0.5885 - val_loss: 3.2924 - val_accuracy: 0.4896 - 2s/epoch - 35ms/step\n",
            "Epoch 7/100\n",
            "54/54 - 1s - loss: 2.6536 - accuracy: 0.6826 - val_loss: 2.8075 - val_accuracy: 0.5486 - 1s/epoch - 22ms/step\n",
            "Epoch 8/100\n",
            "54/54 - 1s - loss: 2.2154 - accuracy: 0.7494 - val_loss: 2.4108 - val_accuracy: 0.6215 - 1s/epoch - 22ms/step\n",
            "Epoch 9/100\n",
            "54/54 - 1s - loss: 1.8627 - accuracy: 0.7946 - val_loss: 2.0874 - val_accuracy: 0.6725 - 1s/epoch - 22ms/step\n",
            "Epoch 10/100\n",
            "54/54 - 1s - loss: 1.5818 - accuracy: 0.8287 - val_loss: 1.8250 - val_accuracy: 0.7176 - 1s/epoch - 21ms/step\n",
            "Epoch 11/100\n",
            "54/54 - 1s - loss: 1.3627 - accuracy: 0.8530 - val_loss: 1.6305 - val_accuracy: 0.7338 - 1s/epoch - 22ms/step\n",
            "Epoch 12/100\n",
            "54/54 - 1s - loss: 1.1852 - accuracy: 0.8620 - val_loss: 1.4539 - val_accuracy: 0.7697 - 1s/epoch - 21ms/step\n",
            "Epoch 13/100\n",
            "54/54 - 1s - loss: 1.0370 - accuracy: 0.8747 - val_loss: 1.3197 - val_accuracy: 0.7824 - 1s/epoch - 21ms/step\n",
            "Epoch 14/100\n",
            "54/54 - 1s - loss: 0.9186 - accuracy: 0.8892 - val_loss: 1.2129 - val_accuracy: 0.8056 - 1s/epoch - 21ms/step\n",
            "Epoch 15/100\n",
            "54/54 - 1s - loss: 0.8169 - accuracy: 0.8990 - val_loss: 1.1135 - val_accuracy: 0.8009 - 1s/epoch - 20ms/step\n",
            "Epoch 16/100\n",
            "54/54 - 1s - loss: 0.7330 - accuracy: 0.9094 - val_loss: 1.0162 - val_accuracy: 0.8194 - 1s/epoch - 22ms/step\n",
            "Epoch 17/100\n",
            "54/54 - 1s - loss: 0.6558 - accuracy: 0.9146 - val_loss: 0.9384 - val_accuracy: 0.8264 - 1s/epoch - 21ms/step\n",
            "Epoch 18/100\n",
            "54/54 - 1s - loss: 0.5913 - accuracy: 0.9216 - val_loss: 0.8821 - val_accuracy: 0.8368 - 1s/epoch - 21ms/step\n",
            "Epoch 19/100\n",
            "54/54 - 1s - loss: 0.5426 - accuracy: 0.9277 - val_loss: 0.8285 - val_accuracy: 0.8426 - 1s/epoch - 21ms/step\n",
            "Epoch 20/100\n",
            "54/54 - 1s - loss: 0.4911 - accuracy: 0.9314 - val_loss: 0.7890 - val_accuracy: 0.8403 - 1s/epoch - 20ms/step\n",
            "Epoch 21/100\n",
            "54/54 - 1s - loss: 0.4464 - accuracy: 0.9421 - val_loss: 0.7451 - val_accuracy: 0.8600 - 1s/epoch - 22ms/step\n",
            "Epoch 22/100\n",
            "54/54 - 1s - loss: 0.4116 - accuracy: 0.9459 - val_loss: 0.6999 - val_accuracy: 0.8623 - 1s/epoch - 21ms/step\n",
            "Epoch 23/100\n",
            "54/54 - 1s - loss: 0.3759 - accuracy: 0.9543 - val_loss: 0.6685 - val_accuracy: 0.8738 - 1s/epoch - 21ms/step\n",
            "Epoch 24/100\n",
            "54/54 - 1s - loss: 0.3453 - accuracy: 0.9578 - val_loss: 0.6368 - val_accuracy: 0.8715 - 1s/epoch - 19ms/step\n",
            "Epoch 25/100\n",
            "54/54 - 1s - loss: 0.3229 - accuracy: 0.9609 - val_loss: 0.6179 - val_accuracy: 0.8727 - 1s/epoch - 19ms/step\n",
            "Epoch 26/100\n",
            "54/54 - 1s - loss: 0.2996 - accuracy: 0.9609 - val_loss: 0.5785 - val_accuracy: 0.8866 - 1s/epoch - 21ms/step\n",
            "Epoch 27/100\n",
            "54/54 - 1s - loss: 0.2744 - accuracy: 0.9705 - val_loss: 0.5568 - val_accuracy: 0.8889 - 1s/epoch - 22ms/step\n",
            "Epoch 28/100\n",
            "54/54 - 1s - loss: 0.2563 - accuracy: 0.9696 - val_loss: 0.5366 - val_accuracy: 0.8889 - 1s/epoch - 19ms/step\n",
            "Epoch 29/100\n",
            "54/54 - 1s - loss: 0.2379 - accuracy: 0.9719 - val_loss: 0.5153 - val_accuracy: 0.8912 - 1s/epoch - 21ms/step\n",
            "Epoch 30/100\n",
            "54/54 - 1s - loss: 0.2228 - accuracy: 0.9771 - val_loss: 0.4960 - val_accuracy: 0.8900 - 1s/epoch - 19ms/step\n",
            "Epoch 31/100\n",
            "54/54 - 1s - loss: 0.2080 - accuracy: 0.9769 - val_loss: 0.4813 - val_accuracy: 0.8970 - 1s/epoch - 22ms/step\n",
            "Epoch 32/100\n",
            "54/54 - 1s - loss: 0.1946 - accuracy: 0.9777 - val_loss: 0.4692 - val_accuracy: 0.8935 - 1s/epoch - 19ms/step\n",
            "Epoch 33/100\n",
            "54/54 - 1s - loss: 0.1825 - accuracy: 0.9797 - val_loss: 0.4640 - val_accuracy: 0.8889 - 999ms/epoch - 19ms/step\n",
            "Epoch 34/100\n",
            "54/54 - 1s - loss: 0.1726 - accuracy: 0.9786 - val_loss: 0.4432 - val_accuracy: 0.8993 - 1s/epoch - 21ms/step\n",
            "Epoch 35/100\n",
            "54/54 - 1s - loss: 0.1607 - accuracy: 0.9821 - val_loss: 0.4321 - val_accuracy: 0.9005 - 1s/epoch - 22ms/step\n",
            "Epoch 36/100\n",
            "54/54 - 1s - loss: 0.1500 - accuracy: 0.9812 - val_loss: 0.4213 - val_accuracy: 0.9005 - 1s/epoch - 20ms/step\n",
            "Epoch 37/100\n",
            "54/54 - 1s - loss: 0.1450 - accuracy: 0.9832 - val_loss: 0.4061 - val_accuracy: 0.9155 - 1s/epoch - 22ms/step\n",
            "Epoch 38/100\n",
            "54/54 - 1s - loss: 0.1359 - accuracy: 0.9852 - val_loss: 0.3965 - val_accuracy: 0.9062 - 1s/epoch - 20ms/step\n",
            "Epoch 39/100\n",
            "54/54 - 1s - loss: 0.1283 - accuracy: 0.9829 - val_loss: 0.3958 - val_accuracy: 0.9109 - 1s/epoch - 19ms/step\n",
            "Epoch 40/100\n",
            "54/54 - 1s - loss: 0.1228 - accuracy: 0.9826 - val_loss: 0.3888 - val_accuracy: 0.9016 - 1s/epoch - 19ms/step\n",
            "Epoch 41/100\n",
            "54/54 - 1s - loss: 0.1139 - accuracy: 0.9838 - val_loss: 0.3739 - val_accuracy: 0.9167 - 1s/epoch - 22ms/step\n",
            "Epoch 42/100\n",
            "54/54 - 1s - loss: 0.1075 - accuracy: 0.9855 - val_loss: 0.3620 - val_accuracy: 0.9132 - 1s/epoch - 19ms/step\n",
            "Epoch 43/100\n",
            "54/54 - 1s - loss: 0.1050 - accuracy: 0.9835 - val_loss: 0.3747 - val_accuracy: 0.9132 - 1s/epoch - 19ms/step\n",
            "Epoch 44/100\n",
            "54/54 - 1s - loss: 0.0983 - accuracy: 0.9870 - val_loss: 0.3515 - val_accuracy: 0.9178 - 1s/epoch - 22ms/step\n",
            "Epoch 45/100\n",
            "54/54 - 1s - loss: 0.0906 - accuracy: 0.9878 - val_loss: 0.3542 - val_accuracy: 0.9167 - 1s/epoch - 19ms/step\n",
            "Epoch 46/100\n",
            "54/54 - 1s - loss: 0.0889 - accuracy: 0.9864 - val_loss: 0.3523 - val_accuracy: 0.9109 - 1s/epoch - 20ms/step\n",
            "Epoch 47/100\n",
            "54/54 - 1s - loss: 0.0893 - accuracy: 0.9855 - val_loss: 0.3382 - val_accuracy: 0.9201 - 1s/epoch - 21ms/step\n",
            "Epoch 48/100\n",
            "54/54 - 1s - loss: 0.0825 - accuracy: 0.9861 - val_loss: 0.3463 - val_accuracy: 0.9178 - 1s/epoch - 19ms/step\n",
            "Epoch 49/100\n",
            "54/54 - 1s - loss: 0.0807 - accuracy: 0.9878 - val_loss: 0.3337 - val_accuracy: 0.9213 - 1s/epoch - 21ms/step\n",
            "Epoch 50/100\n",
            "54/54 - 1s - loss: 0.0756 - accuracy: 0.9873 - val_loss: 0.3310 - val_accuracy: 0.9225 - 1s/epoch - 22ms/step\n",
            "Epoch 51/100\n",
            "54/54 - 1s - loss: 0.0742 - accuracy: 0.9887 - val_loss: 0.3333 - val_accuracy: 0.9236 - 1s/epoch - 21ms/step\n",
            "Epoch 52/100\n",
            "54/54 - 1s - loss: 0.0721 - accuracy: 0.9890 - val_loss: 0.3385 - val_accuracy: 0.9097 - 1s/epoch - 19ms/step\n",
            "Epoch 53/100\n",
            "54/54 - 1s - loss: 0.0676 - accuracy: 0.9870 - val_loss: 0.3245 - val_accuracy: 0.9132 - 1s/epoch - 19ms/step\n",
            "Epoch 54/100\n",
            "54/54 - 1s - loss: 0.0667 - accuracy: 0.9870 - val_loss: 0.3180 - val_accuracy: 0.9225 - 1s/epoch - 19ms/step\n",
            "Epoch 55/100\n",
            "54/54 - 1s - loss: 0.0628 - accuracy: 0.9887 - val_loss: 0.3213 - val_accuracy: 0.9167 - 1s/epoch - 19ms/step\n",
            "Epoch 56/100\n",
            "54/54 - 1s - loss: 0.0631 - accuracy: 0.9881 - val_loss: 0.3256 - val_accuracy: 0.9190 - 1s/epoch - 20ms/step\n",
            "Epoch 57/100\n",
            "54/54 - 1s - loss: 0.0628 - accuracy: 0.9878 - val_loss: 0.3182 - val_accuracy: 0.9225 - 1s/epoch - 20ms/step\n",
            "Epoch 58/100\n",
            "54/54 - 1s - loss: 0.0612 - accuracy: 0.9873 - val_loss: 0.3268 - val_accuracy: 0.9259 - 1s/epoch - 22ms/step\n",
            "Epoch 59/100\n",
            "54/54 - 1s - loss: 0.0571 - accuracy: 0.9884 - val_loss: 0.3236 - val_accuracy: 0.9201 - 1s/epoch - 19ms/step\n",
            "Epoch 60/100\n",
            "54/54 - 1s - loss: 0.0555 - accuracy: 0.9902 - val_loss: 0.3076 - val_accuracy: 0.9282 - 1s/epoch - 21ms/step\n",
            "Epoch 61/100\n",
            "54/54 - 1s - loss: 0.0506 - accuracy: 0.9899 - val_loss: 0.3127 - val_accuracy: 0.9236 - 1s/epoch - 19ms/step\n",
            "Epoch 62/100\n",
            "54/54 - 1s - loss: 0.0514 - accuracy: 0.9899 - val_loss: 0.2989 - val_accuracy: 0.9178 - 1s/epoch - 19ms/step\n",
            "Epoch 63/100\n",
            "54/54 - 1s - loss: 0.0484 - accuracy: 0.9905 - val_loss: 0.3048 - val_accuracy: 0.9259 - 1s/epoch - 19ms/step\n",
            "Epoch 64/100\n",
            "54/54 - 1s - loss: 0.0476 - accuracy: 0.9905 - val_loss: 0.3037 - val_accuracy: 0.9236 - 1s/epoch - 19ms/step\n",
            "Epoch 65/100\n",
            "54/54 - 1s - loss: 0.0449 - accuracy: 0.9910 - val_loss: 0.2909 - val_accuracy: 0.9236 - 1s/epoch - 19ms/step\n",
            "Epoch 66/100\n",
            "54/54 - 1s - loss: 0.0488 - accuracy: 0.9887 - val_loss: 0.3011 - val_accuracy: 0.9167 - 1s/epoch - 20ms/step\n",
            "Epoch 67/100\n",
            "54/54 - 1s - loss: 0.0512 - accuracy: 0.9890 - val_loss: 0.2995 - val_accuracy: 0.9225 - 1s/epoch - 21ms/step\n",
            "Epoch 68/100\n",
            "54/54 - 1s - loss: 0.0485 - accuracy: 0.9893 - val_loss: 0.2903 - val_accuracy: 0.9236 - 1s/epoch - 20ms/step\n",
            "Epoch 69/100\n",
            "54/54 - 1s - loss: 0.0450 - accuracy: 0.9881 - val_loss: 0.3074 - val_accuracy: 0.9155 - 1s/epoch - 19ms/step\n",
            "Epoch 70/100\n",
            "54/54 - 1s - loss: 0.0452 - accuracy: 0.9907 - val_loss: 0.3094 - val_accuracy: 0.9190 - 1s/epoch - 20ms/step\n",
            "Epoch 71/100\n",
            "54/54 - 1s - loss: 0.0452 - accuracy: 0.9899 - val_loss: 0.3041 - val_accuracy: 0.9144 - 1s/epoch - 19ms/step\n",
            "Epoch 72/100\n",
            "54/54 - 1s - loss: 0.0450 - accuracy: 0.9905 - val_loss: 0.3124 - val_accuracy: 0.9178 - 1s/epoch - 19ms/step\n",
            "Epoch 73/100\n",
            "54/54 - 1s - loss: 0.0440 - accuracy: 0.9899 - val_loss: 0.2987 - val_accuracy: 0.9213 - 1s/epoch - 20ms/step\n",
            "Epoch 74/100\n",
            "54/54 - 1s - loss: 0.0433 - accuracy: 0.9896 - val_loss: 0.2919 - val_accuracy: 0.9248 - 1s/epoch - 19ms/step\n",
            "Epoch 75/100\n",
            "54/54 - 1s - loss: 0.0378 - accuracy: 0.9910 - val_loss: 0.2875 - val_accuracy: 0.9213 - 1s/epoch - 19ms/step\n",
            "Epoch 76/100\n",
            "54/54 - 1s - loss: 0.0352 - accuracy: 0.9922 - val_loss: 0.2706 - val_accuracy: 0.9294 - 1s/epoch - 27ms/step\n",
            "Epoch 77/100\n",
            "54/54 - 1s - loss: 0.0368 - accuracy: 0.9907 - val_loss: 0.2877 - val_accuracy: 0.9329 - 1s/epoch - 22ms/step\n",
            "Epoch 78/100\n",
            "54/54 - 1s - loss: 0.0389 - accuracy: 0.9899 - val_loss: 0.2896 - val_accuracy: 0.9213 - 1s/epoch - 19ms/step\n",
            "Epoch 79/100\n",
            "54/54 - 1s - loss: 0.0337 - accuracy: 0.9913 - val_loss: 0.2857 - val_accuracy: 0.9248 - 1s/epoch - 19ms/step\n",
            "Epoch 80/100\n",
            "54/54 - 1s - loss: 0.0319 - accuracy: 0.9913 - val_loss: 0.2866 - val_accuracy: 0.9213 - 1s/epoch - 19ms/step\n",
            "Epoch 81/100\n",
            "54/54 - 2s - loss: 0.0319 - accuracy: 0.9916 - val_loss: 0.2873 - val_accuracy: 0.9259 - 2s/epoch - 29ms/step\n",
            "Epoch 82/100\n",
            "54/54 - 1s - loss: 0.0316 - accuracy: 0.9919 - val_loss: 0.2808 - val_accuracy: 0.9317 - 1s/epoch - 19ms/step\n",
            "Epoch 83/100\n",
            "54/54 - 1s - loss: 0.0306 - accuracy: 0.9933 - val_loss: 0.2731 - val_accuracy: 0.9294 - 1s/epoch - 19ms/step\n",
            "Epoch 84/100\n",
            "54/54 - 1s - loss: 0.0338 - accuracy: 0.9916 - val_loss: 0.3041 - val_accuracy: 0.9201 - 1s/epoch - 20ms/step\n",
            "Epoch 85/100\n",
            "54/54 - 1s - loss: 0.0292 - accuracy: 0.9916 - val_loss: 0.2911 - val_accuracy: 0.9225 - 1s/epoch - 20ms/step\n",
            "Epoch 86/100\n",
            "54/54 - 1s - loss: 0.0289 - accuracy: 0.9922 - val_loss: 0.2820 - val_accuracy: 0.9271 - 1s/epoch - 20ms/step\n",
            "Epoch 87/100\n",
            "54/54 - 1s - loss: 0.0276 - accuracy: 0.9910 - val_loss: 0.2820 - val_accuracy: 0.9271 - 1s/epoch - 19ms/step\n",
            "Epoch 88/100\n",
            "54/54 - 1s - loss: 0.0291 - accuracy: 0.9910 - val_loss: 0.2878 - val_accuracy: 0.9259 - 1s/epoch - 19ms/step\n",
            "Epoch 89/100\n",
            "54/54 - 1s - loss: 0.0304 - accuracy: 0.9913 - val_loss: 0.2846 - val_accuracy: 0.9213 - 1s/epoch - 19ms/step\n",
            "Epoch 90/100\n",
            "54/54 - 1s - loss: 0.0290 - accuracy: 0.9910 - val_loss: 0.2731 - val_accuracy: 0.9306 - 1s/epoch - 19ms/step\n",
            "Epoch 91/100\n",
            "54/54 - 1s - loss: 0.0269 - accuracy: 0.9916 - val_loss: 0.2872 - val_accuracy: 0.9271 - 1s/epoch - 19ms/step\n",
            "Epoch 92/100\n",
            "54/54 - 1s - loss: 0.0272 - accuracy: 0.9913 - val_loss: 0.2805 - val_accuracy: 0.9282 - 1s/epoch - 19ms/step\n",
            "Epoch 93/100\n",
            "54/54 - 1s - loss: 0.0261 - accuracy: 0.9916 - val_loss: 0.2892 - val_accuracy: 0.9236 - 1s/epoch - 20ms/step\n",
            "Epoch 94/100\n",
            "54/54 - 1s - loss: 0.0307 - accuracy: 0.9902 - val_loss: 0.2842 - val_accuracy: 0.9259 - 1s/epoch - 20ms/step\n",
            "Epoch 95/100\n",
            "54/54 - 1s - loss: 0.0366 - accuracy: 0.9887 - val_loss: 0.2856 - val_accuracy: 0.9271 - 1s/epoch - 19ms/step\n",
            "Epoch 96/100\n",
            "54/54 - 1s - loss: 0.0317 - accuracy: 0.9902 - val_loss: 0.2813 - val_accuracy: 0.9236 - 1s/epoch - 19ms/step\n",
            "Epoch 97/100\n",
            "54/54 - 1s - loss: 0.0374 - accuracy: 0.9887 - val_loss: 0.3237 - val_accuracy: 0.9178 - 1s/epoch - 19ms/step\n",
            "Epoch 98/100\n",
            "54/54 - 1s - loss: 0.0435 - accuracy: 0.9884 - val_loss: 0.2855 - val_accuracy: 0.9271 - 1s/epoch - 19ms/step\n",
            "Epoch 99/100\n",
            "54/54 - 1s - loss: 0.0379 - accuracy: 0.9902 - val_loss: 0.2983 - val_accuracy: 0.9329 - 1s/epoch - 19ms/step\n",
