training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Language Classification\n",
    "\n",
    "This notebook contains the code for a CNN that classifies spoken language from Mel spectrograms. There is additional code for other types of models (only some working), but the best results are obtained using the model saved as \"cnn\" in the training segment.\n",
    "\n",
    "Used as reference:\n",
    "\n",
    "https://github.com/pietz/language-recognition/blob/master/Language%20Classifier.ipynb\n",
    "\n",
    "https://github.com/sergeyvilov/MCV-spoken-language-recognition/blob/master/audio_models.py\n",
    "\n",
    "https://towardsdatascience.com/spoken-language-recognition-on-mozilla-common-voice-part-ii-models-b32780ea1ee4\n",
    "\n",
    "https://towardsdatascience.com/spoken-language-recognition-on-mozilla-common-voice-part-i-3f5400bbbcd8\n",
    "\n",
    "https://github.com/zkmkarlsruhe/language-identification/tree/main"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Loading\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.preprocessing import image\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Languages to Detect\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "English\n",
      "French\n"
     ]
    }
   ],
   "source": [
    "np.random.seed(1)\n",
    "from languages import LANGUAGES\n",
    "\n",
    "langs = [\"de\", \"en\", \"es\", \"fr\", \"it\"]\n",
    "# langs = [\"en\", \"fr\"]\n",
    "lang_nums = {lang: index for index, lang in enumerate(langs)}\n",
    "\n",
    "num_to_lang = {v: k for k, v in lang_nums.items()}\n",
    "for lang in langs:\n",
    "    print(LANGUAGES[lang])\n",
    "\n",
    "# IMAGE_DIMENSIONS = (20, 250, 1)\n",
    "IMAGE_DIMENSIONS = (13, 1000, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load Spectrograms\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_images_from_path(path, label, num_images=1000, shuffle=True):\n",
    "    images = []\n",
    "    labels = []\n",
    "\n",
    "    count = 0\n",
    "    files = os.listdir(path)\n",
    "    if shuffle:\n",
    "        files = np.random.choice(files, num_images, replace=False)\n",
    "    else:\n",
    "        files = files[:num_images]\n",
    "    for file in tqdm(files):\n",
    "        # print(f\"{count} - {file}\")\n",
    "        if count >= num_images:\n",
    "            break\n",
    "        try:\n",
    "            img = image.img_to_array(\n",
    "                image.load_img(\n",
    "                    os.path.join(path, file),\n",
    "                    target_size=(IMAGE_DIMENSIONS[0], IMAGE_DIMENSIONS[1], 3),\n",
    "                ).convert(\"L\")\n",
    "            )\n",
    "            img = img.reshape(IMAGE_DIMENSIONS)\n",
    "            # from 255 to 0-1\n",
    "            img = img / 255\n",
    "            images.append(img)\n",
    "            labels.append((label))\n",
    "            count += 1\n",
    "        except Exception as e:\n",
    "            print(f\"Error loading image: {file}\")\n",
    "            print(e)\n",
    "\n",
    "    return images, labels\n",
    "\n",
    "\n",
    "def show_images(images, label):\n",
    "    fig, axes = plt.subplots(\n",
    "        1, 8, figsize=(20, 20), subplot_kw={\"xticks\": [], \"yticks\": []}\n",
    "    )\n",
    "\n",
    "    for i, ax in enumerate(axes.flat):\n",
    "        ax.imshow(images[i])\n",
    "        # ax.imshow(images[i] / 255)\n",
    "        ax.set_title(LANGUAGES[num_to_lang[labels[i]]])  # Set the title for the subplot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "English\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 10000/10000 [00:03<00:00, 2711.49it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(20, 250, 1)\n",
      "French\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 10000/10000 [00:03<00:00, 3038.76it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(20, 250, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x2000 with 8 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x2000 with 8 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "mm = []\n",
    "y = []\n",
    "NUM_IMGS = 92_740\n",
    "# NUM_IMGS = 20\n",
    "# NUM_IMGS = 50_000\n",
    "# NUM_IMGS = 10_000\n",
    "\n",
    "for lang in langs:\n",
    "    print(LANGUAGES[lang])\n",
    "    path = f\"data/spectrogram/{lang}_train_20_mels/\"\n",
    "    images, labels = load_images_from_path(\n",
    "        path, lang_nums[lang], num_images=NUM_IMGS, shuffle=False\n",
    "    )\n",
    "    mm += images\n",
    "    y += labels\n",
    "    print(images[0].shape)\n",
    "    show_images(images, labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Splitting\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf.random.set_seed(1)\n",
