audio_numericMeta_text_cmet.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/284709b96cd84f60930aad7d2ca8bb10\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [25]                : (0.3502214252948761, 0.658504843711853)\n",
      "COMET INFO:     batch_accuracy [1200]        : (0.125, 0.8125)\n",
      "COMET INFO:     batch_loss [1200]            : (0.5997779369354248, 2.7833456993103027)\n",
      "COMET INFO:     epoch_duration [25]          : (5.0, 10.797000000005937)\n",
      "COMET INFO:     loss [25]                    : (0.8623648881912231, 1.9000506401062012)\n",
      "COMET INFO:     val_accuracy [25]            : (0.3466666638851166, 0.5355555415153503)\n",
      "COMET INFO:     val_loss [25]                : (1.3721890449523926, 1.8192678689956665)\n",
      "COMET INFO:     validate_batch_accuracy [50] : (0.28125, 0.625)\n",
      "COMET INFO:     validate_batch_loss [50]     : (1.1516964435577393, 1.9229846000671387)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 1\n",
      "COMET INFO:     optimizer_id           : 6bec93d17d45498f9a5c1f410b0a2417\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.8623648881912231\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 300, \"audio_layer_2\": 500, \"batch_size\": 16, \"epochs\": 14, \"final_layer\": 64, \"meta_layer_1\": 550, \"meta_layer_2\": 500}\n",
      "COMET INFO:     optimizer_pid          : 790f6678bda1167911eed26b591fa6bf9e4f0f8c\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 1830132\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 300\n",
      "COMET INFO:     audio_layer_2      : 500\n",
      "COMET INFO:     batch_size         : 16\n",
      "COMET INFO:     epochs             : 14\n",
      "COMET INFO:     final_layer        : 64\n",
      "COMET INFO:     meta_layer_1       : 550\n",
      "COMET INFO:     meta_layer_2       : 500\n",
      "COMET INFO:     steps              : 480\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/8510b3a45a714fb0855940ac073f9ab1\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# import comet_ml at the top of your file\n",
    "from comet_ml import Experiment\n",
    "experiment = Experiment(\n",
    "            project_name=\"music-genre-multiclass-classification\",\n",
    "            workspace=\"wodenwang820118\",\n",
    "        )\n",
    "import comet_ml\n",
    "import logging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [],
   "source": [
    "logging.basicConfig(level=logging.INFO)\n",
    "LOGGER = logging.getLogger(\"comet_ml\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow import keras\n",
    "\n",
    "%matplotlib inline\n",
    "sns.set_style('whitegrid')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [],
   "source": [
    "# tensorflow 2.7 \n",
    "import tensorflow as tf \n",
    "from tensorflow.keras.layers import Input,Flatten,Dense,Dropout,BatchNormalization,Concatenate\n",
    "from tensorflow.keras.models import Model, Sequential"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [],
   "source": [
    "# scale the numeric data\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "scaler = StandardScaler()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_features = pd.read_csv('data/train_features.csv')\n",
    "df_train_labels = pd.read_csv('data/train_labels.csv')\n",
    "\n",
    "df_valid_features = pd.read_csv('data/valid_features.csv')\n",
    "df_valid_labels = pd.read_csv('data/valid_labels.csv')\n",
    "\n",
    "df_test_features = pd.read_csv('data/test_features.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>trackID</th>\n",
       "      <th>title</th>\n",
       "      <th>tags</th>\n",
       "      <th>loudness</th>\n",
       "      <th>tempo</th>\n",
       "      <th>time_signature</th>\n",
       "      <th>key</th>\n",
       "      <th>mode</th>\n",
       "      <th>duration</th>\n",
       "      <th>vect_1</th>\n",
       "      <th>...</th>\n",
       "      <th>vect_139</th>\n",
       "      <th>vect_140</th>\n",
       "      <th>vect_141</th>\n",
       "      <th>vect_142</th>\n",
       "      <th>vect_143</th>\n",
       "      <th>vect_144</th>\n",
       "      <th>vect_145</th>\n",
       "      <th>vect_146</th>\n",
       "      <th>vect_147</th>\n",
       "      <th>vect_148</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>8424</td>\n",
       "      <td>The Calamity [Explicit]</td>\n",
       "      <td>the, you, to, a, not, in, of, that, are, we, w...</td>\n",
       "      <td>-2.824</td>\n",
       "      <td>131.969</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>298.13506</td>\n",
       "      <td>52.050795</td>\n",
       "      <td>...</td>\n",
       "      <td>0.001097</td>\n",
       "      <td>0.001161</td>\n",
       "      <td>0.001225</td>\n",
       "      <td>0.001251</td>\n",
       "      <td>0.001238</td>\n",
       "      <td>0.001202</td>\n",
       "      <td>0.00113</td>\n",
       "      <td>0.00104</td>\n",
       "      <td>0.104525</td>\n",
       "      <td>0.335818</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 157 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   trackID                    title  \\\n",
       "0     8424  The Calamity [Explicit]   \n",
       "\n",
       "                                                tags  loudness    tempo  \\\n",
       "0  the, you, to, a, not, in, of, that, are, we, w...    -2.824  131.969   \n",
       "\n",
       "   time_signature  key  mode   duration     vect_1  ...  vect_139  vect_140  \\\n",
       "0               4    0     1  298.13506  52.050795  ...  0.001097  0.001161   \n",
       "\n",
       "   vect_141  vect_142  vect_143  vect_144  vect_145  vect_146  vect_147  \\\n",
       "0  0.001225  0.001251  0.001238  0.001202   0.00113   0.00104  0.104525   \n",
       "\n",
       "   vect_148  \n",
       "0  0.335818  \n",
       "\n",
       "[1 rows x 157 columns]"
      ]
     },
     "execution_count": 156,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train_features.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_train_data = df_train_features.iloc[:,9:]\n",
    "num_valid_data = df_valid_features.iloc[:,9:]\n",
    "num_test_data = df_test_features.iloc[:,9:]\n",
    "\n",
    "num_train_meta_data = df_train_features.iloc[:,3:5]\n",
    "num_valid_meta_data = df_valid_features.iloc[:,3:5]\n",
    "\n",
    "num_train_meta_data['duration'] = df_train_features['duration']\n",
    "num_valid_meta_data['duration'] = df_valid_features['duration']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_train = scaler.fit_transform(num_train_data)\n",
    "num_valid = scaler.fit_transform(num_valid_data)\n",
    "\n",
    "num_train_meta = scaler.fit_transform(num_train_meta_data)\n",
    "num_valid_meta = scaler.fit_transform(num_valid_meta_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_train_y = pd.get_dummies(df_train_labels['genre'])\n",
    "num_valid_y = pd.get_dummies(df_valid_labels['genre'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [],
   "source": [
    "def str_remove_comma(string):\n",
    "    str = string.split(\", \")\n",
    "    return str\n",
    "\n",
    "def list_to_string(list):\n",
    "    str = \" \"\n",
    "    return str.join(list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = df_train_features['tags'].apply(lambda x : list_to_string(str_remove_comma(x)))\n",
    "y_train = df_train_labels['genre'].to_numpy()\n",
    "\n",
    "x_test = df_valid_features['tags'].apply(lambda x : list_to_string(str_remove_comma(x)))\n",
    "y_test = df_valid_labels['genre'].to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [],
   "source": [
    "# remove stopwords\n",
    "import nltk\n",
    "stopwords = nltk.corpus.stopwords.words('english')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = x_train.apply(lambda x : ' '.join([word for word in x.split() if word not in (stopwords)]))\n",
    "x_test = x_test.apply(lambda x : ' '.join([word for word in x.split() if word not in (stopwords)]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = np.array(x_train)\n",
    "x_test = np.array(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.utils import class_weight\n",
    "class_weight = class_weight.compute_class_weight(class_weight='balanced', classes=np.unique(df_train_labels['genre']),y=df_train_labels['genre'])\n",
    "class_weight = {i :class_weight[i] for i in range(8)}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "early_stop = EarlyStopping(monitor='val_loss',patience=5,restore_best_weights=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [],
   "source": [
    "import ktrain\n",
    "from ktrain import text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "language: en\n",
      "Word Counts: 4776\n",
      "Nrows: 7678\n",
      "7678 train sequences\n",
      "train sequence lengths:\n",
      "\tmean : 52\n",
      "\t95percentile : 95\n",
      "\t99percentile : 146\n",
      "x_train shape: (7678,150)\n",
      "y_train shape: (7678, 8)\n",
      "Is Multi-Label? False\n",
      "450 test sequences\n",
      "test sequence lengths:\n",
      "\tmean : 50\n",
      "\t95percentile : 101\n",
      "\t99percentile : 129\n",
      "x_test shape: (450,150)\n",
      "y_test shape: (450, 8)\n",
      "task: text classification\n"
     ]
    }
   ],
   "source": [
    "trn2, val2, preproc2 = text.texts_from_array(\n",
    "    x_train=x_train, y_train=y_train,\n",
    "    x_test=x_test, y_test=y_test,\n",
    "    ngram_range=1, \n",
    "    maxlen=150,\n",
    "    preprocess_mode='standard'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Is Multi-Label? False\n",
      "compiling word ID features...\n",
      "maxlen is 150\n",
      "done.\n"
     ]
    }
   ],
   "source": [
    "model2 = text.text_classifier('fasttext', train_data=trn2, preproc=preproc2, metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "class MyCustomDataset(ktrain.SequenceDataset):\n",
    "    def __init__(self, x, y, batch_size=32, shuffle=True):\n",
    "        # error checks\n",
    "        err = False\n",
    "        if type(x) == np.ndarray and len(x.shape) != 2: err = True\n",
    "        elif type(x) == list:\n",
    "            for d in x:\n",