            "Epoch 100/100\n",
            "54/54 - 1s - loss: 0.0325 - accuracy: 0.9910 - val_loss: 0.2790 - val_accuracy: 0.9259 - 1s/epoch - 19ms/step\n",
            "\n",
            "best model based on min training set loss:  acc= 0.9916087985038757  loss= 0.02608380652964115\n",
            "best model based on min test set loss:  acc= 0.9293981194496155  loss= 0.27056607604026794\n",
            "best model based on max training set accuracy:  acc= 0.9933449029922485  loss= 0.0305897518992424\n",
            "best model based on max test set accuracy:  acc= 0.9328703880310059  loss= 0.2876511216163635\n",
            "\n",
            "evaluate dataset with best model based on maximum test set accuracy\n",
            "evaluate train set=  [0.034524187445640564, 0.9939236044883728]\n",
            "evaluate test set=  [0.28765109181404114, 0.9328703880310059]\n",
            "\n",
            " ------------------------------ metrics for traning set ------------------------------\n",
            "confusion matrix: \n",
            " [[ 8  0  0 ...  0  0  0]\n",
            " [ 0  7  0 ...  0  0  0]\n",
            " [ 0  0  6 ...  0  0  0]\n",
            " ...\n",
            " [ 0  0  0 ...  8  0  0]\n",
            " [ 0  0  0 ...  0  9  0]\n",
            " [ 0  0  0 ...  0  0 10]]\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "   A_dim_2_1      1.000     1.000     1.000         8\n",
            "   A_dim_2_0      1.000     1.000     1.000         7\n",
            "   A_maj_5_1      1.000     0.857     0.923         7\n",
            "   A_maj_3_1      1.000     1.000     1.000         9\n",
            "   A_min_2_0      1.000     1.000     1.000        10\n",
            "   A_maj_2_0      1.000     1.000     1.000         8\n",
            "   A_maj_3_0      1.000     1.000     1.000        10\n",
            "   A_maj_5_0      1.000     1.000     1.000         7\n",
            "   A_dim_3_1      1.000     1.000     1.000         8\n",
            "   A_dim_6_1      1.000     1.000     1.000         9\n",
            "   A_min_2_1      1.000     1.000     1.000         9\n",
            "   A_dim_5_1      0.875     1.000     0.933         7\n",
            "   A_dim_7_0      1.000     1.000     1.000        10\n",
            "   A_dim_5_0      1.000     1.000     1.000         8\n",
            "   A_maj_7_0      1.000     1.000     1.000         7\n",
            "   A_maj_6_0      1.000     1.000     1.000         9\n",
            "   A_maj_4_0      1.000     1.000     1.000         8\n",
            "   A_dim_3_0      1.000     1.000     1.000         7\n",
            "   A_dim_4_1      1.000     1.000     1.000         8\n",
            "   A_maj_6_1      1.000     1.000     1.000         8\n",
            "   A_maj_4_1      1.000     1.000     1.000         9\n",
            "   A_dim_4_0      1.000     1.000     1.000         8\n",
            "   A_maj_2_1      1.000     1.000     1.000         8\n",
            "   A_maj_7_1      1.000     1.000     1.000         8\n",
            "   A_dim_7_1      1.000     1.000     1.000         7\n",
            "   A_dim_6_0      1.000     1.000     1.000         8\n",
            "   B_dim_3_1      1.000     1.000     1.000         7\n",
            "   B_dim_7_0      1.000     1.000     1.000         9\n",
            "   B_maj_6_0      1.000     1.000     1.000         9\n",
            "   B_dim_7_1      1.000     1.000     1.000         6\n",
            "   B_min_2_1      1.000     1.000     1.000         7\n",
            "   B_dim_5_0      1.000     0.857     0.923         7\n",
            "   B_min_2_0      1.000     1.000     1.000         9\n",
            "   B_dim_6_1      1.000     1.000     1.000         5\n",
            "   B_maj_7_1      1.000     1.000     1.000        10\n",
            "   B_dim_4_1      1.000     1.000     1.000         9\n",
            "   B_maj_3_0      1.000     1.000     1.000         9\n",
            "   B_dim_3_0      1.000     1.000     1.000         8\n",
            "   A_min_6_0      1.000     1.000     1.000         6\n",
            "   B_maj_6_1      1.000     1.000     1.000         7\n",
            "   B_dim_5_1      1.000     1.000     1.000         2\n",
            "   A_min_6_1      1.000     1.000     1.000         8\n",
            "   A_min_4_1      1.000     1.000     1.000         7\n",
            "   B_maj_5_1      1.000     1.000     1.000         9\n",
            "   B_maj_2_0      1.000     1.000     1.000         7\n",
            "   B_dim_2_0      1.000     1.000     1.000         9\n",
            "   A_min_7_1      1.000     1.000     1.000        10\n",
            "   A_min_4_0      1.000     1.000     1.000         8\n",
            "   A_min_3_0      1.000     1.000     1.000         9\n",
            "   B_dim_4_0      1.000     1.000     1.000        10\n",
            "   B_maj_4_0      1.000     1.000     1.000         9\n",
            "   A_min_7_0      1.000     1.000     1.000         4\n",
            "   B_maj_4_1      1.000     1.000     1.000         9\n",
            "   A_min_3_1      1.000     1.000     1.000         8\n",
            "   B_maj_2_1      1.000     1.000     1.000         8\n",
            "   A_min_5_0      1.000     1.000     1.000         7\n",
            "   B_maj_3_1      1.000     1.000     1.000         7\n",
            "   B_maj_5_0      1.000     1.000     1.000         7\n",
            "   B_dim_2_1      1.000     1.000     1.000         9\n",
            "   B_maj_7_0      1.000     1.000     1.000         9\n",
            "   B_dim_6_0      1.000     1.000     1.000         9\n",
            "   B_min_3_0      1.000     1.000     1.000         8\n",
            "   A_min_5_1      1.000     0.857     0.923         7\n",
            "  Bb_dim_4_1      1.000     1.000     1.000         6\n",
            "  Bb_maj_7_0      1.000     1.000     1.000         8\n",
            "   B_min_5_1      1.000     1.000     1.000         8\n",
            "   B_min_4_0      1.000     1.000     1.000         7\n",
            "  Bb_maj_4_1      1.000     1.000     1.000         9\n",
            "   B_min_7_0      1.000     1.000     1.000         6\n",
            "  Bb_min_2_0      1.000     1.000     1.000         6\n",
            "  Bb_dim_3_0      1.000     1.000     1.000         7\n",
            "  Bb_dim_3_1      1.000     1.000     1.000         8\n",
            "   B_min_3_1      1.000     1.000     1.000         8\n",
            "  Bb_dim_7_1      1.000     1.000     1.000         8\n",
            "  Bb_maj_3_1      1.000     1.000     1.000         7\n",
            "   B_min_5_0      1.000     1.000     1.000         6\n",
            "   B_min_4_1      1.000     1.000     1.000         7\n",
            "   B_min_7_1      1.000     1.000     1.000         7\n",
            "  Bb_dim_5_1      1.000     0.875     0.933         8\n",
            "  Bb_maj_5_0      0.875     1.000     0.933         7\n",
            "  Bb_dim_4_0      1.000     1.000     1.000         8\n",
            "   B_min_6_0      1.000     1.000     1.000         8\n",
            "  Bb_maj_5_1      1.000     1.000     1.000         7\n",
            "  Bb_min_3_1      1.000     1.000     1.000         7\n",
            "  Bb_maj_3_0      1.000     1.000     1.000        10\n",
            "  Bb_dim_6_1      1.000     1.000     1.000         9\n",
            "  Bb_min_2_1      1.000     1.000     1.000        10\n",
            "  Bb_maj_6_1      1.000     1.000     1.000         9\n",
            "  Bb_min_4_0      1.000     1.000     1.000         7\n",
            "  Bb_maj_2_0      1.000     1.000     1.000         7\n",
            "  Bb_dim_2_1      1.000     1.000     1.000         8\n",
            "  Bb_maj_2_1      1.000     1.000     1.000         9\n",
            "  Bb_maj_7_1      1.000     1.000     1.000        10\n",
            "   B_min_6_1      1.000     1.000     1.000         9\n",
            "  Bb_dim_6_0      1.000     1.000     1.000         7\n",
            "  Bb_dim_7_0      1.000     1.000     1.000         8\n",
            "  Bb_maj_6_0      1.000     1.000     1.000         9\n",
            "  Bb_dim_2_0      1.000     1.000     1.000        10\n",
            "  Bb_min_3_0      1.000     1.000     1.000         8\n",
            "  Bb_dim_5_0      1.000     0.900     0.947        10\n",
            "  Bb_maj_4_0      1.000     1.000     1.000         9\n",
            "   C_dim_4_0      1.000     1.000     1.000         7\n",
            "   C_maj_2_0      1.000     1.000     1.000         7\n",
            "   C_dim_4_1      1.000     1.000     1.000         8\n",
            "   C_dim_6_0      1.000     0.800     0.889         5\n",
            "   C_min_2_0      1.000     1.000     1.000        10\n",
            "   C_maj_6_1      1.000     1.000     1.000         8\n",
            "   C_maj_4_0      1.000     1.000     1.000         6\n",
            "   C_min_2_1      1.000     1.000     1.000         9\n",
            "   C_maj_7_1      1.000     1.000     1.000         7\n",
            "   C_dim_7_0      1.000     1.000     1.000         8\n",
            "   C_min_3_1      1.000     1.000     1.000         4\n",
            "   C_maj_2_1      1.000     1.000     1.000         9\n",
            "  Bb_min_6_0      1.000     1.000     1.000         8\n",
            "   C_maj_5_1      1.000     1.000     1.000         7\n",
            "   C_dim_3_1      1.000     1.000     1.000        10\n",
            "  Bb_min_7_1      1.000     1.000     1.000         9\n",
            "  Bb_min_5_1      1.000     1.000     1.000         6\n",
            "   C_maj_3_1      1.000     1.000     1.000         7\n",
            "   C_maj_3_0      0.667     0.889     0.762         9\n",
            "  Bb_min_4_1      1.000     1.000     1.000         7\n",
            "   C_dim_2_1      1.000     1.000     1.000        10\n",
            "   C_dim_3_0      1.000     1.000     1.000         8\n",
            "   C_dim_2_0      1.000     1.000     1.000         9\n",
            "   C_maj_5_0      1.000     1.000     1.000         7\n",
            "   C_dim_5_1      1.000     1.000     1.000         7\n",
            "  Bb_min_6_1      1.000     1.000     1.000        10\n",
            "   C_maj_7_0      1.000     1.000     1.000         8\n",
            "   C_dim_5_0      1.000     1.000     1.000         8\n",
            "  Bb_min_5_0      0.667     1.000     0.800         6\n",
            "   C_maj_6_0      1.000     1.000     1.000         7\n",
            "   C_dim_6_1      1.000     1.000     1.000         7\n",
            "   C_min_4_1      1.000     1.000     1.000         8\n",
            "  Bb_min_7_0      1.000     1.000     1.000         7\n",
            "   C_min_3_0      1.000     1.000     1.000        10\n",
            "   C_maj_4_1      1.000     1.000     1.000         8\n",
            "   C_dim_7_1      1.000     1.000     1.000         7\n",
            "   C_min_4_0      1.000     1.000     1.000         8\n",
            "  Cs_maj_3_0      1.000     1.000     1.000         8\n",
            "  Cs_maj_4_0      1.000     1.000     1.000         8\n",
            "  Cs_maj_2_1      1.000     1.000     1.000         8\n",
            "  Cs_dim_6_0      1.000     0.857     0.923         7\n",
            "   C_min_5_1      1.000     1.000     1.000         9\n",
            "  Cs_min_2_0      1.000     1.000     1.000         9\n",
            "  Cs_maj_3_1      1.000     1.000     1.000         5\n",
            "  Cs_dim_3_1      1.000     1.000     1.000        10\n",
            "  Cs_min_3_0      1.000     1.000     1.000         7\n",
            "  Cs_dim_4_0      1.000     1.000     1.000         7\n",
            "  Cs_maj_2_0      1.000     1.000     1.000         9\n",
            "  Cs_dim_6_1      1.000     1.000     1.000         7\n",
            "  Cs_maj_7_0      1.000     1.000     1.000         7\n",
            "  Cs_dim_5_1      1.000     1.000     1.000         9\n",
            "  Cs_maj_6_1      1.000     0.857     0.923         7\n",
            "  Cs_dim_7_0      1.000     1.000     1.000        10\n",
            "  Cs_dim_2_0      1.000     1.000     1.000        10\n",
            "   C_min_6_1      1.000     1.000     1.000         9\n",
            "  Cs_maj_5_0      1.000     1.000     1.000        10\n",
            "   C_min_7_1      1.000     1.000     1.000         9\n",
            "  Cs_min_4_1      1.000     1.000     1.000         8\n",
            "  Cs_dim_7_1      1.000     1.000     1.000         8\n",
            "  Cs_dim_5_0      1.000     1.000     1.000         7\n",
            "  Cs_maj_6_0      0.750     1.000     0.857         9\n",
            "  Cs_min_5_0      1.000     1.000     1.000         9\n",
            "   C_min_6_0      1.000     1.000     1.000         8\n",
            "  Cs_maj_5_1      1.000     1.000     1.000         7\n",
            "  Cs_min_3_1      1.000     1.000     1.000         8\n",
            "  Cs_maj_7_1      1.000     1.000     1.000         7\n",
            "  Cs_min_2_1      1.000     1.000     1.000         9\n",
            "  Cs_dim_4_1      1.000     1.000     1.000         7\n",
            "  Cs_dim_3_0      1.000     1.000     1.000         6\n",
            "  Cs_dim_2_1      1.000     1.000     1.000         8\n",
            "  Cs_maj_4_1      1.000     1.000     1.000         7\n",
            "  Cs_min_4_0      1.000     1.000     1.000        10\n",
            "   C_min_7_0      1.000     1.000     1.000         7\n",
            "   C_min_5_0      1.000     1.000     1.000        10\n",
            "   D_dim_3_0      1.000     1.000     1.000         7\n",
            "   D_maj_5_0      1.000     1.000     1.000         8\n",
            "   D_dim_4_0      1.000     1.000     1.000         9\n",
            "  Cs_min_7_0      1.000     1.000     1.000         8\n",
            "   D_maj_7_1      1.000     1.000     1.000         8\n",
            "   D_min_5_0      1.000     1.000     1.000         7\n",
            "   D_dim_3_1      1.000     1.000     1.000         8\n",
            "   D_maj_4_0      1.000     1.000     1.000         6\n",
            "   D_dim_7_1      1.000     1.000     1.000        10\n",
            "   D_min_4_1      1.000     1.000     1.000         5\n",
            "   D_min_4_0      1.000     1.000     1.000         4\n",
            "   D_maj_3_0      0.667     0.333     0.444         6\n",