    "\n",
    "x_train, x_test, y_train, y_test = train_test_split(\n",
    "    mm, y, stratify=y, test_size=0.2, random_state=0, shuffle=True\n",
    ")\n",
    "\n",
    "x_train_arr = np.array(x_train)\n",
    "x_test_arr = np.array(x_test)\n",
    "\n",
    "y_train_encoded = to_categorical(y_train)\n",
    "y_test_encoded = to_categorical(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 20, 250, 1)\n",
      "(10000, 2)\n"
     ]
    }
   ],
   "source": [
    "# truncate dataset for faster training:\n",
    "x_train_arr = x_train_arr[:NUM_IMGS]\n",
    "x_test_arr = x_test_arr[:NUM_IMGS]\n",
    "y_train_encoded = y_train_encoded[:NUM_IMGS]\n",
    "y_test_encoded = y_test_encoded[:NUM_IMGS]\n",
    "print(x_train_arr.shape)\n",
    "print(y_train_encoded.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 20, 250, 1)\n",
      "(10000, 2)\n",
      "(4000, 20, 250, 1)\n",
      "(4000, 2)\n",
      "(array([0, 1]), array([8000, 8000]))\n",
      "(array([0., 1.], dtype=float32), array([10000, 10000]))\n"
     ]
    }
   ],
   "source": [
    "print(x_train_arr.shape)\n",
    "print(y_train_encoded.shape)\n",
    "print(x_test_arr.shape)\n",
    "print(y_test_encoded.shape)\n",
    "# print number per class\n",
    "print(np.unique(y_train, return_counts=True))\n",
    "print(np.unique(y_train_encoded, return_counts=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(20, 250, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x2000 with 8 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_images(x_train_arr, y_train)\n",
    "print(IMAGE_DIMENSIONS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Make sure data remains consistent between runs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "OK\n"
     ]
    }
   ],
   "source": [
    "import hashlib\n",
    "\n",
    "hash_x_train = hashlib.sha256(str(x_train_arr).encode(\"utf-8\")).hexdigest()\n",
    "hash_x_test = hashlib.sha256(str(x_test_arr).encode(\"utf-8\")).hexdigest()\n",
    "hash_y_train = hashlib.sha256(str(y_train_encoded).encode(\"utf-8\")).hexdigest()\n",
    "hash_y_test = hashlib.sha256(str(y_test_encoded).encode(\"utf-8\")).hexdigest()\n",
    "\n",
    "# set this to false if you want it to remain consistent run-to-run\n",
    "# for experimentation and modification without being stopped, set to true\n",
    "update_hash = True\n",
    "\n",
    "if update_hash:\n",
    "    with open(\"./hashes/hash_x_train.txt\", \"w\") as f:\n",
    "        f.write(hash_x_train)\n",
    "    with open(\"./hashes/hash_x_test.txt\", \"w\") as f:\n",
    "        f.write(hash_x_test)\n",
    "    with open(\"./hashes/hash_y_train.txt\", \"w\") as f:\n",
    "        f.write(hash_y_train)\n",
    "    with open(\"./hashes/hash_y_test.txt\", \"w\") as f:\n",
    "        f.write(hash_y_test)\n",
    "\n",
    "with open(\"./hashes/hash_x_train.txt\", \"r\") as f:\n",
    "    expected_hash_x_train = f.read()\n",
    "with open(\"./hashes/hash_x_test.txt\", \"r\") as f:\n",
    "    expected_hash_x_test = f.read()\n",
    "with open(\"./hashes/hash_y_train.txt\", \"r\") as f:\n",
    "    expected_hash_y_train = f.read()\n",
    "with open(\"./hashes/hash_y_test.txt\", \"r\") as f:\n",
    "    expected_hash_y_test = f.read()\n",
    "\n",
    "# Check if the computed hashes match the expected values\n",
    "assert (\n",
    "    hash_x_train == expected_hash_x_train\n",
    "), f\"Hash of x does not match the expected value: {expected_hash_x_train}\"\n",
    "assert (\n",
    "    hash_x_test == expected_hash_x_test\n",
    "), f\"Hash of x does not match the expected value: {expected_hash_x_test}\"\n",
    "assert (\n",
    "    hash_y_train == expected_hash_y_train\n",
    "), f\"Hash of y does not match the expected value: {expected_hash_y_train}\"\n",
    "assert (\n",
    "    hash_y_test == expected_hash_y_test\n",
    "), f\"Hash of y does not match the expected value: {expected_hash_y_test}\"\n",
    "print(\"OK\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Models\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Model, Sequential\n",
    "from keras.layers import (\n",
    "    Conv2D,\n",
    "    MaxPooling2D,\n",
    "    Flatten,\n",
    "    Dense,\n",
    "    Dropout,\n",
    "    Input,\n",
    "    TimeDistributed,\n",
    "    GRU,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " conv2d (Conv2D)             (None, 18, 248, 32)       320       \n",
      "                                                                 \n",
      " conv2d_1 (Conv2D)           (None, 16, 123, 64)       18496     \n",
      "                                                                 \n",
      " conv2d_2 (Conv2D)           (None, 14, 121, 64)       36928     \n",