    "                if type(d) != np.ndarray or len(d.shape) != 2:\n",
    "                    err = True\n",
    "                    break\n",
    "        else: err = True\n",
    "        if err:\n",
    "            raise ValueError('x must be a 2d numpy array or a list of 2d numpy arrays')\n",
    "        if type(y) != np.ndarray:\n",
    "            raise ValueError('y must be a numpy array')\n",
    "        if type(x) == np.ndarray:\n",
    "            x = [x]\n",
    "        # set variables\n",
    "\n",
    "        super().__init__(batch_size=batch_size)\n",
    "        self.x, self.y = x, y\n",
    "        self.indices = np.arange(self.x[0].shape[0])\n",
    "        self.n_inputs = len(x)\n",
    "        self.shuffle = shuffle\n",
    "\n",
    "    # required for instances of tf.keras.utils.Sequence\n",
    "    def __len__(self):\n",
    "        \"\"\"Number of batches in the Sequence.\n",
    "        \"\"\"\n",
    "        return math.ceil(self.x[0].shape[0] / self.batch_size)\n",
    "\n",
    "    # required for instances of tf.keras.utils.Sequence\n",
    "    def __getitem__(self, idx):\n",
    "        \"\"\"Gets batch at position `idx`.\n",
    "        \"\"\"\n",
    "        # the starting and ending indices of the batch\n",
    "        inds = self.indices[idx * self.batch_size:(idx + 1) * self.batch_size]\n",
    "        batch_x = []\n",
    "\n",
    "        for i in range(self.n_inputs):\n",
    "            batch_x.append(self.x[i][inds])\n",
    "        batch_y = np.array(self.y[inds])\n",
    "        return tuple(batch_x) , batch_y\n",
    "\n",
    "    # required for instances of ktrain.Dataset\n",
    "    def nsamples(self):\n",
    "        return self.x[0].shape[0]\n",
    "\n",
    "    #required for instances of ktrain.Dataset\n",
    "    def get_y(self):\n",
    "        return self.y\n",
    "\n",
    "    def on_epoch_end(self):\n",
    "        if self.shuffle:  np.random.shuffle(self.indices)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = MyCustomDataset([num_train] + [num_train_meta] + [trn2[0]] , trn2[1])\n",
    "valid_data = MyCustomDataset([num_valid] + [num_train_meta] + [val2[0]] , val2[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [],
   "source": [
    "class AudioNumMetaClassifier:\n",
    "    def __init__(self,num_train,num_train_meta,train_data,num_train_y,num_valid,num_valid_meta,valid_data,num_valid_y,class_weight,early_stop,experiment,text_model):\n",
    "        # audio data\n",
    "        self.num_train = num_train\n",
    "        self.num_valid = num_valid\n",
    "\n",
    "        # meta data\n",
    "        self.num_train_meta = num_train_meta\n",
    "        self.num_valid_meta = num_valid_meta\n",
    "\n",
    "        # all data and ktrain model\n",
    "        self.train_data = train_data\n",
    "        self.valid_data = valid_data\n",
    "        self.text_model = text_model\n",
    "\n",
    "        # labels\n",
    "        self.num_train_y = num_train_y\n",
    "        self.num_valid_y = num_valid_y\n",
    "\n",
    "        # class weights and early stopping\n",
    "        self.class_weight = class_weight\n",
    "        self.early_stop = early_stop\n",
    "\n",
    "        # experiment\n",
    "        self.experiment = experiment\n",
    "    \n",
    "    def build_model(self):\n",
    "        # Build the model\n",
    "        model = Sequential()\n",
    "        model.add(\n",
    "            Dense(\n",
    "                units=self.experiment.get_parameter(\"audio_layer_1\"),\n",
    "                activation='elu',\n",
    "                input_shape=(self.num_train.shape[1],)\n",
    "            )\n",
    "        )\n",
    "        model.add(Dropout(0.5))\n",
    "        model.add(BatchNormalization())\n",
    "\n",
    "        model.add(\n",
    "            Dense(\n",
    "                units=self.experiment.get_parameter(\"audio_layer_2\"),\n",
    "                activation='elu'))\n",
    "        model.add(Dropout(0.3))\n",
    "        model.add(BatchNormalization())\n",
    "\n",
    "        model_meta = Sequential()\n",
    "        model_meta.add(\n",
    "            Dense(\n",
    "                units=self.experiment.get_parameter(\"meta_layer_1\"),\n",
    "                activation='elu',\n",
    "                input_shape=(self.num_train_meta.shape[1],)\n",
    "            )\n",
    "        )\n",
    "        model_meta.add(Dropout(0.5))\n",
    "        model.add(BatchNormalization())\n",
    "\n",
    "        model_meta.add(\n",
    "            Dense(\n",
    "                units=self.experiment.get_parameter(\"meta_layer_2\"),\n",
    "                activation='elu'))\n",
    "        model_meta.add(Dropout(0.3))\n",
    "        model_meta.add(BatchNormalization())\n",
    "\n",
    "\n",
    "        merged_out = Concatenate()([model.output,model_meta.output,self.text_model.output])\n",
    "        merged_out = Dense(\n",
    "            units=self.experiment.get_parameter(\"final_layer\"),\n",
    "            activation='elu')(merged_out)\n",
    "        merged_out = Dense(8,activation='softmax')(merged_out)\n",
    "\n",
    "        combined_model = Model([model.input] + [model_meta.input] + [self.text_model.input], [merged_out])\n",
    "        combined_model.compile(\n",
    "            optimizer='adam',\n",
    "            loss='categorical_crossentropy',\n",
    "            metrics=['accuracy']\n",
    "        )\n",
    "        return combined_model\n",
    "    \n",
    "    def train_model(self):\n",
    "        # Train the model\n",
    "        combined_model = self.build_model()\n",
    "        model = ktrain.get_learner(\n",
    "            combined_model,\n",
    "            train_data=self.train_data,\n",
    "            val_data=self.valid_data,\n",
    "            batch_size=self.experiment.get_parameter(\"batch_size\")\n",
    "        )\n",
    "        model.fit_onecycle(\n",
    "            lr=0.001,\n",
    "            callbacks=[self.early_stop],\n",
    "            epochs=self.experiment.get_parameter(\"epochs\"),\n",
    "            class_weight=self.class_weight,\n",
    "        )\n",
    "        return model\n",
    "    \n",
    "    def evaluate_model(self):\n",
    "        # Evaluate the model\n",
    "        model = self.train_model()\n",
    "        # score = model.evaluate([self.num_valid,self.num_valid_meta],self.num_valid_y)\n",
    "        # LOGGER.info(f\"{ score }\")\n",
    "    \n",
    "    def grid_search(self, config_dict):\n",
    "        opt = comet_ml.Optimizer(config_dict)\n",
    "        for self.experiment in opt.get_experiments(project_name=\"music-genre-multiclass-classification\"):\n",
    "\n",
    "            self.build_model()\n",
    "            self.train_model()\n",
    "            self.evaluate_model()\n",
    "            self.experiment.end()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [],
   "source": [
    "classifier = AudioNumMetaClassifier(num_train,num_train_meta,train_data,num_train_y,num_valid,num_valid_meta,valid_data,num_valid_y,class_weight,early_stop,experiment,model2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: COMET_OPTIMIZER_ID=ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO: Using optimizer config: {'algorithm': 'bayes', 'configSpaceSize': 329280, 'endTime': None, 'id': 'ac9577df43cc4625a3b73696e8632eb2', 'lastUpdateTime': None, 'maxCombo': 10, 'name': 'Optimize Music Classification Network', 'parameters': {'audio_layer_1': {'type': 'discrete', 'values': [450, 500, 550, 600, 700, 900, 950, 1000]}, 'audio_layer_2': {'type': 'discrete', 'values': [128, 256, 300, 400, 450, 500, 550]}, 'batch_size': {'type': 'discrete', 'values': [16, 32, 40, 64]}, 'epochs': {'type': 'discrete', 'values': [13, 14, 15, 16, 20]}, 'final_layer': {'type': 'discrete', 'values': [20, 32, 64, 100, 128, 256]}, 'meta_layer_1': {'type': 'discrete', 'values': [450, 500, 550, 600, 700, 900, 1000]}, 'meta_layer_2': {'type': 'discrete', 'values': [128, 256, 300, 400, 450, 500, 550]}}, 'predictor': None, 'spec': {'gridSize': 10, 'maxCombo': 10, 'metric': 'loss', 'minSampleSize': 100, 'objective': 'minimize', 'retryAssignLimit': 0, 'retryLimit': 1000}, 'startTime': 26935858952, 'state': {'mode': None, 'seed': None, 'sequence': [], 'sequence_i': 0, 'sequence_pid': None, 'sequence_retry': 0, 'sequence_retry_count': 0}, 'status': 'running', 'suggestion_count': 0, 'trials': 1, 'version': '2.0.1'}\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/8510b3a45a714fb0855940ac073f9ab1\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.9\n",
      "COMET INFO:     Adam_beta_2        : 0.999\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 0.001\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     copy               : True\n",
      "COMET INFO:     with_mean          : True\n",
      "COMET INFO:     with_std           : True\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/81950fc95aee44d39c456c1ff42cab1f\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/14\n",
      "192/192 [==============================] - 6s 23ms/step - loss: 1.8766 - accuracy: 0.3583 - val_loss: 1.4829 - val_accuracy: 0.4667\n",
      "Epoch 2/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.5278 - accuracy: 0.4531 - val_loss: 1.6652 - val_accuracy: 0.4200\n",
      "Epoch 3/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.4076 - accuracy: 0.4838 - val_loss: 1.6535 - val_accuracy: 0.4556\n",
      "Epoch 4/14\n",
      "192/192 [==============================] - 4s 22ms/step - loss: 1.3272 - accuracy: 0.5151 - val_loss: 1.6281 - val_accuracy: 0.4889\n",
      "Epoch 5/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.2227 - accuracy: 0.5431 - val_loss: 1.6644 - val_accuracy: 0.4489\n",
      "Epoch 6/14\n",
      "192/192 [==============================] - 4s 22ms/step - loss: 1.1805 - accuracy: 0.5604 - val_loss: 1.6551 - val_accuracy: 0.4356\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/14\n",
      "192/192 [==============================] - 7s 23ms/step - loss: 1.8338 - accuracy: 0.3778 - val_loss: 1.3834 - val_accuracy: 0.5356\n",
      "Epoch 2/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.4900 - accuracy: 0.4654 - val_loss: 1.5476 - val_accuracy: 0.4689\n",