            "   D_maj_7_0      1.000     1.000     1.000         8\n",
            "   D_min_2_1      1.000     1.000     1.000         8\n",
            "   D_dim_2_0      1.000     1.000     1.000         9\n",
            "   D_maj_6_0      1.000     1.000     1.000         8\n",
            "   D_min_3_0      1.000     1.000     1.000         9\n",
            "   D_maj_5_1      1.000     1.000     1.000         4\n",
            "   D_dim_6_0      1.000     1.000     1.000        10\n",
            "   D_dim_5_1      1.000     1.000     1.000         9\n",
            "   D_maj_2_0      1.000     1.000     1.000         8\n",
            "   D_maj_4_1      1.000     1.000     1.000         9\n",
            "  Cs_min_5_1      1.000     1.000     1.000         6\n",
            "   D_maj_2_1      1.000     1.000     1.000         7\n",
            "   D_maj_3_1      1.000     1.000     1.000         7\n",
            "  Cs_min_6_0      1.000     1.000     1.000         6\n",
            "   D_min_2_0      1.000     1.000     1.000         9\n",
            "   D_min_5_1      1.000     1.000     1.000         7\n",
            "   D_dim_4_1      1.000     1.000     1.000         5\n",
            "   D_dim_6_1      1.000     0.875     0.933         8\n",
            "   D_dim_2_1      1.000     1.000     1.000         8\n",
            "   D_dim_5_0      1.000     1.000     1.000         8\n",
            "   D_dim_7_0      1.000     1.000     1.000         8\n",
            "  Cs_min_6_1      1.000     1.000     1.000        10\n",
            "   D_maj_6_1      1.000     1.000     1.000         9\n",
            "   D_min_3_1      1.000     1.000     1.000        10\n",
            "  Cs_min_7_1      1.000     1.000     1.000         8\n",
            "   E_maj_2_1      1.000     1.000     1.000         8\n",
            "   E_dim_6_1      1.000     1.000     1.000         7\n",
            "   D_min_7_0      1.000     1.000     1.000         9\n",
            "   D_min_6_0      1.000     1.000     1.000        10\n",
            "   E_min_3_0      1.000     1.000     1.000         8\n",
            "   D_min_6_1      1.000     0.857     0.923         7\n",
            "   E_dim_7_1      1.000     1.000     1.000         6\n",
            "   E_min_2_1      1.000     1.000     1.000         8\n",
            "   E_dim_2_1      1.000     1.000     1.000         7\n",
            "   E_dim_5_1      1.000     1.000     1.000         8\n",
            "   E_dim_7_0      1.000     1.000     1.000         8\n",
            "   E_dim_3_0      1.000     1.000     1.000         7\n",
            "   E_dim_4_0      1.000     1.000     1.000         6\n",
            "   E_min_2_0      1.000     1.000     1.000         8\n",
            "   E_dim_3_1      1.000     1.000     1.000         8\n",
            "   E_maj_4_0      1.000     1.000     1.000        10\n",
            "   E_min_3_1      1.000     1.000     1.000         9\n",
            "   E_maj_5_0      1.000     1.000     1.000         8\n",
            "   E_maj_7_1      1.000     1.000     1.000         8\n",
            "   E_min_4_1      1.000     1.000     1.000         6\n",
            "   E_min_4_0      1.000     1.000     1.000        10\n",
            "   E_min_6_0      1.000     1.000     1.000         9\n",
            "   E_maj_4_1      1.000     1.000     1.000         8\n",
            "   E_maj_7_0      1.000     1.000     1.000         6\n",
            "   E_dim_4_1      1.000     1.000     1.000         9\n",
            "   E_maj_6_0      1.000     1.000     1.000         9\n",
            "   E_maj_3_1      1.000     1.000     1.000         7\n",
            "   E_dim_5_0      1.000     1.000     1.000        10\n",
            "   E_maj_2_0      1.000     1.000     1.000         7\n",
            "   E_min_5_0      1.000     1.000     1.000         9\n",
            "   E_min_5_1      1.000     1.000     1.000         9\n",
            "   E_dim_2_0      1.000     1.000     1.000         6\n",
            "   E_maj_6_1      1.000     1.000     1.000         6\n",
            "   E_maj_3_0      1.000     1.000     1.000         5\n",
            "   E_maj_5_1      1.000     0.900     0.947        10\n",
            "   E_dim_6_0      1.000     1.000     1.000         8\n",
            "   D_min_7_1      1.000     1.000     1.000         8\n",
            "  Eb_min_4_1      1.000     1.000     1.000         7\n",
            "  Eb_dim_6_1      1.000     1.000     1.000         7\n",
            "  Eb_maj_7_0      1.000     1.000     1.000         8\n",
            "  Eb_min_3_1      1.000     1.000     1.000         9\n",
            "  Eb_min_2_0      1.000     1.000     1.000         9\n",
            "  Eb_maj_5_0      1.000     1.000     1.000         9\n",
            "  Eb_maj_4_0      1.000     1.000     1.000        10\n",
            "  Eb_maj_5_1      1.000     1.000     1.000         8\n",
            "  Eb_dim_5_1      1.000     1.000     1.000         9\n",
            "  Eb_min_5_0      1.000     1.000     1.000         8\n",
            "  Eb_maj_3_0      1.000     1.000     1.000         8\n",
            "  Eb_dim_2_1      1.000     1.000     1.000         9\n",
            "  Eb_maj_3_1      1.000     1.000     1.000         8\n",
            "  Eb_dim_7_0      1.000     1.000     1.000         8\n",
            "  Eb_dim_3_0      1.000     1.000     1.000         8\n",
            "  Eb_dim_4_0      1.000     1.000     1.000         6\n",
            "  Eb_min_5_1      1.000     1.000     1.000        10\n",
            "   E_min_7_0      1.000     1.000     1.000         9\n",
            "  Eb_maj_6_0      1.000     1.000     1.000         5\n",
            "  Eb_maj_2_0      1.000     1.000     1.000         6\n",
            "   E_min_6_1      1.000     1.000     1.000         8\n",
            "   E_min_7_1      1.000     1.000     1.000        10\n",
            "  Eb_dim_2_0      1.000     1.000     1.000         7\n",
            "  Eb_maj_2_1      1.000     1.000     1.000         9\n",
            "  Eb_maj_4_1      1.000     1.000     1.000         9\n",
            "  Eb_min_6_0      1.000     1.000     1.000         6\n",
            "  Eb_maj_7_1      1.000     1.000     1.000         9\n",
            "  Eb_dim_4_1      1.000     1.000     1.000         8\n",
            "  Eb_dim_5_0      1.000     1.000     1.000         4\n",
            "  Eb_min_4_0      1.000     1.000     1.000         7\n",
            "  Eb_min_3_0      1.000     1.000     1.000         7\n",
            "  Eb_min_2_1      1.000     1.000     1.000        10\n",
            "  Eb_dim_7_1      1.000     1.000     1.000         7\n",
            "  Eb_dim_3_1      1.000     1.000     1.000         8\n",
            "  Eb_maj_6_1      1.000     1.000     1.000        10\n",
            "  Eb_dim_6_0      1.000     1.000     1.000         9\n",
            "   F_maj_4_0      1.000     1.000     1.000         9\n",
            "   F_dim_7_1      1.000     1.000     1.000         8\n",
            "   F_dim_2_1      1.000     1.000     1.000         9\n",
            "   F_dim_4_0      1.000     1.000     1.000         8\n",
            "   F_dim_5_1      1.000     1.000     1.000         8\n",
            "   F_dim_7_0      1.000     1.000     1.000         9\n",
            "   F_maj_7_0      1.000     1.000     1.000         9\n",
            "   F_min_4_0      1.000     1.000     1.000         9\n",
            "  Eb_min_7_0      1.000     1.000     1.000         6\n",
            "   F_dim_2_0      1.000     1.000     1.000         8\n",
            "   F_maj_2_0      1.000     1.000     1.000         5\n",
            "   F_maj_6_0      0.833     1.000     0.909         5\n",
            "   F_min_6_1      1.000     1.000     1.000         8\n",
            "   F_min_5_0      1.000     1.000     1.000         8\n",
            "   F_min_2_1      1.000     1.000     1.000         9\n",
            "   F_dim_3_1      1.000     1.000     1.000        10\n",
            "   F_min_6_0      0.727     1.000     0.842         8\n",
            "   F_dim_3_0      1.000     1.000     1.000         8\n",
            "   F_maj_3_1      1.000     1.000     1.000         8\n",
            "   F_dim_6_1      1.000     1.000     1.000         9\n",
            "   F_maj_5_0      1.000     1.000     1.000         9\n",
            "   F_min_4_1      1.000     1.000     1.000         7\n",
            "  Eb_min_7_1      1.000     1.000     1.000         9\n",
            "   F_min_3_0      1.000     1.000     1.000         8\n",
            "   F_min_3_1      1.000     1.000     1.000         8\n",
            "   F_maj_3_0      1.000     1.000     1.000         9\n",
            "   F_min_2_0      1.000     1.000     1.000         7\n",
            "   F_maj_4_1      1.000     1.000     1.000         9\n",
            "   F_dim_4_1      1.000     1.000     1.000         8\n",
            "   F_dim_5_0      1.000     1.000     1.000         8\n",
            "   F_dim_6_0      1.000     0.875     0.933         8\n",
            "   F_min_5_1      1.000     1.000     1.000         9\n",
            "   F_maj_5_1      1.000     1.000     1.000         8\n",
            "   F_maj_7_1      1.000     1.000     1.000         9\n",
            "   F_maj_2_1      1.000     1.000     1.000         9\n",
            "  Eb_min_6_1      1.000     1.000     1.000         8\n",
            "   F_maj_6_1      1.000     1.000     1.000         7\n",
            "  Fs_maj_4_1      1.000     1.000     1.000         9\n",
            "  Fs_dim_6_0      1.000     1.000     1.000         9\n",
            "  Fs_maj_7_1      1.000     1.000     1.000         8\n",
            "  Fs_maj_4_0      1.000     1.000     1.000         6\n",
            "  Fs_min_5_0      1.000     1.000     1.000         7\n",
            "  Fs_maj_5_0      1.000     1.000     1.000         9\n",
            "  Fs_dim_4_1      1.000     1.000     1.000         7\n",
            "  Fs_dim_4_0      1.000     1.000     1.000         9\n",
            "  Fs_maj_3_1      1.000     1.000     1.000         6\n",
            "  Fs_maj_7_0      1.000     1.000     1.000         7\n",
            "  Fs_dim_5_0      1.000     1.000     1.000         9\n",
            "  Fs_maj_6_1      1.000     1.000     1.000         9\n",
            "   F_min_7_0      1.000     1.000     1.000         9\n",
            "  Fs_min_6_1      1.000     1.000     1.000         5\n",
            "  Fs_min_2_0      1.000     1.000     1.000        10\n",
            "  Fs_min_3_1      1.000     1.000     1.000         8\n",
            "  Fs_min_5_1      1.000     1.000     1.000         8\n",
            "  Fs_dim_7_1      1.000     1.000     1.000         9\n",
            "  Fs_maj_2_0      1.000     1.000     1.000         9\n",
            "  Fs_min_2_1      1.000     1.000     1.000         8\n",
            "  Fs_min_6_0      1.000     1.000     1.000         7\n",
            "   F_min_7_1      1.000     1.000     1.000         8\n",
            "  Fs_min_3_0      1.000     1.000     1.000         8\n",
            "  Fs_dim_7_0      1.000     1.000     1.000         9\n",
            "  Fs_dim_2_0      1.000     1.000     1.000         9\n",
            "  Fs_min_4_0      1.000     1.000     1.000        10\n",
            "  Fs_dim_3_1      1.000     1.000     1.000         7\n",
            "  Fs_maj_6_0      1.000     1.000     1.000         9\n",
            "  Fs_maj_2_1      1.000     1.000     1.000         8\n",
            "  Fs_dim_6_1      1.000     1.000     1.000         6\n",
            "  Fs_dim_5_1      1.000     1.000     1.000        10\n",
            "  Fs_min_7_0      1.000     1.000     1.000         8\n",
            "  Fs_min_4_1      1.000     1.000     1.000         8\n",
            "  Fs_maj_5_1      1.000     0.875     0.933         8\n",
            "  Fs_dim_3_0      1.000     1.000     1.000         9\n",
            "  Fs_dim_2_1      1.000     1.000     1.000         8\n",
            "  Fs_maj_3_0      1.000     1.000     1.000         7\n",
            "   G_maj_6_0      1.000     1.000     1.000         7\n",
            "   G_maj_5_0      1.000     1.000     1.000         8\n",
            "   G_min_3_1      1.000     1.000     1.000         8\n",
            "   G_dim_5_1      1.000     0.875     0.933         8\n",
            "   G_dim_4_0      1.000     1.000     1.000         9\n",
            "   G_maj_2_1      1.000     1.000     1.000         9\n",
            "   G_min_5_0      1.000     1.000     1.000         6\n",
            "   G_min_4_0      1.000     1.000     1.000         6\n",
            "   G_min_2_0      1.000     1.000     1.000         8\n",
            "   G_maj_6_1      1.000     1.000     1.000         9\n",
            "   G_maj_3_0      1.000     1.000     1.000        10\n",
            "   G_dim_7_1      1.000     1.000     1.000         8\n",
            "   G_maj_7_1      1.000     1.000     1.000         7\n",
            "   G_maj_2_0      1.000     1.000     1.000         9\n",
            "   G_min_7_0      1.000     1.000     1.000         9\n",
            "   G_min_2_1      1.000     1.000     1.000         8\n",
            "   G_min_5_1      1.000     0.875     0.933         8\n",
            "   G_dim_2_0      1.000     1.000     1.000         7\n",
            "   G_dim_3_0      1.000     1.000     1.000         8\n",
            "   G_dim_6_1      1.000     1.000     1.000         9\n",
            "  Fs_min_7_1      1.000     1.000     1.000         5\n",
            "   G_dim_3_1      1.000     1.000     1.000         7\n",
            "   G_dim_7_0      1.000     1.000     1.000         8\n",
            "   G_min_3_0      1.000     1.000     1.000         9\n",
            "   G_dim_5_0      1.000     1.000     1.000         7\n",
            "   G_dim_6_0      1.000     1.000     1.000         9\n",
            "   G_dim_2_1      1.000     1.000     1.000         8\n",
            "   G_min_6_1      1.000     1.000     1.000         4\n",
            "   G_maj_5_1      1.000     0.875     0.933         8\n",
            "   G_min_4_1      1.000     1.000     1.000         9\n",
            "   G_min_6_0      1.000     1.000     1.000         7\n",
            "   G_maj_3_1      1.000     1.000     1.000         8\n",
            "   G_dim_4_1      1.000     1.000     1.000         9\n",
            "   G_maj_4_0      1.000     1.000     1.000         8\n",