      "Epoch 3/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.3888 - accuracy: 0.4990 - val_loss: 1.4526 - val_accuracy: 0.4867\n",
      "Epoch 4/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.3017 - accuracy: 0.5215 - val_loss: 1.6988 - val_accuracy: 0.4222\n",
      "Epoch 5/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.2441 - accuracy: 0.5423 - val_loss: 1.5874 - val_accuracy: 0.4600\n",
      "Epoch 6/14\n",
      "192/192 [==============================] - 4s 21ms/step - loss: 1.1838 - accuracy: 0.5639 - val_loss: 1.6399 - val_accuracy: 0.4422\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/81950fc95aee44d39c456c1ff42cab1f\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [12]                : (0.3582964241504669, 0.563948929309845)\n",
      "COMET INFO:     batch_accuracy [240]         : (0.07500000298023224, 0.699999988079071)\n",
      "COMET INFO:     batch_loss [240]             : (0.7820491790771484, 2.934331178665161)\n",
      "COMET INFO:     epoch_duration [12]          : (3.9679999999934807, 7.39100000000326)\n",
      "COMET INFO:     loss [12]                    : (1.1804945468902588, 1.8766275644302368)\n",
      "COMET INFO:     val_accuracy [12]            : (0.41999998688697815, 0.5355555415153503)\n",
      "COMET INFO:     val_loss [12]                : (1.3834078311920166, 1.6988356113433838)\n",
      "COMET INFO:     validate_batch_accuracy [24] : (0.28125, 0.59375)\n",
      "COMET INFO:     validate_batch_loss [24]     : (1.1319700479507446, 1.955886960029602)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 1\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 1.1804945468902588\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 1000, \"audio_layer_2\": 450, \"batch_size\": 40, \"epochs\": 14, \"final_layer\": 32, \"meta_layer_1\": 600, \"meta_layer_2\": 500}\n",
      "COMET INFO:     optimizer_pid          : fffde23391dd3feb8a59968e6b996b23214c96e1\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 2227838\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 1000\n",
      "COMET INFO:     audio_layer_2      : 450\n",
      "COMET INFO:     batch_size         : 40\n",
      "COMET INFO:     epochs             : 14\n",
      "COMET INFO:     final_layer        : 32\n",
      "COMET INFO:     meta_layer_1       : 600\n",
      "COMET INFO:     meta_layer_2       : 500\n",
      "COMET INFO:     steps              : 192\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading metrics, params, and assets to Comet before program termination (may take several seconds)\n",
      "COMET INFO: The Python SDK has 3600 seconds to finish before aborting...\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/b933d4ce741c44b9851c6443c8aeb40d\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/13\n",
      "240/240 [==============================] - 8s 23ms/step - loss: 1.8918 - accuracy: 0.3591 - val_loss: 1.5483 - val_accuracy: 0.4533\n",
      "Epoch 2/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 1.5611 - accuracy: 0.4504 - val_loss: 1.6119 - val_accuracy: 0.4711\n",
      "Epoch 3/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 1.4251 - accuracy: 0.4840 - val_loss: 1.6265 - val_accuracy: 0.4422\n",
      "Epoch 4/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 1.3350 - accuracy: 0.5121 - val_loss: 1.4969 - val_accuracy: 0.4889\n",
      "Epoch 5/13\n",
      "240/240 [==============================] - 5s 20ms/step - loss: 1.2552 - accuracy: 0.5428 - val_loss: 1.5980 - val_accuracy: 0.4644\n",
      "Epoch 6/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 1.1891 - accuracy: 0.5619 - val_loss: 1.5530 - val_accuracy: 0.4756\n",
      "Epoch 7/13\n",
      "240/240 [==============================] - 5s 20ms/step - loss: 1.1502 - accuracy: 0.5731 - val_loss: 1.4551 - val_accuracy: 0.5022\n",
      "Epoch 8/13\n",
      "240/240 [==============================] - 5s 22ms/step - loss: 1.0560 - accuracy: 0.6052 - val_loss: 1.4455 - val_accuracy: 0.5311\n",
      "Epoch 9/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 1.0056 - accuracy: 0.6145 - val_loss: 1.5552 - val_accuracy: 0.4867\n",
      "Epoch 10/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 0.9378 - accuracy: 0.6373 - val_loss: 1.4802 - val_accuracy: 0.5067\n",
      "Epoch 11/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 0.8792 - accuracy: 0.6572 - val_loss: 1.4601 - val_accuracy: 0.5289\n",
      "Epoch 12/13\n",
      "240/240 [==============================] - 5s 21ms/step - loss: 0.8193 - accuracy: 0.6796 - val_loss: 1.4092 - val_accuracy: 0.5422\n",
      "Epoch 13/13\n",
      "240/240 [==============================] - 5s 20ms/step - loss: 0.7984 - accuracy: 0.6843 - val_loss: 1.4223 - val_accuracy: 0.5400\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/13\n",
      "240/240 [==============================] - 7s 22ms/step - loss: 1.8425 - accuracy: 0.3617 - val_loss: 1.4595 - val_accuracy: 0.4733\n",
      "Epoch 2/13\n",
      "240/240 [==============================] - 5s 19ms/step - loss: 1.5061 - accuracy: 0.4620 - val_loss: 1.7333 - val_accuracy: 0.4222\n",
      "Epoch 3/13\n",
      "240/240 [==============================] - 5s 22ms/step - loss: 1.3726 - accuracy: 0.4995 - val_loss: 1.5690 - val_accuracy: 0.4311\n",
      "Epoch 4/13\n",
      "240/240 [==============================] - 5s 20ms/step - loss: 1.3022 - accuracy: 0.5249 - val_loss: 1.4643 - val_accuracy: 0.4956\n",
      "Epoch 5/13\n",
      "240/240 [==============================] - 5s 20ms/step - loss: 1.2074 - accuracy: 0.5587 - val_loss: 1.7195 - val_accuracy: 0.4489\n",
      "Epoch 6/13\n",
      "240/240 [==============================] - 5s 22ms/step - loss: 1.1779 - accuracy: 0.5679 - val_loss: 1.6563 - val_accuracy: 0.4156\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/b933d4ce741c44b9851c6443c8aeb40d\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [19]                : (0.3590778708457947, 0.6842927932739258)\n",
      "COMET INFO:     batch_accuracy [456]         : (0.125, 0.75)\n",
      "COMET INFO:     batch_loss [456]             : (0.6482904553413391, 2.811835289001465)\n",
      "COMET INFO:     epoch_duration [19]          : (4.514999999999418, 8.031000000002678)\n",
      "COMET INFO:     loss [19]                    : (0.7984223365783691, 1.8917553424835205)\n",
      "COMET INFO:     val_accuracy [19]            : (0.41555556654930115, 0.5422222018241882)\n",
      "COMET INFO:     val_loss [19]                : (1.4091987609863281, 1.7333006858825684)\n",
      "COMET INFO:     validate_batch_accuracy [38] : (0.34375, 0.625)\n",
      "COMET INFO:     validate_batch_loss [38]     : (1.1905313730239868, 1.995499610900879)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 2\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.7984223365783691\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 550, \"audio_layer_2\": 450, \"batch_size\": 32, \"epochs\": 13, \"final_layer\": 64, \"meta_layer_1\": 550, \"meta_layer_2\": 550}\n",
      "COMET INFO:     optimizer_pid          : 63551b50d609f05fa125d41f2a92bfc1c5f62527\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 1993182\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 550\n",
      "COMET INFO:     audio_layer_2      : 450\n",
      "COMET INFO:     batch_size         : 32\n",
      "COMET INFO:     epochs             : 13\n",
      "COMET INFO:     final_layer        : 64\n",
      "COMET INFO:     meta_layer_1       : 550\n",
      "COMET INFO:     meta_layer_2       : 550\n",
      "COMET INFO:     steps              : 240\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/dffb3ebb561b4bfaaf7d2a84dcd1c181\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/15\n",
      "480/480 [==============================] - 12s 20ms/step - loss: 1.8171 - accuracy: 0.3638 - val_loss: 1.5126 - val_accuracy: 0.4711\n",
      "Epoch 2/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.5228 - accuracy: 0.4526 - val_loss: 1.5432 - val_accuracy: 0.4556\n",
      "Epoch 3/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.3942 - accuracy: 0.4927 - val_loss: 1.4529 - val_accuracy: 0.4867\n",
      "Epoch 4/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.3221 - accuracy: 0.5142 - val_loss: 1.4521 - val_accuracy: 0.4844\n",
      "Epoch 5/15\n",
      "480/480 [==============================] - 9s 20ms/step - loss: 1.2636 - accuracy: 0.5358 - val_loss: 1.3169 - val_accuracy: 0.5356\n",
      "Epoch 6/15\n",
      "480/480 [==============================] - 10s 21ms/step - loss: 1.2189 - accuracy: 0.5514 - val_loss: 1.3400 - val_accuracy: 0.5333\n",
      "Epoch 7/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.1702 - accuracy: 0.5645 - val_loss: 1.3372 - val_accuracy: 0.5378\n",
      "Epoch 8/15\n",
      "480/480 [==============================] - 9s 20ms/step - loss: 1.1357 - accuracy: 0.5798 - val_loss: 1.3189 - val_accuracy: 0.5578\n",
      "Epoch 9/15\n",
      "480/480 [==============================] - 9s 20ms/step - loss: 1.0849 - accuracy: 0.5925 - val_loss: 1.3853 - val_accuracy: 0.5133\n",
      "Epoch 10/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 1.0251 - accuracy: 0.6179 - val_loss: 1.3297 - val_accuracy: 0.5533\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/15\n",
      "480/480 [==============================] - 11s 20ms/step - loss: 1.7943 - accuracy: 0.3724 - val_loss: 1.5269 - val_accuracy: 0.4422\n",
      "Epoch 2/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 1.4973 - accuracy: 0.4538 - val_loss: 1.4779 - val_accuracy: 0.4800\n",
      "Epoch 3/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 1.3533 - accuracy: 0.5064 - val_loss: 1.4521 - val_accuracy: 0.4822\n",
      "Epoch 4/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 1.3123 - accuracy: 0.5194 - val_loss: 1.3809 - val_accuracy: 0.5200\n",
      "Epoch 5/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.2063 - accuracy: 0.5500 - val_loss: 1.3115 - val_accuracy: 0.5311\n",
      "Epoch 6/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.1817 - accuracy: 0.5606 - val_loss: 1.4416 - val_accuracy: 0.4822\n",