            "   G_maj_7_0      1.000     1.000     1.000         7\n",
            "   G_maj_4_1      1.000     1.000     1.000        10\n",
            "  Gs_maj_3_1      1.000     1.000     1.000         9\n",
            "  Gs_maj_6_0      1.000     1.000     1.000         8\n",
            "  Gs_min_4_0      1.000     1.000     1.000         8\n",
            "  Gs_dim_7_0      1.000     1.000     1.000         8\n",
            "  Gs_min_3_1      1.000     1.000     1.000         8\n",
            "  Gs_maj_3_0      1.000     1.000     1.000         8\n",
            "  Gs_dim_4_1      1.000     1.000     1.000         9\n",
            "  Gs_min_5_0      0.900     1.000     0.947         9\n",
            "  Gs_maj_6_1      1.000     1.000     1.000        10\n",
            "  Gs_dim_6_1      1.000     1.000     1.000        10\n",
            "  Gs_maj_7_1      1.000     1.000     1.000         8\n",
            "  Gs_dim_2_0      1.000     1.000     1.000         8\n",
            "  Gs_dim_2_1      1.000     1.000     1.000         7\n",
            "  Gs_dim_6_0      1.000     1.000     1.000         8\n",
            "  Gs_min_7_0      1.000     1.000     1.000         8\n",
            "  Gs_min_7_1      1.000     1.000     1.000        10\n",
            "  Gs_dim_3_0      1.000     1.000     1.000         9\n",
            "  Gs_dim_4_0      1.000     1.000     1.000         8\n",
            "  Gs_maj_4_1      1.000     1.000     1.000         9\n",
            "  Gs_dim_5_0      1.000     1.000     1.000         8\n",
            "  Gs_min_4_1      1.000     1.000     1.000         9\n",
            "  Gs_min_5_1      1.000     1.000     1.000         7\n",
            "  Gs_maj_5_0      1.000     1.000     1.000        10\n",
            "  Gs_min_3_0      1.000     1.000     1.000        10\n",
            "  Gs_min_2_0      1.000     1.000     1.000        10\n",
            "  Gs_dim_5_1      0.818     1.000     0.900         9\n",
            "  Gs_maj_4_0      1.000     1.000     1.000         9\n",
            "  Gs_maj_2_0      1.000     1.000     1.000         9\n",
            "  Gs_maj_5_1      0.909     1.000     0.952        10\n",
            "  Gs_maj_7_0      1.000     1.000     1.000         8\n",
            "   G_min_7_1      1.000     1.000     1.000         9\n",
            "  Gs_min_6_1      1.000     1.000     1.000         8\n",
            "  Gs_dim_3_1      1.000     1.000     1.000         6\n",
            "  Gs_min_6_0      1.000     1.000     1.000         7\n",
            "  Gs_dim_7_1      1.000     1.000     1.000         8\n",
            "  Gs_maj_2_1      1.000     1.000     1.000         9\n",
            "  Gs_min_2_1      1.000     1.000     1.000        10\n",
            "\n",
            "   micro avg      0.994     0.994     0.994      3456\n",
            "   macro avg      0.995     0.993     0.993      3456\n",
            "weighted avg      0.995     0.994     0.994      3456\n",
            " samples avg      0.994     0.994     0.994      3456\n",
            "\n",
            "------------------------------ metrics for test set ------------------------------\n",
            "confusion matrix: \n",
            " [[2 0 0 ... 0 0 0]\n",
            " [0 3 0 ... 0 0 0]\n",
            " [0 0 3 ... 0 0 0]\n",
            " ...\n",
            " [0 0 0 ... 2 0 0]\n",
            " [0 0 0 ... 0 2 0]\n",
            " [0 0 0 ... 0 0 1]]\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "   A_dim_2_1      1.000     1.000     1.000         2\n",
            "   A_dim_2_0      1.000     1.000     1.000         3\n",
            "   A_maj_5_1      1.000     1.000     1.000         3\n",
            "   A_maj_3_1      1.000     1.000     1.000         1\n",
            "   A_min_2_0      0.000     0.000     0.000         0\n",
            "   A_maj_2_0      1.000     1.000     1.000         2\n",
            "   A_maj_3_0      0.000     0.000     0.000         0\n",
            "   A_maj_5_0      1.000     1.000     1.000         3\n",
            "   A_dim_3_1      1.000     1.000     1.000         2\n",
            "   A_dim_6_1      0.500     1.000     0.667         1\n",
            "   A_min_2_1      1.000     1.000     1.000         1\n",
            "   A_dim_5_1      0.750     1.000     0.857         3\n",
            "   A_dim_7_0      0.000     0.000     0.000         0\n",
            "   A_dim_5_0      1.000     0.500     0.667         2\n",
            "   A_maj_7_0      0.600     1.000     0.750         3\n",
            "   A_maj_6_0      1.000     1.000     1.000         1\n",
            "   A_maj_4_0      1.000     1.000     1.000         2\n",
            "   A_dim_3_0      1.000     1.000     1.000         3\n",
            "   A_dim_4_1      1.000     1.000     1.000         2\n",
            "   A_maj_6_1      0.500     1.000     0.667         2\n",
            "   A_maj_4_1      1.000     1.000     1.000         1\n",
            "   A_dim_4_0      1.000     1.000     1.000         2\n",
            "   A_maj_2_1      1.000     1.000     1.000         2\n",
            "   A_maj_7_1      1.000     0.500     0.667         2\n",
            "   A_dim_7_1      1.000     1.000     1.000         3\n",
            "   A_dim_6_0      0.500     1.000     0.667         2\n",
            "   B_dim_3_1      1.000     1.000     1.000         3\n",
            "   B_dim_7_0      0.333     1.000     0.500         1\n",
            "   B_maj_6_0      1.000     1.000     1.000         1\n",
            "   B_dim_7_1      1.000     0.500     0.667         4\n",
            "   B_min_2_1      1.000     1.000     1.000         3\n",
            "   B_dim_5_0      1.000     1.000     1.000         3\n",
            "   B_min_2_0      1.000     1.000     1.000         1\n",
            "   B_dim_6_1      1.000     1.000     1.000         5\n",
            "   B_maj_7_1      0.000     0.000     0.000         0\n",
            "   B_dim_4_1      1.000     1.000     1.000         1\n",
            "   B_maj_3_0      1.000     1.000     1.000         1\n",
            "   B_dim_3_0      1.000     1.000     1.000         2\n",
            "   A_min_6_0      1.000     0.750     0.857         4\n",
            "   B_maj_6_1      1.000     1.000     1.000         3\n",
            "   B_dim_5_1      0.875     0.875     0.875         8\n",
            "   A_min_6_1      1.000     1.000     1.000         2\n",
            "   A_min_4_1      1.000     1.000     1.000         3\n",
            "   B_maj_5_1      1.000     1.000     1.000         1\n",
            "   B_maj_2_0      1.000     1.000     1.000         3\n",
            "   B_dim_2_0      1.000     1.000     1.000         1\n",
            "   A_min_7_1      0.000     0.000     0.000         0\n",
            "   A_min_4_0      1.000     1.000     1.000         2\n",
            "   A_min_3_0      1.000     1.000     1.000         1\n",
            "   B_dim_4_0      0.000     0.000     0.000         0\n",
            "   B_maj_4_0      1.000     1.000     1.000         1\n",
            "   A_min_7_0      1.000     0.667     0.800         6\n",
            "   B_maj_4_1      1.000     1.000     1.000         1\n",
            "   A_min_3_1      1.000     1.000     1.000         2\n",
            "   B_maj_2_1      1.000     1.000     1.000         2\n",
            "   A_min_5_0      0.750     1.000     0.857         3\n",
            "   B_maj_3_1      1.000     1.000     1.000         3\n",
            "   B_maj_5_0      1.000     1.000     1.000         3\n",
            "   B_dim_2_1      1.000     1.000     1.000         1\n",
            "   B_maj_7_0      0.500     1.000     0.667         1\n",
            "   B_dim_6_0      0.500     1.000     0.667         1\n",
            "   B_min_3_0      1.000     1.000     1.000         2\n",
            "   A_min_5_1      1.000     1.000     1.000         3\n",
            "  Bb_dim_4_1      1.000     1.000     1.000         4\n",
            "  Bb_maj_7_0      1.000     1.000     1.000         2\n",
            "   B_min_5_1      1.000     1.000     1.000         2\n",
            "   B_min_4_0      1.000     1.000     1.000         3\n",
            "  Bb_maj_4_1      1.000     1.000     1.000         1\n",
            "   B_min_7_0      0.750     0.750     0.750         4\n",
            "  Bb_min_2_0      1.000     1.000     1.000         4\n",
            "  Bb_dim_3_0      1.000     1.000     1.000         3\n",
            "  Bb_dim_3_1      1.000     1.000     1.000         2\n",
            "   B_min_3_1      1.000     1.000     1.000         2\n",
            "  Bb_dim_7_1      1.000     1.000     1.000         2\n",
            "  Bb_maj_3_1      1.000     1.000     1.000         3\n",
            "   B_min_5_0      1.000     0.750     0.857         4\n",
            "   B_min_4_1      1.000     1.000     1.000         3\n",
            "   B_min_7_1      0.750     1.000     0.857         3\n",
            "  Bb_dim_5_1      1.000     1.000     1.000         2\n",
            "  Bb_maj_5_0      1.000     1.000     1.000         3\n",
            "  Bb_dim_4_0      1.000     1.000     1.000         2\n",
            "   B_min_6_0      1.000     1.000     1.000         2\n",
            "  Bb_maj_5_1      1.000     0.333     0.500         3\n",
            "  Bb_min_3_1      1.000     1.000     1.000         3\n",
            "  Bb_maj_3_0      0.000     0.000     0.000         0\n",
            "  Bb_dim_6_1      1.000     1.000     1.000         1\n",
            "  Bb_min_2_1      0.000     0.000     0.000         0\n",
            "  Bb_maj_6_1      1.000     1.000     1.000         1\n",
            "  Bb_min_4_0      1.000     0.667     0.800         3\n",
            "  Bb_maj_2_0      1.000     1.000     1.000         3\n",
            "  Bb_dim_2_1      1.000     1.000     1.000         2\n",
            "  Bb_maj_2_1      1.000     1.000     1.000         1\n",
            "  Bb_maj_7_1      0.000     0.000     0.000         0\n",
            "   B_min_6_1      1.000     1.000     1.000         1\n",
            "  Bb_dim_6_0      1.000     1.000     1.000         3\n",
            "  Bb_dim_7_0      1.000     1.000     1.000         2\n",
            "  Bb_maj_6_0      1.000     1.000     1.000         1\n",
            "  Bb_dim_2_0      0.000     0.000     0.000         0\n",
            "  Bb_min_3_0      1.000     1.000     1.000         2\n",
            "  Bb_dim_5_0      0.000     0.000     0.000         0\n",
            "  Bb_maj_4_0      1.000     1.000     1.000         1\n",
            "   C_dim_4_0      1.000     1.000     1.000         3\n",
            "   C_maj_2_0      1.000     1.000     1.000         3\n",
            "   C_dim_4_1      1.000     1.000     1.000         2\n",
            "   C_dim_6_0      1.000     0.800     0.889         5\n",
            "   C_min_2_0      0.000     0.000     0.000         0\n",
            "   C_maj_6_1      1.000     0.500     0.667         2\n",
            "   C_maj_4_0      1.000     1.000     1.000         4\n",
            "   C_min_2_1      1.000     1.000     1.000         1\n",
            "   C_maj_7_1      1.000     1.000     1.000         3\n",
            "   C_dim_7_0      0.667     1.000     0.800         2\n",
            "   C_min_3_1      1.000     1.000     1.000         6\n",
            "   C_maj_2_1      1.000     1.000     1.000         1\n",
            "  Bb_min_6_0      1.000     0.500     0.667         2\n",
            "   C_maj_5_1      1.000     1.000     1.000         3\n",
            "   C_dim_3_1      0.000     0.000     0.000         0\n",
            "  Bb_min_7_1      1.000     1.000     1.000         1\n",
            "  Bb_min_5_1      1.000     0.750     0.857         4\n",
            "   C_maj_3_1      1.000     1.000     1.000         3\n",
            "   C_maj_3_0      0.200     1.000     0.333         1\n",
            "  Bb_min_4_1      1.000     1.000     1.000         3\n",
            "   C_dim_2_1      0.000     0.000     0.000         0\n",
            "   C_dim_3_0      1.000     1.000     1.000         2\n",
            "   C_dim_2_0      1.000     1.000     1.000         1\n",
            "   C_maj_5_0      1.000     1.000     1.000         3\n",
            "   C_dim_5_1      1.000     1.000     1.000         3\n",
            "  Bb_min_6_1      0.000     0.000     0.000         0\n",
            "   C_maj_7_0      1.000     0.500     0.667         2\n",
            "   C_dim_5_0      1.000     1.000     1.000         2\n",
            "  Bb_min_5_0      1.000     1.000     1.000         4\n",
            "   C_maj_6_0      1.000     0.667     0.800         3\n",
            "   C_dim_6_1      1.000     1.000     1.000         3\n",
            "   C_min_4_1      1.000     1.000     1.000         2\n",
            "  Bb_min_7_0      1.000     1.000     1.000         3\n",
            "   C_min_3_0      0.000     0.000     0.000         0\n",
            "   C_maj_4_1      1.000     1.000     1.000         2\n",
            "   C_dim_7_1      1.000     1.000     1.000         3\n",
            "   C_min_4_0      1.000     1.000     1.000         2\n",
            "  Cs_maj_3_0      1.000     1.000     1.000         2\n",
            "  Cs_maj_4_0      1.000     1.000     1.000         2\n",
            "  Cs_maj_2_1      1.000     1.000     1.000         2\n",
            "  Cs_dim_6_0      1.000     1.000     1.000         3\n",
            "   C_min_5_1      1.000     1.000     1.000         1\n",
            "  Cs_min_2_0      1.000     1.000     1.000         1\n",
            "  Cs_maj_3_1      1.000     1.000     1.000         5\n",
            "  Cs_dim_3_1      0.000     0.000     0.000         0\n",
            "  Cs_min_3_0      1.000     1.000     1.000         3\n",
            "  Cs_dim_4_0      1.000     1.000     1.000         3\n",
            "  Cs_maj_2_0      1.000     1.000     1.000         1\n",
            "  Cs_dim_6_1      0.500     1.000     0.667         3\n",
            "  Cs_maj_7_0      1.000     1.000     1.000         3\n",
            "  Cs_dim_5_1      1.000     1.000     1.000         1\n",
            "  Cs_maj_6_1      1.000     1.000     1.000         3\n",
            "  Cs_dim_7_0      0.000     0.000     0.000         0\n",
            "  Cs_dim_2_0      0.000     0.000     0.000         0\n",
            "   C_min_6_1      1.000     1.000     1.000         1\n",
            "  Cs_maj_5_0      0.000     0.000     0.000         0\n",