      "Epoch 7/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.1302 - accuracy: 0.5667 - val_loss: 1.2904 - val_accuracy: 0.5578\n",
      "Epoch 8/15\n",
      "480/480 [==============================] - 10s 20ms/step - loss: 1.1177 - accuracy: 0.5794 - val_loss: 1.2839 - val_accuracy: 0.5733\n",
      "Epoch 9/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 1.0586 - accuracy: 0.6045 - val_loss: 1.2395 - val_accuracy: 0.5667\n",
      "Epoch 10/15\n",
      "480/480 [==============================] - 9s 20ms/step - loss: 1.0101 - accuracy: 0.6131 - val_loss: 1.2871 - val_accuracy: 0.5356\n",
      "Epoch 11/15\n",
      "480/480 [==============================] - 9s 20ms/step - loss: 0.9678 - accuracy: 0.6253 - val_loss: 1.2447 - val_accuracy: 0.5533\n",
      "Epoch 12/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 0.9165 - accuracy: 0.6446 - val_loss: 1.2135 - val_accuracy: 0.5844\n",
      "Epoch 13/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 0.8854 - accuracy: 0.6582 - val_loss: 1.2931 - val_accuracy: 0.5600\n",
      "Epoch 14/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 0.8356 - accuracy: 0.6650 - val_loss: 1.2807 - val_accuracy: 0.5600\n",
      "Epoch 15/15\n",
      "480/480 [==============================] - 9s 19ms/step - loss: 0.8224 - accuracy: 0.6787 - val_loss: 1.2829 - val_accuracy: 0.5578\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/dffb3ebb561b4bfaaf7d2a84dcd1c181\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [25]                : (0.363766610622406, 0.6786923408508301)\n",
      "COMET INFO:     batch_accuracy [1200]        : (0.0, 0.8125)\n",
      "COMET INFO:     batch_loss [1200]            : (0.6122347116470337, 2.769400119781494)\n",
      "COMET INFO:     epoch_duration [25]          : (9.14100000000326, 11.593000000008033)\n",
      "COMET INFO:     loss [25]                    : (0.8224328756332397, 1.8171194791793823)\n",
      "COMET INFO:     val_accuracy [25]            : (0.4422222077846527, 0.5844444632530212)\n",
      "COMET INFO:     val_loss [25]                : (1.213472843170166, 1.5432382822036743)\n",
      "COMET INFO:     validate_batch_accuracy [50] : (0.375, 0.65625)\n",
      "COMET INFO:     validate_batch_loss [50]     : (1.0114142894744873, 1.8411327600479126)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 3\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.8224328756332397\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 700, \"audio_layer_2\": 400, \"batch_size\": 16, \"epochs\": 15, \"final_layer\": 20, \"meta_layer_1\": 550, \"meta_layer_2\": 128}\n",
      "COMET INFO:     optimizer_pid          : 6baa08e8974f86db1ed39abb76fdc2ac006f84af\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 1759784\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 700\n",
      "COMET INFO:     audio_layer_2      : 400\n",
      "COMET INFO:     batch_size         : 16\n",
      "COMET INFO:     epochs             : 15\n",
      "COMET INFO:     final_layer        : 20\n",
      "COMET INFO:     meta_layer_1       : 550\n",
      "COMET INFO:     meta_layer_2       : 128\n",
      "COMET INFO:     steps              : 480\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 16.57 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/b8cf98d9fa8a4fc1b8abb1d93bfc71e5\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/15\n",
      "480/480 [==============================] - 10s 18ms/step - loss: 1.8195 - accuracy: 0.3741 - val_loss: 1.5941 - val_accuracy: 0.4556\n",
      "Epoch 2/15\n",
      "480/480 [==============================] - 5s 10ms/step - loss: 1.4779 - accuracy: 0.4654 - val_loss: 1.6824 - val_accuracy: 0.4222\n",
      "Epoch 3/15\n",
      "480/480 [==============================] - 5s 10ms/step - loss: 1.3677 - accuracy: 0.5148 - val_loss: 1.3753 - val_accuracy: 0.5400\n",
      "Epoch 4/15\n",
      "480/480 [==============================] - 5s 11ms/step - loss: 1.3236 - accuracy: 0.5328 - val_loss: 1.3879 - val_accuracy: 0.5289\n",
      "Epoch 5/15\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.2785 - accuracy: 0.5479 - val_loss: 1.6082 - val_accuracy: 0.4600\n",
      "Epoch 6/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.1994 - accuracy: 0.5693 - val_loss: 1.6886 - val_accuracy: 0.4378\n",
      "Epoch 7/15\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.1484 - accuracy: 0.5783 - val_loss: 1.5182 - val_accuracy: 0.4867\n",
      "Epoch 8/15\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.1186 - accuracy: 0.5856 - val_loss: 1.6332 - val_accuracy: 0.4467\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/15\n",
      "480/480 [==============================] - 7s 11ms/step - loss: 1.7689 - accuracy: 0.3821 - val_loss: 1.5505 - val_accuracy: 0.4689\n",
      "Epoch 2/15\n",
      "480/480 [==============================] - 5s 11ms/step - loss: 1.4307 - accuracy: 0.4898 - val_loss: 1.5127 - val_accuracy: 0.4889\n",
      "Epoch 3/15\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.3455 - accuracy: 0.5176 - val_loss: 1.5035 - val_accuracy: 0.5156\n",
      "Epoch 4/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.3101 - accuracy: 0.5362 - val_loss: 1.4808 - val_accuracy: 0.4933\n",
      "Epoch 5/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.2424 - accuracy: 0.5612 - val_loss: 1.5298 - val_accuracy: 0.4911\n",
      "Epoch 6/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.2137 - accuracy: 0.5623 - val_loss: 1.3261 - val_accuracy: 0.5711\n",
      "Epoch 7/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.1609 - accuracy: 0.5741 - val_loss: 1.4220 - val_accuracy: 0.5467\n",
      "Epoch 8/15\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.1194 - accuracy: 0.5864 - val_loss: 1.5757 - val_accuracy: 0.4711\n",
      "Epoch 9/15\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.0458 - accuracy: 0.6069 - val_loss: 1.2874 - val_accuracy: 0.5644\n",
      "Epoch 10/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 0.9864 - accuracy: 0.6274 - val_loss: 1.5087 - val_accuracy: 0.5244\n",
      "Epoch 11/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 0.9293 - accuracy: 0.6438 - val_loss: 1.3485 - val_accuracy: 0.5511\n",
      "Epoch 12/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 0.8911 - accuracy: 0.6573 - val_loss: 1.3820 - val_accuracy: 0.5444\n",
      "Epoch 13/15\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 0.8269 - accuracy: 0.6737 - val_loss: 1.3567 - val_accuracy: 0.5289\n",
      "Epoch 14/15\n",
      "480/480 [==============================] - 5s 9ms/step - loss: 0.7942 - accuracy: 0.6879 - val_loss: 1.3933 - val_accuracy: 0.5400\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/b8cf98d9fa8a4fc1b8abb1d93bfc71e5\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [22]                : (0.37405574321746826, 0.6879395842552185)\n",
      "COMET INFO:     batch_accuracy [1056]        : (0.0625, 0.8125)\n",
      "COMET INFO:     batch_loss [1056]            : (0.5057334303855896, 2.5949959754943848)\n",
      "COMET INFO:     epoch_duration [22]          : (3.797000000005937, 10.311999999990803)\n",
      "COMET INFO:     loss [22]                    : (0.7941743731498718, 1.819540023803711)\n",
      "COMET INFO:     val_accuracy [22]            : (0.42222222685813904, 0.5711110830307007)\n",
      "COMET INFO:     val_loss [22]                : (1.2874257564544678, 1.6886146068572998)\n",
      "COMET INFO:     validate_batch_accuracy [44] : (0.34375, 0.6875)\n",
      "COMET INFO:     validate_batch_loss [44]     : (0.9597406983375549, 2.055755615234375)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 4\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.7941743731498718\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 900, \"audio_layer_2\": 256, \"batch_size\": 16, \"epochs\": 15, \"final_layer\": 256, \"meta_layer_1\": 700, \"meta_layer_2\": 500}\n",
      "COMET INFO:     optimizer_pid          : 4277de37e73853eccbc5be8443b2e876ffcc31dd\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 2208536\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 900\n",
      "COMET INFO:     audio_layer_2      : 256\n",
      "COMET INFO:     batch_size         : 16\n",
      "COMET INFO:     epochs             : 15\n",
      "COMET INFO:     final_layer        : 256\n",
      "COMET INFO:     meta_layer_1       : 700\n",
      "COMET INFO:     meta_layer_2       : 500\n",
      "COMET INFO:     steps              : 480\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/0ccec2e245c04c2590fe2fcc5d6fdbbf\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/13\n",
      "240/240 [==============================] - 4s 11ms/step - loss: 1.8184 - accuracy: 0.3708 - val_loss: 1.4261 - val_accuracy: 0.4889\n",
      "Epoch 2/13\n",
      "240/240 [==============================] - 3s 11ms/step - loss: 1.4766 - accuracy: 0.4651 - val_loss: 1.5633 - val_accuracy: 0.4822\n",
      "Epoch 3/13\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.3100 - accuracy: 0.5296 - val_loss: 1.5496 - val_accuracy: 0.4667\n",
      "Epoch 4/13\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.2159 - accuracy: 0.5529 - val_loss: 1.4333 - val_accuracy: 0.5067\n",
      "Epoch 5/13\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.1487 - accuracy: 0.5749 - val_loss: 1.5281 - val_accuracy: 0.4489\n",
      "Epoch 6/13\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.0809 - accuracy: 0.5991 - val_loss: 1.4667 - val_accuracy: 0.5289\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/13\n",
      "240/240 [==============================] - 6s 20ms/step - loss: 1.8791 - accuracy: 0.3532 - val_loss: 1.5061 - val_accuracy: 0.4467\n",
      "Epoch 2/13\n",
      "240/240 [==============================] - 4s 18ms/step - loss: 1.4868 - accuracy: 0.4689 - val_loss: 1.7021 - val_accuracy: 0.4200\n",
      "Epoch 3/13\n",
      "240/240 [==============================] - 5s 19ms/step - loss: 1.3279 - accuracy: 0.5203 - val_loss: 1.5153 - val_accuracy: 0.4778\n",
      "Epoch 4/13\n",
      "240/240 [==============================] - 5s 19ms/step - loss: 1.2157 - accuracy: 0.5612 - val_loss: 1.4080 - val_accuracy: 0.5333\n",
      "Epoch 5/13\n",
      "240/240 [==============================] - 5s 19ms/step - loss: 1.1614 - accuracy: 0.5748 - val_loss: 1.5363 - val_accuracy: 0.4556\n",