            "   C_min_7_1      0.200     1.000     0.333         1\n",
            "  Cs_min_4_1      1.000     1.000     1.000         2\n",
            "  Cs_dim_7_1      1.000     1.000     1.000         2\n",
            "  Cs_dim_5_0      1.000     1.000     1.000         3\n",
            "  Cs_maj_6_0      0.333     1.000     0.500         1\n",
            "  Cs_min_5_0      1.000     1.000     1.000         1\n",
            "   C_min_6_0      1.000     0.500     0.667         2\n",
            "  Cs_maj_5_1      1.000     1.000     1.000         3\n",
            "  Cs_min_3_1      1.000     1.000     1.000         2\n",
            "  Cs_maj_7_1      1.000     1.000     1.000         3\n",
            "  Cs_min_2_1      1.000     1.000     1.000         1\n",
            "  Cs_dim_4_1      1.000     1.000     1.000         3\n",
            "  Cs_dim_3_0      1.000     1.000     1.000         4\n",
            "  Cs_dim_2_1      1.000     1.000     1.000         2\n",
            "  Cs_maj_4_1      1.000     1.000     1.000         3\n",
            "  Cs_min_4_0      0.000     0.000     0.000         0\n",
            "   C_min_7_0      0.500     0.333     0.400         3\n",
            "   C_min_5_0      0.000     0.000     0.000         0\n",
            "   D_dim_3_0      1.000     1.000     1.000         3\n",
            "   D_maj_5_0      1.000     1.000     1.000         2\n",
            "   D_dim_4_0      1.000     1.000     1.000         1\n",
            "  Cs_min_7_0      1.000     1.000     1.000         2\n",
            "   D_maj_7_1      0.667     1.000     0.800         2\n",
            "   D_min_5_0      1.000     1.000     1.000         3\n",
            "   D_dim_3_1      1.000     1.000     1.000         2\n",
            "   D_maj_4_0      1.000     1.000     1.000         4\n",
            "   D_dim_7_1      0.000     0.000     0.000         0\n",
            "   D_min_4_1      1.000     1.000     1.000         5\n",
            "   D_min_4_0      1.000     1.000     1.000         6\n",
            "   D_maj_3_0      0.000     0.000     0.000         4\n",
            "   D_maj_7_0      1.000     0.500     0.667         2\n",
            "   D_min_2_1      1.000     1.000     1.000         2\n",
            "   D_dim_2_0      1.000     1.000     1.000         1\n",
            "   D_maj_6_0      1.000     1.000     1.000         2\n",
            "   D_min_3_0      1.000     1.000     1.000         1\n",
            "   D_maj_5_1      1.000     0.833     0.909         6\n",
            "   D_dim_6_0      0.000     0.000     0.000         0\n",
            "   D_dim_5_1      1.000     1.000     1.000         1\n",
            "   D_maj_2_0      1.000     1.000     1.000         2\n",
            "   D_maj_4_1      1.000     1.000     1.000         1\n",
            "  Cs_min_5_1      1.000     1.000     1.000         4\n",
            "   D_maj_2_1      1.000     0.667     0.800         3\n",
            "   D_maj_3_1      1.000     1.000     1.000         3\n",
            "  Cs_min_6_0      1.000     0.750     0.857         4\n",
            "   D_min_2_0      1.000     1.000     1.000         1\n",
            "   D_min_5_1      1.000     1.000     1.000         3\n",
            "   D_dim_4_1      1.000     1.000     1.000         5\n",
            "   D_dim_6_1      1.000     1.000     1.000         2\n",
            "   D_dim_2_1      1.000     1.000     1.000         2\n",
            "   D_dim_5_0      1.000     1.000     1.000         2\n",
            "   D_dim_7_0      1.000     0.500     0.667         2\n",
            "  Cs_min_6_1      0.000     0.000     0.000         0\n",
            "   D_maj_6_1      1.000     1.000     1.000         1\n",
            "   D_min_3_1      0.000     0.000     0.000         0\n",
            "  Cs_min_7_1      1.000     1.000     1.000         2\n",
            "   E_maj_2_1      1.000     1.000     1.000         2\n",
            "   E_dim_6_1      1.000     1.000     1.000         3\n",
            "   D_min_7_0      1.000     1.000     1.000         1\n",
            "   D_min_6_0      0.000     0.000     0.000         0\n",
            "   E_min_3_0      1.000     1.000     1.000         2\n",
            "   D_min_6_1      0.750     1.000     0.857         3\n",
            "   E_dim_7_1      1.000     0.750     0.857         4\n",
            "   E_min_2_1      1.000     1.000     1.000         2\n",
            "   E_dim_2_1      1.000     1.000     1.000         3\n",
            "   E_dim_5_1      1.000     1.000     1.000         2\n",
            "   E_dim_7_0      1.000     1.000     1.000         2\n",
            "   E_dim_3_0      1.000     1.000     1.000         3\n",
            "   E_dim_4_0      1.000     1.000     1.000         4\n",
            "   E_min_2_0      1.000     1.000     1.000         2\n",
            "   E_dim_3_1      1.000     1.000     1.000         2\n",
            "   E_maj_4_0      0.000     0.000     0.000         0\n",
            "   E_min_3_1      1.000     1.000     1.000         1\n",
            "   E_maj_5_0      1.000     1.000     1.000         2\n",
            "   E_maj_7_1      1.000     1.000     1.000         2\n",
            "   E_min_4_1      1.000     1.000     1.000         4\n",
            "   E_min_4_0      0.000     0.000     0.000         0\n",
            "   E_min_6_0      0.000     0.000     0.000         1\n",
            "   E_maj_4_1      1.000     1.000     1.000         2\n",
            "   E_maj_7_0      1.000     1.000     1.000         4\n",
            "   E_dim_4_1      1.000     1.000     1.000         1\n",
            "   E_maj_6_0      1.000     1.000     1.000         1\n",
            "   E_maj_3_1      1.000     1.000     1.000         3\n",
            "   E_dim_5_0      0.000     0.000     0.000         0\n",
            "   E_maj_2_0      1.000     1.000     1.000         3\n",
            "   E_min_5_0      1.000     1.000     1.000         1\n",
            "   E_min_5_1      1.000     1.000     1.000         1\n",
            "   E_dim_2_0      1.000     1.000     1.000         4\n",
            "   E_maj_6_1      1.000     0.500     0.667         4\n",
            "   E_maj_3_0      1.000     1.000     1.000         5\n",
            "   E_maj_5_1      0.000     0.000     0.000         0\n",
            "   E_dim_6_0      1.000     1.000     1.000         2\n",
            "   D_min_7_1      1.000     0.500     0.667         2\n",
            "  Eb_min_4_1      1.000     1.000     1.000         3\n",
            "  Eb_dim_6_1      1.000     1.000     1.000         3\n",
            "  Eb_maj_7_0      1.000     0.500     0.667         2\n",
            "  Eb_min_3_1      1.000     1.000     1.000         1\n",
            "  Eb_min_2_0      1.000     1.000     1.000         1\n",
            "  Eb_maj_5_0      1.000     1.000     1.000         1\n",
            "  Eb_maj_4_0      0.000     0.000     0.000         0\n",
            "  Eb_maj_5_1      1.000     1.000     1.000         2\n",
            "  Eb_dim_5_1      0.500     1.000     0.667         1\n",
            "  Eb_min_5_0      1.000     1.000     1.000         2\n",
            "  Eb_maj_3_0      1.000     1.000     1.000         2\n",
            "  Eb_dim_2_1      1.000     1.000     1.000         1\n",
            "  Eb_maj_3_1      1.000     1.000     1.000         2\n",
            "  Eb_dim_7_0      1.000     1.000     1.000         2\n",
            "  Eb_dim_3_0      1.000     1.000     1.000         2\n",
            "  Eb_dim_4_0      1.000     1.000     1.000         4\n",
            "  Eb_min_5_1      0.000     0.000     0.000         0\n",
            "   E_min_7_0      1.000     1.000     1.000         1\n",
            "  Eb_maj_6_0      1.000     0.800     0.889         5\n",
            "  Eb_maj_2_0      1.000     1.000     1.000         4\n",
            "   E_min_6_1      1.000     1.000     1.000         2\n",
            "   E_min_7_1      0.000     0.000     0.000         0\n",
            "  Eb_dim_2_0      1.000     0.667     0.800         3\n",
            "  Eb_maj_2_1      0.000     0.000     0.000         1\n",
            "  Eb_maj_4_1      1.000     1.000     1.000         1\n",
            "  Eb_min_6_0      1.000     1.000     1.000         4\n",
            "  Eb_maj_7_1      1.000     1.000     1.000         1\n",
            "  Eb_dim_4_1      1.000     1.000     1.000         2\n",
            "  Eb_dim_5_0      1.000     1.000     1.000         6\n",
            "  Eb_min_4_0      1.000     1.000     1.000         3\n",
            "  Eb_min_3_0      1.000     1.000     1.000         3\n",
            "  Eb_min_2_1      0.000     0.000     0.000         0\n",
            "  Eb_dim_7_1      1.000     0.667     0.800         3\n",
            "  Eb_dim_3_1      1.000     1.000     1.000         2\n",
            "  Eb_maj_6_1      0.000     0.000     0.000         0\n",
            "  Eb_dim_6_0      1.000     1.000     1.000         1\n",
            "   F_maj_4_0      1.000     1.000     1.000         1\n",
            "   F_dim_7_1      1.000     1.000     1.000         2\n",
            "   F_dim_2_1      1.000     1.000     1.000         1\n",
            "   F_dim_4_0      1.000     1.000     1.000         2\n",
            "   F_dim_5_1      1.000     1.000     1.000         2\n",
            "   F_dim_7_0      1.000     1.000     1.000         1\n",
            "   F_maj_7_0      0.500     1.000     0.667         1\n",
            "   F_min_4_0      1.000     1.000     1.000         1\n",
            "  Eb_min_7_0      1.000     1.000     1.000         4\n",
            "   F_dim_2_0      1.000     1.000     1.000         2\n",
            "   F_maj_2_0      1.000     1.000     1.000         5\n",
            "   F_maj_6_0      1.000     0.800     0.889         5\n",
            "   F_min_6_1      1.000     1.000     1.000         2\n",
            "   F_min_5_0      1.000     1.000     1.000         2\n",
            "   F_min_2_1      1.000     1.000     1.000         1\n",
            "   F_dim_3_1      0.000     0.000     0.000         0\n",
            "   F_min_6_0      1.000     1.000     1.000         2\n",
            "   F_dim_3_0      1.000     1.000     1.000         2\n",
            "   F_maj_3_1      1.000     1.000     1.000         2\n",
            "   F_dim_6_1      0.000     0.000     0.000         1\n",
            "   F_maj_5_0      1.000     1.000     1.000         1\n",
            "   F_min_4_1      1.000     1.000     1.000         3\n",
            "  Eb_min_7_1      1.000     1.000     1.000         1\n",
            "   F_min_3_0      1.000     1.000     1.000         2\n",
            "   F_min_3_1      1.000     1.000     1.000         2\n",
            "   F_maj_3_0      1.000     1.000     1.000         1\n",
            "   F_min_2_0      1.000     1.000     1.000         3\n",
            "   F_maj_4_1      1.000     1.000     1.000         1\n",
            "   F_dim_4_1      1.000     1.000     1.000         2\n",
            "   F_dim_5_0      1.000     1.000     1.000         2\n",
            "   F_dim_6_0      1.000     1.000     1.000         2\n",
            "   F_min_5_1      1.000     1.000     1.000         1\n",
            "   F_maj_5_1      1.000     0.500     0.667         2\n",
            "   F_maj_7_1      1.000     1.000     1.000         1\n",
            "   F_maj_2_1      1.000     1.000     1.000         1\n",
            "  Eb_min_6_1      1.000     1.000     1.000         2\n",
            "   F_maj_6_1      1.000     1.000     1.000         3\n",
            "  Fs_maj_4_1      1.000     1.000     1.000         1\n",
            "  Fs_dim_6_0      1.000     1.000     1.000         1\n",
            "  Fs_maj_7_1      1.000     1.000     1.000         2\n",
            "  Fs_maj_4_0      1.000     0.750     0.857         4\n",
            "  Fs_min_5_0      1.000     1.000     1.000         3\n",
            "  Fs_maj_5_0      1.000     1.000     1.000         1\n",
            "  Fs_dim_4_1      1.000     1.000     1.000         3\n",
            "  Fs_dim_4_0      1.000     1.000     1.000         1\n",
            "  Fs_maj_3_1      1.000     1.000     1.000         4\n",
            "  Fs_maj_7_0      1.000     1.000     1.000         3\n",
            "  Fs_dim_5_0      1.000     1.000     1.000         1\n",
            "  Fs_maj_6_1      1.000     1.000     1.000         1\n",
            "   F_min_7_0      1.000     1.000     1.000         1\n",
            "  Fs_min_6_1      1.000     0.800     0.889         5\n",
            "  Fs_min_2_0      0.000     0.000     0.000         0\n",
            "  Fs_min_3_1      1.000     1.000     1.000         2\n",
            "  Fs_min_5_1      1.000     1.000     1.000         2\n",
            "  Fs_dim_7_1      0.000     0.000     0.000         1\n",
            "  Fs_maj_2_0      1.000     1.000     1.000         1\n",
            "  Fs_min_2_1      1.000     1.000     1.000         2\n",
            "  Fs_min_6_0      1.000     1.000     1.000         3\n",
            "   F_min_7_1      1.000     1.000     1.000         2\n",
            "  Fs_min_3_0      1.000     1.000     1.000         2\n",
            "  Fs_dim_7_0      1.000     1.000     1.000         1\n",
            "  Fs_dim_2_0      0.500     1.000     0.667         1\n",
            "  Fs_min_4_0      0.000     0.000     0.000         0\n",
            "  Fs_dim_3_1      1.000     1.000     1.000         3\n",
            "  Fs_maj_6_0      1.000     1.000     1.000         1\n",
            "  Fs_maj_2_1      0.667     1.000     0.800         2\n",
            "  Fs_dim_6_1      1.000     1.000     1.000         4\n",
            "  Fs_dim_5_1      0.000     0.000     0.000         0\n",
            "  Fs_min_7_0      0.667     1.000     0.800         2\n",
            "  Fs_min_4_1      1.000     0.500     0.667         2\n",
            "  Fs_maj_5_1      1.000     1.000     1.000         2\n",
            "  Fs_dim_3_0      1.000     1.000     1.000         1\n",
            "  Fs_dim_2_1      1.000     1.000     1.000         2\n",
            "  Fs_maj_3_0      1.000     1.000     1.000         3\n",
            "   G_maj_6_0      1.000     0.667     0.800         3\n",
            "   G_maj_5_0      1.000     1.000     1.000         2\n",
            "   G_min_3_1      1.000     1.000     1.000         2\n",
            "   G_dim_5_1      1.000     1.000     1.000         2\n",