      "Epoch 6/13\n",
      "240/240 [==============================] - 5s 19ms/step - loss: 1.0924 - accuracy: 0.6011 - val_loss: 1.4887 - val_accuracy: 0.4911\n",
      "Epoch 7/13\n",
      "240/240 [==============================] - 3s 14ms/step - loss: 1.0331 - accuracy: 0.6168 - val_loss: 1.4639 - val_accuracy: 0.5178\n",
      "Epoch 8/13\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 0.9805 - accuracy: 0.6330 - val_loss: 1.4922 - val_accuracy: 0.5044\n",
      "Epoch 9/13\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.9065 - accuracy: 0.6545 - val_loss: 1.4099 - val_accuracy: 0.5200\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/0ccec2e245c04c2590fe2fcc5d6fdbbf\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [15]                : (0.3532169759273529, 0.6544672846794128)\n",
      "COMET INFO:     batch_accuracy [360]         : (0.0625, 0.6915322542190552)\n",
      "COMET INFO:     batch_loss [360]             : (0.7774025201797485, 3.1096343994140625)\n",
      "COMET INFO:     epoch_duration [15]          : (2.0309999999881256, 6.375)\n",
      "COMET INFO:     loss [15]                    : (0.906499981880188, 1.8791028261184692)\n",
      "COMET INFO:     val_accuracy [15]            : (0.41999998688697815, 0.5333333611488342)\n",
      "COMET INFO:     val_loss [15]                : (1.4080137014389038, 1.7021414041519165)\n",
      "COMET INFO:     validate_batch_accuracy [30] : (0.3125, 0.5625)\n",
      "COMET INFO:     validate_batch_loss [30]     : (1.119657039642334, 1.7564234733581543)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 5\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.906499981880188\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 900, \"audio_layer_2\": 256, \"batch_size\": 32, \"epochs\": 13, \"final_layer\": 64, \"meta_layer_1\": 1000, \"meta_layer_2\": 500}\n",
      "COMET INFO:     optimizer_pid          : b6437a14b1033434557dfaa09dee979e0f82bdc9\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 2211320\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 900\n",
      "COMET INFO:     audio_layer_2      : 256\n",
      "COMET INFO:     batch_size         : 32\n",
      "COMET INFO:     epochs             : 13\n",
      "COMET INFO:     final_layer        : 64\n",
      "COMET INFO:     meta_layer_1       : 1000\n",
      "COMET INFO:     meta_layer_2       : 500\n",
      "COMET INFO:     steps              : 240\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/aec17d0080364183989e46bcdca227df\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/14\n",
      "120/120 [==============================] - 2s 12ms/step - loss: 1.9465 - accuracy: 0.3089 - val_loss: 1.5414 - val_accuracy: 0.4356\n",
      "Epoch 2/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.5544 - accuracy: 0.4411 - val_loss: 1.4574 - val_accuracy: 0.4867\n",
      "Epoch 3/14\n",
      "120/120 [==============================] - 1s 12ms/step - loss: 1.3873 - accuracy: 0.4876 - val_loss: 1.4523 - val_accuracy: 0.5089\n",
      "Epoch 4/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.2959 - accuracy: 0.5225 - val_loss: 1.4210 - val_accuracy: 0.4978\n",
      "Epoch 5/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.2006 - accuracy: 0.5586 - val_loss: 1.4165 - val_accuracy: 0.5089\n",
      "Epoch 6/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.1163 - accuracy: 0.5828 - val_loss: 1.4158 - val_accuracy: 0.5178\n",
      "Epoch 7/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.0539 - accuracy: 0.6098 - val_loss: 1.3224 - val_accuracy: 0.5467\n",
      "Epoch 8/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.9848 - accuracy: 0.6331 - val_loss: 1.3768 - val_accuracy: 0.5533\n",
      "Epoch 9/14\n",
      "120/120 [==============================] - 1s 10ms/step - loss: 0.9267 - accuracy: 0.6538 - val_loss: 1.2695 - val_accuracy: 0.5489\n",
      "Epoch 10/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.8673 - accuracy: 0.6687 - val_loss: 1.3578 - val_accuracy: 0.5400\n",
      "Epoch 11/14\n",
      "120/120 [==============================] - 1s 10ms/step - loss: 0.8374 - accuracy: 0.6778 - val_loss: 1.2611 - val_accuracy: 0.5622\n",
      "Epoch 12/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 0.8050 - accuracy: 0.6918 - val_loss: 1.2807 - val_accuracy: 0.5400\n",
      "Epoch 13/14\n",
      "120/120 [==============================] - 2s 18ms/step - loss: 0.7759 - accuracy: 0.6987 - val_loss: 1.2492 - val_accuracy: 0.5556\n",
      "Epoch 14/14\n",
      "120/120 [==============================] - 2s 18ms/step - loss: 0.7528 - accuracy: 0.7050 - val_loss: 1.2831 - val_accuracy: 0.5467\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/14\n",
      "120/120 [==============================] - 4s 22ms/step - loss: 1.9045 - accuracy: 0.3235 - val_loss: 1.4489 - val_accuracy: 0.4622\n",
      "Epoch 2/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 1.5557 - accuracy: 0.4349 - val_loss: 1.4672 - val_accuracy: 0.4956\n",
      "Epoch 3/14\n",
      "120/120 [==============================] - 2s 20ms/step - loss: 1.3820 - accuracy: 0.4897 - val_loss: 1.4537 - val_accuracy: 0.5044\n",
      "Epoch 4/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 1.2682 - accuracy: 0.5259 - val_loss: 1.4103 - val_accuracy: 0.5000\n",
      "Epoch 5/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 1.1808 - accuracy: 0.5610 - val_loss: 1.4425 - val_accuracy: 0.4667\n",
      "Epoch 6/14\n",
      "120/120 [==============================] - 2s 18ms/step - loss: 1.0958 - accuracy: 0.5909 - val_loss: 1.3919 - val_accuracy: 0.5044\n",
      "Epoch 7/14\n",
      "120/120 [==============================] - 2s 18ms/step - loss: 1.0286 - accuracy: 0.6095 - val_loss: 1.3444 - val_accuracy: 0.5378\n",
      "Epoch 8/14\n",
      "120/120 [==============================] - 2s 18ms/step - loss: 0.9723 - accuracy: 0.6382 - val_loss: 1.4085 - val_accuracy: 0.5089\n",
      "Epoch 9/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 0.9124 - accuracy: 0.6517 - val_loss: 1.4326 - val_accuracy: 0.4889\n",
      "Epoch 10/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 0.8606 - accuracy: 0.6670 - val_loss: 1.3335 - val_accuracy: 0.5378\n",
      "Epoch 11/14\n",
      "120/120 [==============================] - 2s 18ms/step - loss: 0.8127 - accuracy: 0.6809 - val_loss: 1.3064 - val_accuracy: 0.5289\n",
      "Epoch 12/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 0.7878 - accuracy: 0.6951 - val_loss: 1.3047 - val_accuracy: 0.5333\n",
      "Epoch 13/14\n",
      "120/120 [==============================] - 2s 18ms/step - loss: 0.7552 - accuracy: 0.7040 - val_loss: 1.3224 - val_accuracy: 0.5333\n",
      "Epoch 14/14\n",
      "120/120 [==============================] - 2s 19ms/step - loss: 0.7300 - accuracy: 0.7110 - val_loss: 1.3304 - val_accuracy: 0.5289\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/aec17d0080364183989e46bcdca227df\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [28]                : (0.30893462896347046, 0.7109924554824829)\n",
      "COMET INFO:     batch_accuracy [336]         : (0.0625, 0.734375)\n",
      "COMET INFO:     batch_loss [336]             : (0.6267034411430359, 2.5626168251037598)\n",
      "COMET INFO:     epoch_duration [28]          : (1.1089999999967404, 4.297000000005937)\n",
      "COMET INFO:     loss [28]                    : (0.7300055027008057, 1.9464826583862305)\n",
      "COMET INFO:     val_accuracy [28]            : (0.4355555474758148, 0.5622222423553467)\n",
      "COMET INFO:     val_loss [28]                : (1.2492315769195557, 1.5413715839385986)\n",
      "COMET INFO:     validate_batch_accuracy [56] : (0.40625, 0.6875)\n",
      "COMET INFO:     validate_batch_loss [56]     : (1.0571486949920654, 1.911614179611206)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 6\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.7300055027008057\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 550, \"audio_layer_2\": 300, \"batch_size\": 64, \"epochs\": 14, \"final_layer\": 32, \"meta_layer_1\": 600, \"meta_layer_2\": 128}\n",
      "COMET INFO:     optimizer_pid          : 107a20f8b115f3eb941d77e152fa93fa35990b97\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 1630874\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 550\n",
      "COMET INFO:     audio_layer_2      : 300\n",
      "COMET INFO:     batch_size         : 64\n",
      "COMET INFO:     epochs             : 14\n",
      "COMET INFO:     final_layer        : 32\n",
      "COMET INFO:     meta_layer_1       : 600\n",
      "COMET INFO:     meta_layer_2       : 128\n",
      "COMET INFO:     steps              : 120\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/12232c8e6e3d44bfbe600f3d46354cd1\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/20\n",
      "192/192 [==============================] - 5s 20ms/step - loss: 1.8239 - accuracy: 0.3751 - val_loss: 1.4755 - val_accuracy: 0.4622\n",
      "Epoch 2/20\n",
      "192/192 [==============================] - 3s 17ms/step - loss: 1.4432 - accuracy: 0.4703 - val_loss: 1.4785 - val_accuracy: 0.4978\n",
      "Epoch 3/20\n",
      "192/192 [==============================] - 2s 10ms/step - loss: 1.2660 - accuracy: 0.5384 - val_loss: 1.6179 - val_accuracy: 0.4556\n",
      "Epoch 4/20\n",
      "192/192 [==============================] - 2s 10ms/step - loss: 1.1953 - accuracy: 0.5653 - val_loss: 1.4123 - val_accuracy: 0.5111\n",
      "Epoch 5/20\n",
      "192/192 [==============================] - 2s 11ms/step - loss: 1.0944 - accuracy: 0.5989 - val_loss: 1.6286 - val_accuracy: 0.4311\n",
      "Epoch 6/20\n",
      "192/192 [==============================] - 2s 11ms/step - loss: 1.0366 - accuracy: 0.6140 - val_loss: 1.6863 - val_accuracy: 0.4378\n",
      "Epoch 7/20\n",
      "192/192 [==============================] - 3s 14ms/step - loss: 1.0070 - accuracy: 0.6310 - val_loss: 1.4472 - val_accuracy: 0.5178\n",
      "Epoch 8/20\n",
      "192/192 [==============================] - 2s 12ms/step - loss: 0.9729 - accuracy: 0.6395 - val_loss: 1.5555 - val_accuracy: 0.5156\n",
      "Epoch 9/20\n",
      "192/192 [==============================] - 2s 10ms/step - loss: 0.9463 - accuracy: 0.6437 - val_loss: 1.5304 - val_accuracy: 0.5333\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/20\n",
      "192/192 [==============================] - 4s 14ms/step - loss: 1.8149 - accuracy: 0.3715 - val_loss: 1.3830 - val_accuracy: 0.5200\n",
      "Epoch 2/20\n",
      "192/192 [==============================] - 2s 12ms/step - loss: 1.4360 - accuracy: 0.4792 - val_loss: 1.4411 - val_accuracy: 0.5133\n",
      "Epoch 3/20\n",
      "192/192 [==============================] - 3s 17ms/step - loss: 1.2767 - accuracy: 0.5366 - val_loss: 1.3569 - val_accuracy: 0.5267\n",
      "Epoch 4/20\n",
      "192/192 [==============================] - 2s 9ms/step - loss: 1.1564 - accuracy: 0.5736 - val_loss: 1.4063 - val_accuracy: 0.5289\n",
      "Epoch 5/20\n",
      "192/192 [==============================] - 2s 11ms/step - loss: 1.0675 - accuracy: 0.6000 - val_loss: 1.4295 - val_accuracy: 0.5333\n",
      "Epoch 6/20\n",
      "192/192 [==============================] - 2s 11ms/step - loss: 1.0231 - accuracy: 0.6214 - val_loss: 1.4247 - val_accuracy: 0.5267\n",
      "Epoch 7/20\n",
      "192/192 [==============================] - 2s 10ms/step - loss: 0.9965 - accuracy: 0.6340 - val_loss: 1.4597 - val_accuracy: 0.4933\n",
      "Epoch 8/20\n",
      "192/192 [==============================] - 2s 10ms/step - loss: 0.9873 - accuracy: 0.6323 - val_loss: 1.5859 - val_accuracy: 0.5111\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/12232c8e6e3d44bfbe600f3d46354cd1\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [17]                : (0.37145090103149414, 0.6436572074890137)\n",
      "COMET INFO:     batch_accuracy [340]         : (0.07500000298023224, 0.699999988079071)\n",
      "COMET INFO:     batch_loss [340]             : (0.5899624824523926, 2.5186054706573486)\n",
      "COMET INFO:     epoch_duration [17]          : (1.7660000000032596, 5.4689999999973224)\n",
      "COMET INFO:     loss [17]                    : (0.9462659358978271, 1.8238894939422607)\n",
      "COMET INFO:     val_accuracy [17]            : (0.43111109733581543, 0.5333333611488342)\n",
      "COMET INFO:     val_loss [17]                : (1.3568576574325562, 1.6862506866455078)\n",
      "COMET INFO:     validate_batch_accuracy [34] : (0.375, 0.75)\n",
      "COMET INFO:     validate_batch_loss [34]     : (1.0880088806152344, 1.9522091150283813)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 7\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.9462659358978271\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 500, \"audio_layer_2\": 400, \"batch_size\": 40, \"epochs\": 20, \"final_layer\": 256, \"meta_layer_1\": 900, \"meta_layer_2\": 400}\n",
      "COMET INFO:     optimizer_pid          : 9ed5cbd3afdde1d72fc8b6cb1849e6e49ea26d76\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 2139796\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 500\n",
      "COMET INFO:     audio_layer_2      : 400\n",
      "COMET INFO:     batch_size         : 40\n",
      "COMET INFO:     epochs             : 20\n",
      "COMET INFO:     final_layer        : 256\n",
      "COMET INFO:     meta_layer_1       : 900\n",
      "COMET INFO:     meta_layer_2       : 400\n",
      "COMET INFO:     steps              : 192\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/ec54896543c2421ba9f7270661b6cec1\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/20\n",
      "240/240 [==============================] - 4s 11ms/step - loss: 1.9094 - accuracy: 0.3200 - val_loss: 1.6253 - val_accuracy: 0.4133\n",
      "Epoch 2/20\n",
      "240/240 [==============================] - 3s 11ms/step - loss: 1.5861 - accuracy: 0.4195 - val_loss: 1.6488 - val_accuracy: 0.4511\n",
      "Epoch 3/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.4245 - accuracy: 0.4760 - val_loss: 1.6215 - val_accuracy: 0.4733\n",
      "Epoch 4/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.3298 - accuracy: 0.5150 - val_loss: 1.6222 - val_accuracy: 0.4378\n",
      "Epoch 5/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.2249 - accuracy: 0.5387 - val_loss: 1.4256 - val_accuracy: 0.5111\n",
      "Epoch 6/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.1678 - accuracy: 0.5675 - val_loss: 1.5833 - val_accuracy: 0.4689\n",
      "Epoch 7/20\n",
      "240/240 [==============================] - 3s 10ms/step - loss: 1.1105 - accuracy: 0.5835 - val_loss: 1.5512 - val_accuracy: 0.4644\n",
      "Epoch 8/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.0650 - accuracy: 0.6059 - val_loss: 1.4929 - val_accuracy: 0.5067\n",
      "Epoch 9/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.0029 - accuracy: 0.6244 - val_loss: 1.5419 - val_accuracy: 0.5022\n",
      "Epoch 10/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 0.9886 - accuracy: 0.6241 - val_loss: 1.4476 - val_accuracy: 0.5178\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/20\n",
      "240/240 [==============================] - 4s 11ms/step - loss: 1.8352 - accuracy: 0.3563 - val_loss: 1.5584 - val_accuracy: 0.4378\n",
      "Epoch 2/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.5456 - accuracy: 0.4488 - val_loss: 1.5566 - val_accuracy: 0.4400\n",
      "Epoch 3/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.4028 - accuracy: 0.4956 - val_loss: 1.6011 - val_accuracy: 0.4600\n",
      "Epoch 4/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.2969 - accuracy: 0.5314 - val_loss: 1.4665 - val_accuracy: 0.4844\n",
      "Epoch 5/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.2317 - accuracy: 0.5494 - val_loss: 1.6084 - val_accuracy: 0.4756\n",
      "Epoch 6/20\n",
      "240/240 [==============================] - 2s 10ms/step - loss: 1.1513 - accuracy: 0.5762 - val_loss: 1.4382 - val_accuracy: 0.5111\n",
      "Epoch 7/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.1082 - accuracy: 0.5815 - val_loss: 1.4916 - val_accuracy: 0.5200\n",
      "Epoch 8/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.0561 - accuracy: 0.6065 - val_loss: 1.3912 - val_accuracy: 0.5600\n",
      "Epoch 9/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 1.0239 - accuracy: 0.6210 - val_loss: 1.5255 - val_accuracy: 0.4822\n",
      "Epoch 10/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 0.9645 - accuracy: 0.6302 - val_loss: 1.4591 - val_accuracy: 0.5378\n",
      "Epoch 11/20\n",
      "240/240 [==============================] - 3s 10ms/step - loss: 0.9549 - accuracy: 0.6355 - val_loss: 1.4367 - val_accuracy: 0.5156\n",
      "Epoch 12/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 0.9089 - accuracy: 0.6512 - val_loss: 1.3956 - val_accuracy: 0.5489\n",
      "Epoch 13/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.8657 - accuracy: 0.6685 - val_loss: 1.5529 - val_accuracy: 0.4956\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/ec54896543c2421ba9f7270661b6cec1\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [23]                : (0.3200052082538605, 0.668533444404602)\n",
      "COMET INFO:     batch_accuracy [552]         : (0.125, 0.75)\n",
      "COMET INFO:     batch_loss [552]             : (0.808335542678833, 2.5297203063964844)\n",
      "COMET INFO:     epoch_duration [23]          : (1.9380000000091968, 3.7810000000026776)\n",
      "COMET INFO:     loss [23]                    : (0.8657002449035645, 1.9094011783599854)\n",
      "COMET INFO:     val_accuracy [23]            : (0.41333332657814026, 0.5600000023841858)\n",
      "COMET INFO:     val_loss [23]                : (1.3912405967712402, 1.6488409042358398)\n",
      "COMET INFO:     validate_batch_accuracy [46] : (0.375, 0.6875)\n",
      "COMET INFO:     validate_batch_loss [46]     : (1.0947449207305908, 1.908547282218933)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 8\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.8657002449035645\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 550, \"audio_layer_2\": 400, \"batch_size\": 32, \"epochs\": 20, \"final_layer\": 20, \"meta_layer_1\": 500, \"meta_layer_2\": 500}\n",
      "COMET INFO:     optimizer_pid          : 57ea10a92749762f228b629a0997fde994318a4b\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 1865534\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 550\n",
      "COMET INFO:     audio_layer_2      : 400\n",
      "COMET INFO:     batch_size         : 32\n",
      "COMET INFO:     epochs             : 20\n",
      "COMET INFO:     final_layer        : 20\n",
      "COMET INFO:     meta_layer_1       : 500\n",
      "COMET INFO:     meta_layer_2       : 500\n",
      "COMET INFO:     steps              : 240\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/41bbbb907c054e90b53374340f38e532\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/14\n",
      "480/480 [==============================] - 6s 10ms/step - loss: 1.8687 - accuracy: 0.3588 - val_loss: 1.6932 - val_accuracy: 0.3733\n",
      "Epoch 2/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.5476 - accuracy: 0.4430 - val_loss: 1.5552 - val_accuracy: 0.4356\n",
      "Epoch 3/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.3915 - accuracy: 0.4858 - val_loss: 1.4765 - val_accuracy: 0.4889\n",
      "Epoch 4/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.2851 - accuracy: 0.5280 - val_loss: 1.4627 - val_accuracy: 0.5178\n",
      "Epoch 5/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.2137 - accuracy: 0.5465 - val_loss: 1.4252 - val_accuracy: 0.5378\n",
      "Epoch 6/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.1638 - accuracy: 0.5647 - val_loss: 1.6265 - val_accuracy: 0.4489\n",
      "Epoch 7/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.1153 - accuracy: 0.5787 - val_loss: 1.3889 - val_accuracy: 0.5244\n",
      "Epoch 8/14\n",
      "480/480 [==============================] - 5s 10ms/step - loss: 1.0565 - accuracy: 0.6037 - val_loss: 1.4298 - val_accuracy: 0.4844\n",
      "Epoch 9/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.0104 - accuracy: 0.6229 - val_loss: 1.3087 - val_accuracy: 0.5356\n",
      "Epoch 10/14\n",
      "480/480 [==============================] - 5s 10ms/step - loss: 0.9555 - accuracy: 0.6352 - val_loss: 1.2181 - val_accuracy: 0.5933\n",