            "   G_dim_4_0      1.000     1.000     1.000         1\n",
            "   G_maj_2_1      1.000     1.000     1.000         1\n",
            "   G_min_5_0      1.000     1.000     1.000         4\n",
            "   G_min_4_0      1.000     1.000     1.000         4\n",
            "   G_min_2_0      1.000     1.000     1.000         2\n",
            "   G_maj_6_1      0.000     0.000     0.000         1\n",
            "   G_maj_3_0      0.000     0.000     0.000         0\n",
            "   G_dim_7_1      1.000     1.000     1.000         2\n",
            "   G_maj_7_1      1.000     0.667     0.800         3\n",
            "   G_maj_2_0      1.000     1.000     1.000         1\n",
            "   G_min_7_0      0.500     1.000     0.667         1\n",
            "   G_min_2_1      1.000     1.000     1.000         2\n",
            "   G_min_5_1      1.000     1.000     1.000         2\n",
            "   G_dim_2_0      1.000     1.000     1.000         3\n",
            "   G_dim_3_0      1.000     1.000     1.000         2\n",
            "   G_dim_6_1      1.000     1.000     1.000         1\n",
            "  Fs_min_7_1      1.000     0.800     0.889         5\n",
            "   G_dim_3_1      1.000     1.000     1.000         3\n",
            "   G_dim_7_0      1.000     1.000     1.000         2\n",
            "   G_min_3_0      1.000     1.000     1.000         1\n",
            "   G_dim_5_0      1.000     1.000     1.000         3\n",
            "   G_dim_6_0      0.500     1.000     0.667         1\n",
            "   G_dim_2_1      1.000     1.000     1.000         2\n",
            "   G_min_6_1      1.000     1.000     1.000         6\n",
            "   G_maj_5_1      1.000     1.000     1.000         2\n",
            "   G_min_4_1      1.000     1.000     1.000         1\n",
            "   G_min_6_0      1.000     1.000     1.000         3\n",
            "   G_maj_3_1      1.000     1.000     1.000         2\n",
            "   G_dim_4_1      1.000     1.000     1.000         1\n",
            "   G_maj_4_0      1.000     1.000     1.000         2\n",
            "   G_maj_7_0      1.000     0.333     0.500         3\n",
            "   G_maj_4_1      0.000     0.000     0.000         0\n",
            "  Gs_maj_3_1      1.000     1.000     1.000         1\n",
            "  Gs_maj_6_0      1.000     0.500     0.667         2\n",
            "  Gs_min_4_0      1.000     1.000     1.000         2\n",
            "  Gs_dim_7_0      1.000     1.000     1.000         2\n",
            "  Gs_min_3_1      1.000     1.000     1.000         2\n",
            "  Gs_maj_3_0      1.000     1.000     1.000         2\n",
            "  Gs_dim_4_1      1.000     1.000     1.000         1\n",
            "  Gs_min_5_0      1.000     1.000     1.000         1\n",
            "  Gs_maj_6_1      0.000     0.000     0.000         0\n",
            "  Gs_dim_6_1      0.000     0.000     0.000         0\n",
            "  Gs_maj_7_1      1.000     1.000     1.000         2\n",
            "  Gs_dim_2_0      1.000     1.000     1.000         2\n",
            "  Gs_dim_2_1      1.000     1.000     1.000         3\n",
            "  Gs_dim_6_0      1.000     0.500     0.667         2\n",
            "  Gs_min_7_0      1.000     1.000     1.000         2\n",
            "  Gs_min_7_1      0.000     0.000     0.000         0\n",
            "  Gs_dim_3_0      1.000     1.000     1.000         1\n",
            "  Gs_dim_4_0      1.000     1.000     1.000         2\n",
            "  Gs_maj_4_1      0.500     1.000     0.667         1\n",
            "  Gs_dim_5_0      1.000     1.000     1.000         2\n",
            "  Gs_min_4_1      1.000     1.000     1.000         1\n",
            "  Gs_min_5_1      1.000     1.000     1.000         3\n",
            "  Gs_maj_5_0      0.000     0.000     0.000         0\n",
            "  Gs_min_3_0      0.000     0.000     0.000         0\n",
            "  Gs_min_2_0      0.000     0.000     0.000         0\n",
            "  Gs_dim_5_1      0.500     1.000     0.667         1\n",
            "  Gs_maj_4_0      0.000     0.000     0.000         1\n",
            "  Gs_maj_2_0      1.000     1.000     1.000         1\n",
            "  Gs_maj_5_1      0.000     0.000     0.000         0\n",
            "  Gs_maj_7_0      1.000     1.000     1.000         2\n",
            "   G_min_7_1      1.000     1.000     1.000         1\n",
            "  Gs_min_6_1      1.000     1.000     1.000         2\n",
            "  Gs_dim_3_1      1.000     1.000     1.000         4\n",
            "  Gs_min_6_0      0.667     0.667     0.667         3\n",
            "  Gs_dim_7_1      0.667     1.000     0.800         2\n",
            "  Gs_maj_2_1      1.000     1.000     1.000         1\n",
            "  Gs_min_2_1      0.000     0.000     0.000         0\n",
            "\n",
            "   micro avg      0.933     0.933     0.933       864\n",
            "   macro avg      0.836     0.831     0.824       864\n",
            "weighted avg      0.961     0.933     0.937       864\n",
            " samples avg      0.933     0.933     0.933       864\n",
            "\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1318: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1318: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "with tf.device('/GPU:0'):\n",
        "  NR_history = NR_model.fit(X_train, y_train, validation_data=(X_test, y_test), batch_size=64, epochs=100, verbose=2, callbacks=[checkpoint]).history\n",
        "  NR_model = load_model(filepath)\n",
        "  outputResult(NR_model, NR_history, X_train, y_train, X_test, y_test, noteDic)\n",
        "  showPlots(NR_model, NR_history)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "with tf.device('/GPU:0'):\n",
        "  NR_history = NR_model.fit(X_NR, y_true_NR_aug, batch_size=64, epochs=100, verbose=2, callbacks=[checkpoint]).history\n",
        "  NR_model = load_model(filepath)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CFPGW1Zr5qOq",
        "outputId": "9968d96e-c035-41e1-9f01-3a7629f1c078"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "68/68 - 3s - loss: 6.1153 - accuracy: 0.0060 - 3s/epoch - 42ms/step\n",
            "Epoch 2/100\n",
            "68/68 - 4s - loss: 5.6587 - accuracy: 0.0799 - 4s/epoch - 53ms/step\n",
            "Epoch 3/100\n",
            "68/68 - 2s - loss: 4.9076 - accuracy: 0.2632 - 2s/epoch - 34ms/step\n",
            "Epoch 4/100\n",
            "68/68 - 2s - loss: 4.0415 - accuracy: 0.4757 - 2s/epoch - 27ms/step\n",
            "Epoch 5/100\n",
            "68/68 - 2s - loss: 3.2564 - accuracy: 0.6086 - 2s/epoch - 28ms/step\n",
            "Epoch 6/100\n",
            "68/68 - 2s - loss: 2.6052 - accuracy: 0.7104 - 2s/epoch - 29ms/step\n",
            "Epoch 7/100\n",
            "68/68 - 2s - loss: 2.0947 - accuracy: 0.7713 - 2s/epoch - 28ms/step\n",
            "Epoch 8/100\n",
            "68/68 - 2s - loss: 1.7116 - accuracy: 0.8201 - 2s/epoch - 28ms/step\n",
            "Epoch 9/100\n",
            "68/68 - 2s - loss: 1.4137 - accuracy: 0.8463 - 2s/epoch - 27ms/step\n",
            "Epoch 10/100\n",
            "68/68 - 2s - loss: 1.1889 - accuracy: 0.8713 - 2s/epoch - 27ms/step\n",
            "Epoch 11/100\n",
            "68/68 - 2s - loss: 1.0140 - accuracy: 0.8838 - 2s/epoch - 28ms/step\n",
            "Epoch 12/100\n",
            "68/68 - 2s - loss: 0.8689 - accuracy: 0.9044 - 2s/epoch - 27ms/step\n",
            "Epoch 13/100\n",
            "68/68 - 2s - loss: 0.7538 - accuracy: 0.9162 - 2s/epoch - 27ms/step\n",
            "Epoch 14/100\n",
            "68/68 - 2s - loss: 0.6614 - accuracy: 0.9236 - 2s/epoch - 27ms/step\n",
            "Epoch 15/100\n",
            "68/68 - 2s - loss: 0.5806 - accuracy: 0.9336 - 2s/epoch - 27ms/step\n",
            "Epoch 16/100\n",
            "68/68 - 2s - loss: 0.5150 - accuracy: 0.9407 - 2s/epoch - 28ms/step\n",
            "Epoch 17/100\n",
            "68/68 - 2s - loss: 0.4571 - accuracy: 0.9458 - 2s/epoch - 28ms/step\n",
            "Epoch 18/100\n",
            "68/68 - 2s - loss: 0.4105 - accuracy: 0.9495 - 2s/epoch - 27ms/step\n",
            "Epoch 19/100\n",
            "68/68 - 2s - loss: 0.3691 - accuracy: 0.9549 - 2s/epoch - 27ms/step\n",
            "Epoch 20/100\n",
            "68/68 - 2s - loss: 0.3328 - accuracy: 0.9563 - 2s/epoch - 27ms/step\n",
            "Epoch 21/100\n",
            "68/68 - 2s - loss: 0.3052 - accuracy: 0.9641 - 2s/epoch - 27ms/step\n",
            "Epoch 22/100\n",
            "68/68 - 2s - loss: 0.2788 - accuracy: 0.9653 - 2s/epoch - 27ms/step\n",
            "Epoch 23/100\n",
            "68/68 - 2s - loss: 0.2539 - accuracy: 0.9676 - 2s/epoch - 28ms/step\n",
            "Epoch 24/100\n",
            "68/68 - 2s - loss: 0.2324 - accuracy: 0.9701 - 2s/epoch - 27ms/step\n",
            "Epoch 25/100\n",
            "68/68 - 2s - loss: 0.2127 - accuracy: 0.9738 - 2s/epoch - 28ms/step\n",
            "Epoch 26/100\n",
            "68/68 - 2s - loss: 0.1967 - accuracy: 0.9750 - 2s/epoch - 27ms/step\n",
            "Epoch 27/100\n",
            "68/68 - 2s - loss: 0.1834 - accuracy: 0.9755 - 2s/epoch - 27ms/step\n",
            "Epoch 28/100\n",
            "68/68 - 2s - loss: 0.1704 - accuracy: 0.9766 - 2s/epoch - 27ms/step\n",
            "Epoch 29/100\n",
            "68/68 - 2s - loss: 0.1571 - accuracy: 0.9799 - 2s/epoch - 27ms/step\n",
            "Epoch 30/100\n",
            "68/68 - 2s - loss: 0.1513 - accuracy: 0.9794 - 2s/epoch - 26ms/step\n",
            "Epoch 31/100\n",
            "68/68 - 2s - loss: 0.1405 - accuracy: 0.9785 - 2s/epoch - 25ms/step\n",
            "Epoch 32/100\n",
            "68/68 - 2s - loss: 0.1328 - accuracy: 0.9789 - 2s/epoch - 26ms/step\n",
            "Epoch 33/100\n",
            "68/68 - 2s - loss: 0.1209 - accuracy: 0.9829 - 2s/epoch - 33ms/step\n",
            "Epoch 34/100\n",
            "68/68 - 2s - loss: 0.1148 - accuracy: 0.9838 - 2s/epoch - 28ms/step\n",
            "Epoch 35/100\n",
            "68/68 - 2s - loss: 0.1094 - accuracy: 0.9836 - 2s/epoch - 26ms/step\n",
            "Epoch 36/100\n",
            "68/68 - 2s - loss: 0.1062 - accuracy: 0.9831 - 2s/epoch - 25ms/step\n",
            "Epoch 37/100\n",
            "68/68 - 2s - loss: 0.1009 - accuracy: 0.9836 - 2s/epoch - 26ms/step\n",
            "Epoch 38/100\n",
            "68/68 - 2s - loss: 0.0971 - accuracy: 0.9831 - 2s/epoch - 27ms/step\n",
            "Epoch 39/100\n",
            "68/68 - 2s - loss: 0.0923 - accuracy: 0.9836 - 2s/epoch - 26ms/step\n",
            "Epoch 40/100\n",
            "68/68 - 2s - loss: 0.0871 - accuracy: 0.9852 - 2s/epoch - 32ms/step\n",
            "Epoch 41/100\n",
            "68/68 - 2s - loss: 0.0859 - accuracy: 0.9843 - 2s/epoch - 26ms/step\n",
            "Epoch 42/100\n",
            "68/68 - 2s - loss: 0.0815 - accuracy: 0.9866 - 2s/epoch - 27ms/step\n",
            "Epoch 43/100\n",
            "68/68 - 2s - loss: 0.0796 - accuracy: 0.9843 - 2s/epoch - 26ms/step\n",
            "Epoch 44/100\n",
            "68/68 - 2s - loss: 0.0773 - accuracy: 0.9856 - 2s/epoch - 26ms/step\n",
            "Epoch 45/100\n",
            "68/68 - 2s - loss: 0.0724 - accuracy: 0.9852 - 2s/epoch - 26ms/step\n",
            "Epoch 46/100\n",
            "68/68 - 2s - loss: 0.0694 - accuracy: 0.9859 - 2s/epoch - 27ms/step\n",
            "Epoch 47/100\n",
            "68/68 - 2s - loss: 0.0609 - accuracy: 0.9889 - 2s/epoch - 32ms/step\n",
            "Epoch 48/100\n",
            "68/68 - 2s - loss: 0.0629 - accuracy: 0.9863 - 2s/epoch - 32ms/step\n",
            "Epoch 49/100\n",
            "68/68 - 3s - loss: 0.0604 - accuracy: 0.9875 - 3s/epoch - 40ms/step\n",
            "Epoch 50/100\n",
            "68/68 - 2s - loss: 0.0606 - accuracy: 0.9873 - 2s/epoch - 35ms/step\n",
            "Epoch 51/100\n",
            "68/68 - 2s - loss: 0.0584 - accuracy: 0.9868 - 2s/epoch - 26ms/step\n",
            "Epoch 52/100\n",
            "68/68 - 2s - loss: 0.0552 - accuracy: 0.9889 - 2s/epoch - 27ms/step\n",
            "Epoch 53/100\n",
            "68/68 - 2s - loss: 0.0518 - accuracy: 0.9903 - 2s/epoch - 32ms/step\n",
            "Epoch 54/100\n",
            "68/68 - 2s - loss: 0.0537 - accuracy: 0.9887 - 2s/epoch - 35ms/step\n",
            "Epoch 55/100\n",
            "68/68 - 2s - loss: 0.0516 - accuracy: 0.9887 - 2s/epoch - 26ms/step\n",
            "Epoch 56/100\n",
            "68/68 - 2s - loss: 0.0487 - accuracy: 0.9912 - 2s/epoch - 28ms/step\n",
            "Epoch 57/100\n",
            "68/68 - 2s - loss: 0.0481 - accuracy: 0.9880 - 2s/epoch - 26ms/step\n",
            "Epoch 58/100\n",
            "68/68 - 2s - loss: 0.0455 - accuracy: 0.9889 - 2s/epoch - 26ms/step\n",
            "Epoch 59/100\n",
            "68/68 - 2s - loss: 0.0439 - accuracy: 0.9898 - 2s/epoch - 26ms/step\n",
            "Epoch 60/100\n",
            "68/68 - 2s - loss: 0.0438 - accuracy: 0.9894 - 2s/epoch - 26ms/step\n",
            "Epoch 61/100\n",
            "68/68 - 2s - loss: 0.0438 - accuracy: 0.9891 - 2s/epoch - 26ms/step\n",
            "Epoch 62/100\n",
            "68/68 - 2s - loss: 0.0434 - accuracy: 0.9889 - 2s/epoch - 26ms/step\n",
            "Epoch 63/100\n",
            "68/68 - 2s - loss: 0.0433 - accuracy: 0.9884 - 2s/epoch - 26ms/step\n",
            "Epoch 64/100\n",
            "68/68 - 2s - loss: 0.0363 - accuracy: 0.9912 - 2s/epoch - 26ms/step\n",
            "Epoch 65/100\n",
            "68/68 - 2s - loss: 0.0393 - accuracy: 0.9907 - 2s/epoch - 26ms/step\n",
            "Epoch 66/100\n",
            "68/68 - 2s - loss: 0.0408 - accuracy: 0.9894 - 2s/epoch - 25ms/step\n",
            "Epoch 67/100\n",
            "68/68 - 2s - loss: 0.0355 - accuracy: 0.9905 - 2s/epoch - 25ms/step\n",
            "Epoch 68/100\n",
            "68/68 - 2s - loss: 0.0412 - accuracy: 0.9882 - 2s/epoch - 25ms/step\n",
            "Epoch 69/100\n",
            "68/68 - 2s - loss: 0.0437 - accuracy: 0.9898 - 2s/epoch - 26ms/step\n",
            "Epoch 70/100\n",
            "68/68 - 2s - loss: 0.0370 - accuracy: 0.9912 - 2s/epoch - 26ms/step\n",
            "Epoch 71/100\n",
            "68/68 - 2s - loss: 0.0354 - accuracy: 0.9907 - 2s/epoch - 26ms/step\n",
            "Epoch 72/100\n",
            "68/68 - 2s - loss: 0.0341 - accuracy: 0.9917 - 2s/epoch - 31ms/step\n",
            "Epoch 73/100\n",
            "68/68 - 2s - loss: 0.0377 - accuracy: 0.9894 - 2s/epoch - 26ms/step\n",
            "Epoch 74/100\n",
            "68/68 - 2s - loss: 0.0362 - accuracy: 0.9907 - 2s/epoch - 26ms/step\n",