      "Epoch 11/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 0.8929 - accuracy: 0.6555 - val_loss: 1.3381 - val_accuracy: 0.5400\n",
      "Epoch 12/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 0.8641 - accuracy: 0.6623 - val_loss: 1.3247 - val_accuracy: 0.5489\n",
      "Epoch 13/14\n",
      "480/480 [==============================] - 5s 10ms/step - loss: 0.8264 - accuracy: 0.6795 - val_loss: 1.3395 - val_accuracy: 0.5444\n",
      "Epoch 14/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 0.8005 - accuracy: 0.6865 - val_loss: 1.3318 - val_accuracy: 0.5400\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/14\n",
      "480/480 [==============================] - 6s 10ms/step - loss: 1.8137 - accuracy: 0.3599 - val_loss: 1.5983 - val_accuracy: 0.4467\n",
      "Epoch 2/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.4935 - accuracy: 0.4508 - val_loss: 1.5781 - val_accuracy: 0.4400\n",
      "Epoch 3/14\n",
      "480/480 [==============================] - 5s 10ms/step - loss: 1.3735 - accuracy: 0.5014 - val_loss: 1.5684 - val_accuracy: 0.4444\n",
      "Epoch 4/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.2975 - accuracy: 0.5160 - val_loss: 1.4873 - val_accuracy: 0.4911\n",
      "Epoch 5/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.2308 - accuracy: 0.5375 - val_loss: 1.4867 - val_accuracy: 0.5111\n",
      "Epoch 6/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.1458 - accuracy: 0.5759 - val_loss: 1.2478 - val_accuracy: 0.5689\n",
      "Epoch 7/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 1.1161 - accuracy: 0.5823 - val_loss: 1.3682 - val_accuracy: 0.5067\n",
      "Epoch 8/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 1.0440 - accuracy: 0.6055 - val_loss: 1.3106 - val_accuracy: 0.5422\n",
      "Epoch 9/14\n",
      "480/480 [==============================] - 4s 8ms/step - loss: 0.9782 - accuracy: 0.6301 - val_loss: 1.4228 - val_accuracy: 0.5200\n",
      "Epoch 10/14\n",
      "480/480 [==============================] - 5s 10ms/step - loss: 0.9329 - accuracy: 0.6371 - val_loss: 1.3229 - val_accuracy: 0.5422\n",
      "Epoch 11/14\n",
      "480/480 [==============================] - 4s 9ms/step - loss: 0.8767 - accuracy: 0.6636 - val_loss: 1.3175 - val_accuracy: 0.5467\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/41bbbb907c054e90b53374340f38e532\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [25]                : (0.3588173985481262, 0.6865069270133972)\n",
      "COMET INFO:     batch_accuracy [1200]        : (0.0625, 0.875)\n",
      "COMET INFO:     batch_loss [1200]            : (0.4730539619922638, 2.784681797027588)\n",
      "COMET INFO:     epoch_duration [25]          : (3.7339999999967404, 6.10899999999674)\n",
      "COMET INFO:     loss [25]                    : (0.8005383610725403, 1.8686538934707642)\n",
      "COMET INFO:     val_accuracy [25]            : (0.3733333349227905, 0.5933333039283752)\n",
      "COMET INFO:     val_loss [25]                : (1.2181159257888794, 1.6931582689285278)\n",
      "COMET INFO:     validate_batch_accuracy [50] : (0.34375, 0.6875)\n",
      "COMET INFO:     validate_batch_loss [50]     : (1.0125460624694824, 1.7873506546020508)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 9\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.8005383610725403\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 550, \"audio_layer_2\": 550, \"batch_size\": 16, \"epochs\": 14, \"final_layer\": 20, \"meta_layer_1\": 550, \"meta_layer_2\": 500}\n",
      "COMET INFO:     optimizer_pid          : 4c26ef39d2110f2ea1d443faf8996930e9fc42fc\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 1977584\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 550\n",
      "COMET INFO:     audio_layer_2      : 550\n",
      "COMET INFO:     batch_size         : 16\n",
      "COMET INFO:     epochs             : 14\n",
      "COMET INFO:     final_layer        : 20\n",
      "COMET INFO:     meta_layer_1       : 550\n",
      "COMET INFO:     meta_layer_2       : 500\n",
      "COMET INFO:     steps              : 480\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET WARNING: As you are running in a Jupyter environment, you will need to call `experiment.end()` when finished to ensure all metrics and code are logged before exiting.\n",
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/e64c2fdef597479ebe88b30d6cc82544\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/20\n",
      "240/240 [==============================] - 3s 10ms/step - loss: 1.7918 - accuracy: 0.3832 - val_loss: 1.4246 - val_accuracy: 0.5178\n",
      "Epoch 2/20\n",
      "240/240 [==============================] - 2s 10ms/step - loss: 1.4503 - accuracy: 0.4753 - val_loss: 1.5492 - val_accuracy: 0.4867\n",
      "Epoch 3/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.2924 - accuracy: 0.5255 - val_loss: 1.4941 - val_accuracy: 0.4756\n",
      "Epoch 4/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.1624 - accuracy: 0.5742 - val_loss: 1.4789 - val_accuracy: 0.4800\n",
      "Epoch 5/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.0847 - accuracy: 0.5964 - val_loss: 1.4036 - val_accuracy: 0.5022\n",
      "Epoch 6/20\n",
      "240/240 [==============================] - 2s 10ms/step - loss: 1.0050 - accuracy: 0.6192 - val_loss: 1.4876 - val_accuracy: 0.4644\n",
      "Epoch 7/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.9462 - accuracy: 0.6447 - val_loss: 1.3256 - val_accuracy: 0.5467\n",
      "Epoch 8/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 0.9125 - accuracy: 0.6528 - val_loss: 1.4761 - val_accuracy: 0.5089\n",
      "Epoch 9/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 0.8848 - accuracy: 0.6607 - val_loss: 1.4404 - val_accuracy: 0.5156\n",
      "Epoch 10/20\n",
      "240/240 [==============================] - 2s 8ms/step - loss: 0.8767 - accuracy: 0.6674 - val_loss: 1.5055 - val_accuracy: 0.5044\n",
      "Epoch 11/20\n",
      "240/240 [==============================] - 3s 11ms/step - loss: 0.8406 - accuracy: 0.6720 - val_loss: 1.4368 - val_accuracy: 0.5111\n",
      "Epoch 12/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.7835 - accuracy: 0.6925 - val_loss: 1.4862 - val_accuracy: 0.5467\n",
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/20\n",
      "240/240 [==============================] - 3s 9ms/step - loss: 1.8265 - accuracy: 0.3794 - val_loss: 1.4521 - val_accuracy: 0.5089\n",
      "Epoch 2/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.4768 - accuracy: 0.4729 - val_loss: 1.4934 - val_accuracy: 0.4800\n",
      "Epoch 3/20\n",
      "240/240 [==============================] - 3s 11ms/step - loss: 1.2925 - accuracy: 0.5215 - val_loss: 1.6065 - val_accuracy: 0.4556\n",
      "Epoch 4/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.1904 - accuracy: 0.5595 - val_loss: 1.4967 - val_accuracy: 0.5067\n",
      "Epoch 5/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 1.0735 - accuracy: 0.5978 - val_loss: 1.4828 - val_accuracy: 0.4711\n",
      "Epoch 6/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.9883 - accuracy: 0.6258 - val_loss: 1.4100 - val_accuracy: 0.4978\n",
      "Epoch 7/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.9567 - accuracy: 0.6343 - val_loss: 1.3106 - val_accuracy: 0.5578\n",
      "Epoch 8/20\n",
      "240/240 [==============================] - 2s 10ms/step - loss: 0.8898 - accuracy: 0.6508 - val_loss: 1.5807 - val_accuracy: 0.5111\n",
      "Epoch 9/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.9007 - accuracy: 0.6569 - val_loss: 1.4211 - val_accuracy: 0.5356\n",
      "Epoch 10/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.8655 - accuracy: 0.6689 - val_loss: 1.5761 - val_accuracy: 0.4644\n",
      "Epoch 11/20\n",
      "240/240 [==============================] - 2s 9ms/step - loss: 0.8455 - accuracy: 0.6694 - val_loss: 1.4597 - val_accuracy: 0.5356\n",
      "Epoch 12/20\n",
      "240/240 [==============================] - 3s 11ms/step - loss: 0.7692 - accuracy: 0.6981 - val_loss: 1.4871 - val_accuracy: 0.5222\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     display_summary_level : 1\n",
      "COMET INFO:     url                   : https://www.comet.ml/wodenwang820118/music-genre-multiclass-classification/e64c2fdef597479ebe88b30d6cc82544\n",
      "COMET INFO:   Metrics [count] (min, max):\n",
      "COMET INFO:     accuracy [24]                : (0.37939566373825073, 0.6980984807014465)\n",
      "COMET INFO:     batch_accuracy [576]         : (0.15625, 0.78125)\n",
      "COMET INFO:     batch_loss [576]             : (0.43188461661338806, 2.5712742805480957)\n",
      "COMET INFO:     epoch_duration [24]          : (2.0, 3.4530000000086147)\n",
      "COMET INFO:     loss [24]                    : (0.7691603302955627, 1.826533555984497)\n",
      "COMET INFO:     val_accuracy [24]            : (0.4555555582046509, 0.5577777624130249)\n",
      "COMET INFO:     val_loss [24]                : (1.3105894327163696, 1.6065045595169067)\n",
      "COMET INFO:     validate_batch_accuracy [48] : (0.375, 0.71875)\n",
      "COMET INFO:     validate_batch_loss [48]     : (1.0258684158325195, 2.0093631744384766)\n",
      "COMET INFO:   Others:\n",
      "COMET INFO:     optimizer_count        : 10\n",
      "COMET INFO:     optimizer_id           : ac9577df43cc4625a3b73696e8632eb2\n",
      "COMET INFO:     optimizer_metric       : loss\n",
      "COMET INFO:     optimizer_metric_value : 0.7691603302955627\n",
      "COMET INFO:     optimizer_objective    : minimum\n",
      "COMET INFO:     optimizer_parameters   : {\"audio_layer_1\": 1000, \"audio_layer_2\": 500, \"batch_size\": 32, \"epochs\": 20, \"final_layer\": 64, \"meta_layer_1\": 900, \"meta_layer_2\": 300}\n",
      "COMET INFO:     optimizer_pid          : 768453565823ea37b1d22a1599fcf30e6806e90b\n",
      "COMET INFO:     optimizer_process      : 25788\n",
      "COMET INFO:     optimizer_trial        : 1\n",
      "COMET INFO:     optimizer_version      : 2.0.1\n",
      "COMET INFO:     trainable_params       : 2269832\n",
      "COMET INFO:   Parameters:\n",
      "COMET INFO:     Adam_amsgrad       : False\n",
      "COMET INFO:     Adam_beta_1        : 0.8999999761581421\n",
      "COMET INFO:     Adam_beta_2        : 0.9990000128746033\n",