            "Epoch 75/100\n",
            "68/68 - 2s - loss: 0.0338 - accuracy: 0.9907 - 2s/epoch - 26ms/step\n",
            "Epoch 76/100\n",
            "68/68 - 2s - loss: 0.0427 - accuracy: 0.9889 - 2s/epoch - 26ms/step\n",
            "Epoch 77/100\n",
            "68/68 - 2s - loss: 0.0358 - accuracy: 0.9889 - 2s/epoch - 26ms/step\n",
            "Epoch 78/100\n",
            "68/68 - 2s - loss: 0.0373 - accuracy: 0.9912 - 2s/epoch - 26ms/step\n",
            "Epoch 79/100\n",
            "68/68 - 2s - loss: 0.0319 - accuracy: 0.9894 - 2s/epoch - 26ms/step\n",
            "Epoch 80/100\n",
            "68/68 - 2s - loss: 0.0384 - accuracy: 0.9884 - 2s/epoch - 27ms/step\n",
            "Epoch 81/100\n",
            "68/68 - 2s - loss: 0.0308 - accuracy: 0.9896 - 2s/epoch - 26ms/step\n",
            "Epoch 82/100\n",
            "68/68 - 2s - loss: 0.0468 - accuracy: 0.9889 - 2s/epoch - 26ms/step\n",
            "Epoch 83/100\n",
            "68/68 - 2s - loss: 0.0470 - accuracy: 0.9889 - 2s/epoch - 26ms/step\n",
            "Epoch 84/100\n",
            "68/68 - 2s - loss: 0.0392 - accuracy: 0.9898 - 2s/epoch - 26ms/step\n",
            "Epoch 85/100\n",
            "68/68 - 2s - loss: 0.0315 - accuracy: 0.9921 - 2s/epoch - 33ms/step\n",
            "Epoch 86/100\n",
            "68/68 - 2s - loss: 0.0319 - accuracy: 0.9910 - 2s/epoch - 26ms/step\n",
            "Epoch 87/100\n",
            "68/68 - 2s - loss: 0.0300 - accuracy: 0.9917 - 2s/epoch - 26ms/step\n",
            "Epoch 88/100\n",
            "68/68 - 2s - loss: 0.0276 - accuracy: 0.9914 - 2s/epoch - 26ms/step\n",
            "Epoch 89/100\n",
            "68/68 - 2s - loss: 0.0263 - accuracy: 0.9914 - 2s/epoch - 26ms/step\n",
            "Epoch 90/100\n",
            "68/68 - 2s - loss: 0.0243 - accuracy: 0.9912 - 2s/epoch - 26ms/step\n",
            "Epoch 91/100\n",
            "68/68 - 2s - loss: 0.0283 - accuracy: 0.9914 - 2s/epoch - 27ms/step\n",
            "Epoch 92/100\n",
            "68/68 - 2s - loss: 0.0276 - accuracy: 0.9921 - 2s/epoch - 26ms/step\n",
            "Epoch 93/100\n",
            "68/68 - 2s - loss: 0.0245 - accuracy: 0.9917 - 2s/epoch - 26ms/step\n",
            "Epoch 94/100\n",
            "68/68 - 2s - loss: 0.0233 - accuracy: 0.9912 - 2s/epoch - 26ms/step\n",
            "Epoch 95/100\n",
            "68/68 - 2s - loss: 0.0250 - accuracy: 0.9924 - 2s/epoch - 32ms/step\n",
            "Epoch 96/100\n",
            "68/68 - 2s - loss: 0.0236 - accuracy: 0.9914 - 2s/epoch - 26ms/step\n",
            "Epoch 97/100\n",
            "68/68 - 2s - loss: 0.0238 - accuracy: 0.9926 - 2s/epoch - 28ms/step\n",
            "Epoch 98/100\n",
            "68/68 - 2s - loss: 0.0225 - accuracy: 0.9926 - 2s/epoch - 26ms/step\n",
            "Epoch 99/100\n",
            "68/68 - 2s - loss: 0.0287 - accuracy: 0.9896 - 2s/epoch - 26ms/step\n",
            "Epoch 100/100\n",
            "68/68 - 2s - loss: 0.0414 - accuracy: 0.9887 - 2s/epoch - 26ms/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5SSUxxwszK5A"
      },
      "source": [
        "# **Note Prediction**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mhXsTPYt_zul"
      },
      "source": [
        "## **Preprocessing**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "4FzL2U9F_zuq"
      },
      "outputs": [],
      "source": [
        "sample_rate = 4000  # sample rate"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l4IfV2BW_zus"
      },
      "source": [
        "**Paths**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "5wqFHlbC_zuu"
      },
      "outputs": [],
      "source": [
        "songsPath = '/content/drive/MyDrive/Deep Learning/Music Generator/piano_pieces'\n",
        "notesPath = '/content/drive/MyDrive/Deep Learning/Music Generator/piano_triads'"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "24lvrZd0CSey"
      },
      "source": [
        "**Note encoding**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "hu09ybI0hMKn"
      },
      "outputs": [],
      "source": [
        "noteLabel = {}\n",
        "noteLabel['NA'] = 0 # NO note code\n",
        "noteLabel['START'] = 1 # start note code\n",
        "noteLabel['END'] = 2 # start note code\n",
        "\n",
        "i = 3\n",
        "\n",
        "for filename in os.listdir(notesPath): # notes code\n",
        "  noteLabel[filename[:-4]] = i\n",
        "  i+=1\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xI8VZau3_zux"
      },
      "source": [
        "**Helper Functions**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "17zqWstf_zuz"
      },
      "outputs": [],
      "source": [
        "# add padding to piece to grow up its length to max_length\n",
        "def addPadding(piece):\n",
        "  piece = np.array(piece)\n",
        "  max_length = 31\n",
        "  return np.pad(piece, (0,max_length-piece.shape[0]))\n",
        "\n",
        "# split np.array to multiple specific length array\n",
        "def splitToSpecificLength(arr, length):\n",
        "  arr = np.array(arr)\n",
        "  numberOfPiece = arr.shape[0]//length + 1 if arr.shape[0]%length != 0 else arr.shape[0]//length\n",
        "  return np.split(arr, [length*i for i in range(1, numberOfPiece)])\n",
        "\n",
        "# Generate Dataset with specific length for Note Prediction\n",
        "def generateXYForSpecificNumberOfNotes(song, length):\n",
        "  arr = splitToSpecificLength(song, length)\n",
        "  NP_XY = [(['START'] + list(arr[i]), arr[i+1][0]) for i in range(len(arr)-1)]\n",
        "  NP_XY.append((['START'] + list(arr[-1]), 'END'))\n",
        "  return NP_XY\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Y_Nub38Y_zu9"
      },
      "source": [
        "**Read all song notes**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Y1WeNl_V_zvB"
      },
      "outputs": [],
      "source": [
        "songs = []\n",
        "for song in os.listdir(songsPath):\n",
        "  data  = open(songsPath + '/' + song, 'r')\n",
        "  data = data.readlines()\n",
        "  if data:\n",
        "    songs.append(data[0].split(' / '))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kQ479cx0mZXG",
        "outputId": "427029df-f048-4d63-9d59-736fc1f79cbf"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "435"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ],
      "source": [
        "len(noteLabel)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vKF6ttuO_zvG"
      },
      "source": [
        "**Estimate number of dataset**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "J3muOcFG_zvI",
        "outputId": "fde34685-38cb-451a-c7e5-4fc821b69217"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "21995.0\n"
          ]
        }
      ],
      "source": [
        "songsLength = [len(x) for x in songs]\n",
        "songsLength = np.array(songsLength)\n",
        "numberOfPiecesPerSong = np.ceil(songsLength/3) + np.ceil(songsLength/4) #+ np.ceil(songsLength/5) + np.ceil(songsLength/6)\\\n",
        "                     #+ np.ceil(songsLength/7) + np.ceil(songsLength/8)\n",
        "\n",
        "numberOfPieces = np.sum(numberOfPiecesPerSong)\n",
        "\n",
        "print(numberOfPieces)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rRI0a5wE_zvL"
      },
      "source": [
        "**NP(S2S) Data Generator**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "y4oeSjO7MEyS",
        "outputId": "1a78e235-e6d9-44a2-86c1-8345f06ddd1c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['START', 'Eb_min_5_0', 'Cs_maj_6_0', 'Cs_maj_7_1']\n",
            "----------------------------------------------------------------------------------------------------\n",
            "B_maj_7_1\n"
          ]
        }
      ],
      "source": [
        "XY_NP = generateXYForSpecificNumberOfNotes(songs[0], 3)\n",
        "\n",
        "X_NP = [x[0] for x in XY_NP]\n",
        "y_true_NP = [x[1] for x in XY_NP]\n",
        "\n",
        "print(X_NP[15])\n",
        "print('-' * 100)\n",
        "print(y_true_NP[15])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "M7oaJDp1onFo",
        "outputId": "2c3d4326-9bb1-4b30-9558-c45931d7cd33"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((31, 435), (435,))"
            ]
          },
          "metadata": {},
          "execution_count": 29
        }
      ],
      "source": [
        "# test one hot\n",
        "x0 = X_NP[15]\n",
        "\n",
        "y0 = y_true_NP[15]\n",
        "\n",
        "x0 = [noteLabel[n] for n in x0]\n",
        "x0_pad = addPadding(x0)\n",
        "y0 = noteLabel[y0]\n",
        "\n",
        "x0_hot = to_categorical(x0_pad, num_classes=len(noteLabel))\n",
        "y0_hot = to_categorical(y0, num_classes=len(noteLabel))\n",
        "x0_hot.shape, y0_hot.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "bu0Nbg9G_zvM"
      },
      "outputs": [],
      "source": [
        "def NP_data_generator(songs, batchsize):\n",
        "  X_NP_batch, y_NP_batch = [], []\n",
        "  batchcount = 0\n",
        "  while True:\n",
        "    for song in songs:\n",
        "      # separate the notes of song with 30 length\n",
        "      XY_NP = generateXYForSpecificNumberOfNotes(song, 3) + generateXYForSpecificNumberOfNotes(song, 4)\n",
        "            # + generateXYForSpecificNumberOfNotes(song, 5) + generateXYForSpecificNumberOfNotes(song, 5) \\\n",
        "            # + generateXYForSpecificNumberOfNotes(song, 7) + generateXYForSpecificNumberOfNotes(song, 8) \\\n",
        "\n",
        "\n",
        "      # create all pieces in the song\n",
        "      for sample in XY_NP:\n",
        "        # sample[0] => X_NP\n",
        "        # sample[1] => y_NP\n",
        "        X_sample = [noteLabel[note] for note in sample[0]]\n",
        "        y_sample = noteLabel[sample[1]]\n",
        "        #print(y_sample)\n",
        "\n",
        "\n",
        "        X_sample_pad = addPadding(X_sample) # Add padding and grow its length up to 780\n",
        "\n",
        "        # make X and Y One Hot\n",
        "        X_sample_HOT = to_categorical(X_sample_pad, num_classes=len(noteLabel))\n",
        "        y_sample_HOT = to_categorical(y_sample, num_classes=len(noteLabel))\n",
        "\n",
        "        X_NP_batch.append(X_sample_HOT)\n",
        "        y_NP_batch.append(y_sample_HOT)\n",
        "\n",
        "        batchcount += 1\n",
        "        if batchcount > batchsize:\n",
        "          X_NP_batch = np.array(X_NP_batch)\n",
        "          y_NP_batch = np.array(y_NP_batch)\n",
        "          yield (X_NP_batch, y_NP_batch)\n",
        "          X_NP_batch = []\n",
        "          y_NP_batch = []\n",
        "          batchcount = 0"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DDHY6enp8D42"
      },
      "source": [
        "## **Implementation**"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 533
        },
        "id": "TxPMfAyl8LtB",
        "outputId": "02aa2815-f003-4b1d-fc66-06071a477b9b"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<IPython.core.display.Image object>"
            ],
            "image/png": "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\n"
          },
          "metadata": {},
          "execution_count": 233
        }
      ],
      "source": [
        "batchsize = 250\n",
        "NPInputShape = (31, 435)\n",
        "\n",
        "input = Input(shape=NPInputShape, name=\"input\")\n",
        "\n",
        "l = LSTM(128, activation=\"relu\", return_sequences=True)(input)\n",
        "l = LSTM(64, activation=\"relu\")(l)\n",
        "\n",
        "d = Dense(64, activation=\"relu\")(l)\n",
        "\n",
        "output = Dense(NPInputShape[1], activation=\"softmax\")(d)\n",
        "\n",
        "NP_model = Model(inputs=input, outputs=output)\n",
        "\n",
        "# Compile the Model\n",
        "NP_model.compile(loss=\"categorical_crossentropy\", optimizer=Adam(0.001), metrics=['accuracy'])\n",
        "\n",
        "filepath = '/content/drive/MyDrive/Deep Learning/Music Generator/NP_best_model.hdf5'\n",
        "checkpoint = ModelCheckpoint(filepath=filepath,\n",
        "                              monitor='accuracy',\n",
        "                              verbose=0,\n",
        "                              save_best_only=True,\n",
        "                              mode='max')\n",
        "\n",
        "plot_model(NP_model, to_file=\"NP_model.png\", show_shapes=True)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iP-acl_IC6D7",
        "outputId": "8bdfbf87-d444-48fc-ef72-f8d52213cb59"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/50\n",
            "87/87 [==============================] - 34s 331ms/step - loss: 5.9511 - accuracy: 0.0334\n",
            "Epoch 2/50\n",
            "87/87 [==============================] - 29s 332ms/step - loss: 5.8845 - accuracy: 0.0363\n",
            "Epoch 3/50\n",
            "87/87 [==============================] - 29s 333ms/step - loss: 5.7910 - accuracy: 0.0364\n",
            "Epoch 4/50\n",
            "87/87 [==============================] - 30s 336ms/step - loss: 5.7655 - accuracy: 0.0362\n",
            "Epoch 5/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.7517 - accuracy: 0.0363\n",
            "Epoch 6/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.7387 - accuracy: 0.0362\n",
            "Epoch 7/50\n",
            "87/87 [==============================] - 29s 333ms/step - loss: 5.7363 - accuracy: 0.0364\n",
            "Epoch 8/50\n",
            "87/87 [==============================] - 30s 338ms/step - loss: 5.7227 - accuracy: 0.0364\n",
            "Epoch 9/50\n",
            "87/87 [==============================] - 29s 331ms/step - loss: 5.7106 - accuracy: 0.0365\n",
            "Epoch 10/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.6939 - accuracy: 0.0358\n",
            "Epoch 11/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.6860 - accuracy: 0.0358\n",
            "Epoch 12/50\n",
            "87/87 [==============================] - 29s 327ms/step - loss: 5.6691 - accuracy: 0.0360\n",
            "Epoch 13/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.6598 - accuracy: 0.0364\n",
            "Epoch 14/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.6521 - accuracy: 0.0360\n",
            "Epoch 15/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.6460 - accuracy: 0.0359\n",
            "Epoch 16/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.6422 - accuracy: 0.0363\n",
            "Epoch 17/50\n",
            "87/87 [==============================] - 29s 333ms/step - loss: 5.6305 - accuracy: 0.0365\n",
            "Epoch 18/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.6262 - accuracy: 0.0364\n",
            "Epoch 19/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.6298 - accuracy: 0.0362\n",
            "Epoch 20/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.6193 - accuracy: 0.0364\n",
            "Epoch 21/50\n",
            "87/87 [==============================] - 29s 332ms/step - loss: 5.6268 - accuracy: 0.0368\n",
            "Epoch 22/50\n",
            "87/87 [==============================] - 29s 330ms/step - loss: 5.6221 - accuracy: 0.0366\n",
            "Epoch 23/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.6086 - accuracy: 0.0365\n",
            "Epoch 24/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.5946 - accuracy: 0.0367\n",
            "Epoch 25/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.5911 - accuracy: 0.0367\n",
            "Epoch 26/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.5836 - accuracy: 0.0366\n",
            "Epoch 27/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.5699 - accuracy: 0.0365\n",
            "Epoch 28/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.5754 - accuracy: 0.0365\n",
            "Epoch 29/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.5623 - accuracy: 0.0367\n",
            "Epoch 30/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.5575 - accuracy: 0.0367\n",
            "Epoch 31/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.5495 - accuracy: 0.0367\n",
            "Epoch 32/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.5400 - accuracy: 0.0366\n",
            "Epoch 33/50\n",
            "87/87 [==============================] - 29s 327ms/step - loss: 5.5364 - accuracy: 0.0364\n",
            "Epoch 34/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.5276 - accuracy: 0.0364\n",
            "Epoch 35/50\n",
            "87/87 [==============================] - 29s 327ms/step - loss: 5.5121 - accuracy: 0.0366\n",
            "Epoch 36/50\n",
            "87/87 [==============================] - 29s 326ms/step - loss: 5.5199 - accuracy: 0.0366\n",
            "Epoch 37/50\n",
            "87/87 [==============================] - 29s 327ms/step - loss: 5.5375 - accuracy: 0.0368\n",
            "Epoch 38/50\n",
            "87/87 [==============================] - 29s 326ms/step - loss: 5.5374 - accuracy: 0.0362\n",
            "Epoch 39/50\n",
            "87/87 [==============================] - 29s 327ms/step - loss: 5.5110 - accuracy: 0.0364\n",
            "Epoch 40/50\n",
            "87/87 [==============================] - 29s 330ms/step - loss: 5.4856 - accuracy: 0.0369\n",
            "Epoch 41/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.4707 - accuracy: 0.0364\n",
            "Epoch 42/50\n",
            "87/87 [==============================] - 29s 327ms/step - loss: 5.4578 - accuracy: 0.0363\n",
            "Epoch 43/50\n",
            "87/87 [==============================] - 29s 328ms/step - loss: 5.4592 - accuracy: 0.0366\n",
            "Epoch 44/50\n",
            "87/87 [==============================] - 29s 326ms/step - loss: 5.5051 - accuracy: 0.0359\n",
            "Epoch 45/50\n",
            "87/87 [==============================] - 29s 329ms/step - loss: 5.4717 - accuracy: 0.0367\n",
            "Epoch 46/50\n",
            "87/87 [==============================] - 29s 326ms/step - loss: 5.4451 - accuracy: 0.0368\n",
            "Epoch 47/50\n",
            "87/87 [==============================] - 32s 361ms/step - loss: 5.4321 - accuracy: 0.0364\n",
            "Epoch 48/50\n",
            "87/87 [==============================] - 29s 330ms/step - loss: 5.4035 - accuracy: 0.0364\n",
            "Epoch 49/50\n",
            "87/87 [==============================] - 29s 326ms/step - loss: 5.3889 - accuracy: 0.0363\n",
            "Epoch 50/50\n",
            "87/87 [==============================] - 29s 330ms/step - loss: 5.3753 - accuracy: 0.0369\n"
          ]
        }
      ],
      "source": [
        "with tf.device('/GPU:0'):\n",
        "  NP_history = NP_model.fit(NP_data_generator(songs, batchsize), steps_per_epoch=numberOfPieces/batchsize, epochs=50, callbacks=[checkpoint]).history\n",
        "  NP_model = load_model(filepath)\n",
        "  # outputResult(NP_model, NP_history, X_train, y_train, X_test, y_test, noteDic)\n",
        "  # showPlots(NP_model, NP_history)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# **Generate Song**"
      ],
      "metadata": {
        "id": "LxVgYWN6TpeU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "notesPath = '/content/drive/MyDrive/Deep Learning/Music Generator/piano_triads'\n",
        "noisesPath = '/content/drive/MyDrive/Deep Learning/Music Generator/artifacts'\n",
        "songsPath = '/content/drive/MyDrive/Deep Learning/Music Generator/piano_pieces'\n",
        "\n",
        "DAE_path = '/content/drive/MyDrive/Deep Learning/Music Generator/DAE_best_model_3.hdf5'\n",
        "NP_path = '/content/drive/MyDrive/Deep Learning/Music Generator/NP_best_model.hdf5'\n",
        "NR_path = '/content/drive/MyDrive/Deep Learning/Music Generator/NR_best_model.hdf5'\n",
        "\n",
        "sample_rate = 4000"
      ],
      "metadata": {
        "id": "7MnPbInRfF0E"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "DAE_model = load_model(DAE_path)"
      ],
      "metadata": {
        "id": "6NbpqJNZBYCA"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "NR_model = load_model(NR_path)"
      ],
      "metadata": {
        "id": "D56w4DTHB8JU"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "NP_model = load_model(NP_path)"
      ],
      "metadata": {
        "id": "eXmP65tECAiW"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import random\n",
        "\n",
        "# generate 3 random notes as initial piece\n",
        "def generateRandomNotes(songs):\n",
        "  songNumber = random.randint(0, len(songs))  # random song\n",
        "\n",
        "  song = songs[songNumber]  # song\n",
        "  songsLength = len(song)\n",
        "  numberOfPiecesPerSong = np.ceil(songsLength/10) + np.ceil(songsLength/20)\n",
        "\n",
        "  pieceNumber = random.randint(0, numberOfPiecesPerSong-1)  # random piece in the song\n",
        "\n",
        "  songSplitted = splitToSpecificLength(song, 10) + splitToSpecificLength(song, 20)  # pieces in the song\n",
        "\n",
        "  if len(songSplitted[pieceNumber]) >= 3:\n",
        "    return songSplitted[pieceNumber][:3]\n",
        "  else:\n",
        "    return generateRandomNotes(songs)\n",
        "\n",
        "# mix piece and noise together\n",
        "def AddNoiseToPiece(piece, noise, noise_factor):\n",
        "  noisy_piece = piece + (noise_factor*noise)\n",
        "  return noisy_piece\n",
        "\n",
        "# append wav np.array of notes by having note names in a piece\n",
        "def appendNotes(piece, notes, notesLabel):\n",
        "  appendedNotes = []\n",
        "  for piece_note in piece:\n",
        "    index = notesLabel.index(piece_note)\n",
        "    appendedNotes.extend(notes[index])\n",
        "  return np.array(appendedNotes)\n",
        "\n",
        "# add padding to wav piece to grow up its length to max_length\n",
        "def addPaddingWav(piece_wav):\n",
        "  max_length = sample_rate * 90\n",
        "  return np.pad(piece_wav, (0,max_length-piece_wav.shape[0]))\n",
        "\n",
        "# add padding to piece to grow up its length to max_length\n",
        "def addPadding(piece):\n",
        "  piece = np.array(piece)\n",
        "  max_length = 31\n",
        "  return np.pad(piece, (0,max_length-piece.shape[0]))\n",
        "\n",
        "# add random noise to piece\n",
        "def addRandomNoise(noises, initial_piece):\n",
        "  noiseNumber = random.randint(0, len(noises)-1)\n",
        "  noisy_piece = AddNoiseToPiece(initial_piece, noises[noiseNumber], 0.25)\n",
        "  return noisy_piece\n",
        "\n",
        "# break piece down to its notes\n",
        "def breakDownPiece(denoised_piece, sample_rate, numberOfNotes):\n",
        "  no_padding_piece_length = sample_rate * 3 * numberOfNotes\n",
        "  breakdown_pieces = splitToSpecificLength(denoised_piece[:no_padding_piece_length], sample_rate * 3)\n",
        "  return breakdown_pieces\n",
        "\n",
        "\n",
        "def notesToVector(piece, dic): # piece should be one hot do argmaxkeep dimention\n",
        "  encoded_piece = [dic['START']] + [dic[x] for x in piece]\n",
        "  encoded_piece_pad = addPadding(encoded_piece)\n",
        "  vectorized_piece = to_categorical(encoded_piece_pad, num_classes=len(dic))\n",
        "  return vectorized_piece\n",
        "\n",
        "\n",
        "def DAE(DAE_model, noisy_piece):\n",
        "  return DAE_model.predict(np.array([noisy_piece]))[0]\n",
        "\n",
        "def NID(NR_model, note, notesLabel):\n",
        "  noteName = notesLabel[np.argmax(NR_model.predict(np.array([note]))[0])]\n",
        "  return noteName\n",
        "\n",
        "def LM(NP_model, sub_notes_vec, dic):\n",
        "  next_note = NP_model.predict(np.array([sub_notes_vec]))[0]\n",
        "  next_note_name = predictionToNoteName(dic, next_note)\n",
        "  return next_note_name\n",
        "\n",
        "def predictionToNoteName(dic, note):\n",
        "  noteIndex = np.argmax(note)\n",
        "  if random.getrandbits(1) and noteIndex == 2:\n",
        "    noteIndex = random.randint(3, 434)\n",
        "  noteName = list(dic.keys())[list(dic.values()).index(noteIndex)]\n",
        "  return noteName\n",
        "\n",
        "\n",
        "def OHtoNoteName(dic, note):\n",
        "  noteIndex = np.argmax(note)\n",
        "  #print(noteIndex)\n",
        "  noteName = list(dic.keys())[list(dic.values()).index(noteIndex)]\n",
        "  return noteName\n"
      ],
      "metadata": {
        "id": "PMieNNDTUMbO"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "notes, notesLabel, noises = [], [], []\n",
        "sample_rate = 4000\n",
        "\n",
        "noteDict = {}\n",
        "noteDict['NA'] = 0 # NO note code\n",
        "noteDict['START'] = 1 # start note code\n",
        "noteDict['END'] = 2 # start note code\n",
        "\n",
        "i = 3\n",
        "\n",
        "for note in os.listdir(notesPath):\n",
        "  data  = librosa.load(notesPath+'/'+note, sr=sample_rate, dtype=np.float32)[0]\n",
        "  notes.append(data)\n",
        "  notesLabel.append(note[:-4])\n",
        "  noteDict[note[:-4]] = i\n",
        "  i+=1\n",
        "\n",
        "\n",
        "for noise in os.listdir(noisesPath):\n",
        "  # noises must have 1/3 note sample rate because their length are 3X\n",
        "  noises.append(librosa.load(noisesPath + '/'+ noise, sr=sample_rate*10,dtype=np.float32)[0]) # sample rate for a 90s song: 10 noises per song\n",
        "\n",
        "notes = np.array(notes)\n",
        "noises = np.array(noises)\n",
        "# notesLabel = np.array(notesLabel)\n",
        "print(notes.shape)\n",
        "print(noises.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3jkJboG_ec3y",
        "outputId": "03782f42-3afc-45e4-b467-68b3729c1c78"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(432, 12000)\n",
            "(40, 360000)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "sub_notes_classes = None\n",
        "\n",
        "note_list = generateRandomNotes(songs)\n",
        "NOTES_GROUND_TRUTH = list(note_list)\n",
        "initial_piece = appendNotes(note_list, notes, notesLabel)\n",
        "initial_piece_pad = addPaddingWav(initial_piece)\n",
        "noisy_piece = addRandomNoise(noises, initial_piece_pad)\n",
        "final_piece = None\n",
        "\n",
        "# print([noisy_piece])\n",
        "numberOfNotes = 3\n",
        "piece_limit = 30\n",
        "while True:\n",
        "  denoised_piece = DAE(DAE_model, noisy_piece)\n",
        "  sub_notes = breakDownPiece(denoised_piece, sample_rate, numberOfNotes)\n",
        "  sub_notes_classes = [NID(NR_model, note, notesLabel) for note in sub_notes]\n",
        "  sub_notes_vec = notesToVector(sub_notes_classes, noteDict)\n",
        "  next_note = LM(NP_model, sub_notes_vec, noteDict)\n",
        "  NOTES_GROUND_TRUTH.append(next_note)\n",
        "  new_piece =  appendNotes(sub_notes_classes, notes, notesLabel)\n",
        "  note_list = sub_notes_classes + [next_note]\n",
        "  numberOfNotes += 1\n",
        "\n",
        "  if next_note == 'END' or numberOfNotes == 30:\n",
        "    final_piece = new_piece\n",
        "    break\n",
        "  else:\n",
        "    new_piece = appendNotes(note_list, notes, notesLabel)\n",
        "    new_piece_pad = addPaddingWav(new_piece)\n",
        "    noisy_piece = addRandomNoise(noises, new_piece_pad)\n"
      ],
      "metadata": {
        "id": "SszQoUKITxj-"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sub_notes_classes"
      ],
      "metadata": {
        "id": "Zdu5yoAVJOvC",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fb5dd957-c773-495d-d518-576eadbdb5a4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['Gs_maj_6_1', 'A_maj_6_1', 'Bb_dim_7_1', 'F_maj_4_1', 'C_min_5_0']"
            ]
          },
          "metadata": {},
          "execution_count": 60
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "NOTES_GROUND_TRUTH"
      ],
      "metadata": {
        "id": "CvV48XlwJU6p",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "61e8d945-78c2-406c-e3b4-d7918337ecfa"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['G_min_7_1', 'A_dim_5_1', 'Bb_maj_4_1', 'B_min_6_1', 'Cs_min_4_1', 'END']"
            ]
          },
          "metadata": {},
          "execution_count": 61
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "wavfile.write('final_piece.wav', sample_rate, final_piece)"
      ],
      "metadata": {
        "id": "_RzNNHHGJHy-"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [
        "rG540utTBHVi",
        "Q1qXvuLDsHy-",
        "mHrW-d9w4Css",
        "LXvHNESIZ_BV",
        "9g2bT1KDsTU1",
        "WcHmHbcQsYPY",
        "6YZs2_NCO_2K",
        "jHKA9UHazJkI",
        "vIqIjSi41bS8",
        "plg3bsOfjJYn",
        "WuHsZrgdmTgO",
        "NmUGeK9MoHyt",
        "1t344qwo0F-o",
        "KGWOhtvqgfQI",
        "xzGuRMZc2UUT",
        "US2y4cnYzL3I",
        "5SSUxxwszK5A",
        "mhXsTPYt_zul",
        "DDHY6enp8D42",
        "LxVgYWN6TpeU"
      ],
      "provenance": [],
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}