      "COMET INFO:     Adam_decay         : 0.0\n",
      "COMET INFO:     Adam_epsilon       : 1e-07\n",
      "COMET INFO:     Adam_learning_rate : 9.999999747378752e-05\n",
      "COMET INFO:     Adam_name          : Adam\n",
      "COMET INFO:     Optimizer          : Adam\n",
      "COMET INFO:     audio_layer_1      : 1000\n",
      "COMET INFO:     audio_layer_2      : 500\n",
      "COMET INFO:     batch_size         : 32\n",
      "COMET INFO:     epochs             : 20\n",
      "COMET INFO:     final_layer        : 64\n",
      "COMET INFO:     meta_layer_1       : 900\n",
      "COMET INFO:     meta_layer_2       : 300\n",
      "COMET INFO:     steps              : 240\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:     environment details      : 1\n",
      "COMET INFO:     filename                 : 1\n",
      "COMET INFO:     git metadata             : 1\n",
      "COMET INFO:     git-patch (uncompressed) : 1 (826 bytes)\n",
      "COMET INFO:     installed packages       : 1\n",
      "COMET INFO:     model graph              : 1\n",
      "COMET INFO:     notebook                 : 1\n",
      "COMET INFO:     source_code              : 1\n",
      "COMET INFO: ---------------------------\n",
      "COMET INFO: Uploading 1 metrics, params and output messages\n",
      "COMET INFO: Waiting for completion of the file uploads (may take several seconds)\n",
      "COMET INFO: The Python SDK has 10800 seconds to finish before aborting...\n",
      "COMET INFO: Still uploading 2 file(s), remaining 165.75 KB/166.34 KB\n",
      "COMET INFO: Optimizer search ac9577df43cc4625a3b73696e8632eb2 has completed\n"
     ]
    }
   ],
   "source": [
    "classifier.grid_search({\n",
    "    \"algorithm\": \"bayes\",\n",
    "    \"name\": \"Optimize Music Classification Network\",\n",
    "    \"spec\": {\"maxCombo\": 10, \"objective\": \"minimize\", \"metric\": \"loss\"},\n",
    "    \"parameters\": {\n",
    "        \"audio_layer_1\": {\"type\": \"discrete\", \"values\": [450,500,550,600,700,900,950,1000]},\n",
    "        \"audio_layer_2\": {\"type\": \"discrete\", \"values\": [128,256,300,400,450,500,550]},\n",
    "        \"meta_layer_1\": {\"type\": \"discrete\", \"values\": [450,500,550,600,700,900,1000]},\n",
    "        \"meta_layer_2\": {\"type\": \"discrete\", \"values\": [128,256,300,400,450,500,550]},\n",
    "        \"final_layer\": {\"type\": \"discrete\", \"values\": [20,32,64,100,128,256]},\n",
    "        \"batch_size\": {\"type\": \"discrete\", \"values\": [16,32,40,64]},\n",
    "        \"epochs\": {\"type\": \"discrete\", \"values\": [13,14,15,16,20]},\n",
    "    },\n",
    "    \"trials\": 1,\n",
    "})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build the model\n",
    "model = Sequential()\n",
    "model.add(\n",
    "    Dense(\n",
    "        units=550,\n",
    "        activation='elu',\n",
    "        input_shape=(num_train.shape[1],)\n",
    "    )\n",
    ")\n",
    "model.add(Dropout(0.5))\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "model.add(\n",
    "    Dense(\n",
    "        units=300,\n",
    "        activation='elu'))\n",
    "model.add(Dropout(0.3))\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "model_meta = Sequential()\n",
    "model_meta.add(\n",
    "    Dense(\n",
    "        units=600,\n",
    "        activation='elu',\n",
    "        input_shape=(num_train_meta.shape[1],)\n",
    "    )\n",
    ")\n",
    "model_meta.add(Dropout(0.5))\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "model_meta.add(\n",
    "    Dense(\n",
    "        units=128,\n",
    "        activation='elu'))\n",
    "model_meta.add(Dropout(0.3))\n",
    "model_meta.add(BatchNormalization())\n",
    "\n",
    "\n",
    "merged_out = Concatenate()([model.output,model_meta.output,model2.output])\n",
    "merged_out = Dense(\n",
    "    units=32,\n",
    "    activation='elu')(merged_out)\n",
    "merged_out = Dense(8,activation='softmax')(merged_out)\n",
    "\n",
    "combined_model = Model([model.input] + [model_meta.input] + [model2.input], [merged_out])\n",
    "combined_model.compile(\n",
    "    optimizer='adam',\n",
    "    loss='categorical_crossentropy',\n",
    "    metrics=['accuracy']\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "begin training using onecycle policy with max lr of 0.001...\n",
      "Epoch 1/14\n",
      "120/120 [==============================] - 2s 12ms/step - loss: 1.8585 - accuracy: 0.3545 - val_loss: 1.5298 - val_accuracy: 0.4356\n",
      "Epoch 2/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.5040 - accuracy: 0.4459 - val_loss: 1.4578 - val_accuracy: 0.4578\n",
      "Epoch 3/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.3634 - accuracy: 0.4996 - val_loss: 1.4273 - val_accuracy: 0.4911\n",
      "Epoch 4/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.2566 - accuracy: 0.5343 - val_loss: 1.4809 - val_accuracy: 0.4756\n",
      "Epoch 5/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.1648 - accuracy: 0.5698 - val_loss: 1.4371 - val_accuracy: 0.4800\n",
      "Epoch 6/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.0945 - accuracy: 0.5944 - val_loss: 1.3483 - val_accuracy: 0.5289\n",
      "Epoch 7/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 1.0083 - accuracy: 0.6266 - val_loss: 1.3746 - val_accuracy: 0.5178\n",
      "Epoch 8/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.9368 - accuracy: 0.6444 - val_loss: 1.2550 - val_accuracy: 0.5644\n",
      "Epoch 9/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.8635 - accuracy: 0.6763 - val_loss: 1.2358 - val_accuracy: 0.5578\n",
      "Epoch 10/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.8039 - accuracy: 0.6896 - val_loss: 1.3060 - val_accuracy: 0.5533\n",
      "Epoch 11/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.7611 - accuracy: 0.7008 - val_loss: 1.2736 - val_accuracy: 0.5600\n",
      "Epoch 12/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.7276 - accuracy: 0.7106 - val_loss: 1.2528 - val_accuracy: 0.5800\n",
      "Epoch 13/14\n",
      "120/120 [==============================] - 1s 8ms/step - loss: 0.6977 - accuracy: 0.7201 - val_loss: 1.2596 - val_accuracy: 0.5778\n",
      "Epoch 14/14\n",
      "120/120 [==============================] - 1s 9ms/step - loss: 0.6941 - accuracy: 0.7279 - val_loss: 1.2599 - val_accuracy: 0.5778\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1e62da07820>"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learner = ktrain.get_learner(\n",
    "            combined_model,\n",
    "            train_data=train_data,\n",
    "            val_data=valid_data,\n",
    "            batch_size=64\n",
    "        )\n",
    "learner.fit_onecycle(\n",
    "    lr=0.001,\n",
    "    callbacks=[early_stop],\n",
    "    epochs=14,\n",
    "    class_weight=class_weight,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_test_data = df_test_features.iloc[:,9:]\n",
    "num_test_meta_data = df_test_features.iloc[:,3:5]\n",
    "num_test_meta_data['duration'] = df_test_features['duration']\n",
    "\n",
    "num_test = scaler.fit_transform(num_test_data)\n",
    "num_test_meta = scaler.fit_transform(num_test_meta_data)\n",
    "\n",
    "df_test_label = pd.read_csv('test_labels.csv')\n",
    "num_test_y = pd.get_dummies(df_test_label['genre'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_test2 = df_test_features['tags'].apply(lambda x : list_to_string(str_remove_comma(x)))\n",
    "x_test2 = x_test2.apply(lambda x : ' '.join([word for word in x.split() if word not in (stopwords)]))\n",
    "x_test2 = np.array(x_test2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "428 test sequences\n",
      "test sequence lengths:\n",
      "\tmean : 54\n",
      "\t95percentile : 109\n",
      "\t99percentile : 150\n",
      "x_test shape: (428,150)\n"
     ]
    }
   ],
   "source": [
    "x_text_input = preproc2.preprocess_test(x_test2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_dataset = [num_test] + [num_test_meta] + [x_text_input[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction = learner.model.predict(test_dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                       precision    recall  f1-score   support\n",
      "\n",
      " classic pop and rock       0.31      0.47      0.37        64\n",
      "dance and electronica       0.37      0.16      0.22        64\n",
      "                 folk       0.57      0.59      0.58        73\n",
      "       jazz and blues       0.18      0.11      0.14        36\n",
      "                metal       0.62      0.55      0.58        42\n",
      "                  pop       0.67      0.98      0.80        44\n",
      "                 punk       0.51      0.47      0.49        40\n",
      "      soul and reggae       0.41      0.43      0.42        65\n",
      "\n",
      "             accuracy                           0.47       428\n",
      "            macro avg       0.46      0.47      0.45       428\n",
      "         weighted avg       0.45      0.47      0.45       428\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predict_labels = [ label.argmax() for label in prediction]\n",
    "df = pd.DataFrame(predict_labels, columns=['genre_pred'])\n",
    "classes_mapping = {\n",
    "    0: 'classic pop and rock',\n",
    "    1: 'dance and electronica',\n",
    "    2: 'folk',\n",
    "    3: 'jazz and blues',\n",
    "    4: 'metal',\n",
    "    5: 'pop',\n",
    "    6: 'punk',\n",
    "    7: 'soul and reggae',\n",
    "}\n",
    "df['genre_pred'] = df['genre_pred'].map(classes_mapping)\n",
    "df['trackID'] = df_test_features['trackID']\n",
    "\n",
    "correct_label = pd.read_csv('test_labels.csv')\n",
    "# merge df and correct_label by trackID\n",
    "df_merged = pd.merge(df, correct_label, on='trackID')\n",
    "\n",
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(df_merged['genre'], df_merged['genre_pred']))"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "984944ce4eefbed249e2a3501e72ee0777f56e66d10a0a53283275062444400f"
  },
  "kernelspec": {
   "display_name": "Python 3.9.6 64-bit",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.6"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}