      "                                                                 \n",
      " conv2d_3 (Conv2D)           (None, 14, 60, 128)       24704     \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 107520)            0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 1024)              110101504 \n",
      "                                                                 \n",
      " dropout (Dropout)           (None, 1024)              0         \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 2)                 2050      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 110184002 (420.32 MB)\n",
      "Trainable params: 110184002 (420.32 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# sequentialModel = Sequential()\n",
    "# sequentialModel.add(Conv2D(32, (3, 3), activation=\"relu\", input_shape=(IMAGE_DIMENSIONS)))\n",
    "# sequentialModel.add(MaxPooling2D(2, 2))\n",
    "# sequentialModel.add(Conv2D(128, (3, 3), activation=\"relu\"))\n",
    "# sequentialModel.add(MaxPooling2D(2, 2))\n",
    "# sequentialModel.add(Conv2D(128, (3, 3), activation=\"relu\"))\n",
    "# sequentialModel.add(MaxPooling2D(2, 2))\n",
    "# sequentialModel.add(Conv2D(128, (3, 3), activation=\"relu\"))\n",
    "# sequentialModel.add(MaxPooling2D(2, 2))\n",
    "# sequentialModel.add(Flatten())\n",
    "# sequentialModel.add(Dense(1024, activation=\"relu\"))\n",
    "# sequentialModel.add(Dense(len(langs), activation=\"softmax\"))\n",
    "sequentialModel = Sequential()\n",
    "sequentialModel.add(\n",
    "    Conv2D(32, (3, 3), activation=\"relu\", strides=(1, 1), input_shape=IMAGE_DIMENSIONS)\n",
    ")\n",
    "sequentialModel.add(Conv2D(64, (3, 3), activation=\"relu\", strides=(1, 2)))\n",
    "sequentialModel.add(Conv2D(64, (3, 3), activation=\"relu\", strides=(1, 1)))\n",
    "sequentialModel.add(Conv2D(128, (1, 3), activation=\"relu\", strides=(1, 2)))\n",
    "sequentialModel.add(Flatten())\n",
    "sequentialModel.add(Dense(1024, activation=\"relu\"))\n",
    "sequentialModel.add(Dropout(0.5))\n",
    "sequentialModel.add(Dense(len(langs), activation=\"softmax\"))\n",
    "\n",
    "\n",
    "seqOpt = tf.keras.optimizers.legacy.Nadam(learning_rate=1e-3)\n",
    "sequentialModel.compile(\n",
    "    optimizer=seqOpt, loss=\"categorical_crossentropy\", metrics=[\"accuracy\"]\n",
    ")\n",
    "sequentialModel.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_1 (InputLayer)        [(None, 20, 250, 1)]      0         \n",
      "                                                                 \n",
      " conv2d_4 (Conv2D)           (None, 20, 250, 16)       160       \n",
      "                                                                 \n",
      " max_pooling2d (MaxPooling2  (None, 10, 125, 16)       0         \n",
      " D)                                                              \n",
      "                                                                 \n",
      " conv2d_5 (Conv2D)           (None, 10, 125, 32)       4640      \n",
      "                                                                 \n",
      " max_pooling2d_1 (MaxPoolin  (None, 5, 62, 32)         0         \n",
      " g2D)                                                            \n",
      "                                                                 \n",
      " conv2d_6 (Conv2D)           (None, 5, 62, 64)         18496     \n",
      "                                                                 \n",
      " conv2d_7 (Conv2D)           (None, 5, 62, 128)        73856     \n",
      "                                                                 \n",
      " conv2d_8 (Conv2D)           (None, 5, 62, 256)        295168    \n",
      "                                                                 \n",
      " flatten_1 (Flatten)         (None, 79360)             0         \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 512)               40632832  \n",
      "                                                                 \n",
      " dropout_1 (Dropout)         (None, 512)               0         \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 2)                 1026      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 41026178 (156.50 MB)\n",
      "Trainable params: 41026178 (156.50 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# architecture from\n",
    "# https://github.com/pietz/language-recognition/blob/master/Language%20Classifier.ipynb\n",
    "in_dim = IMAGE_DIMENSIONS  # 13x1000 grayscale images (1 color channel)\n",
    "\n",
    "i = Input(shape=in_dim)\n",
    "m = Conv2D(16, (3, 3), activation=\"elu\", padding=\"same\")(i)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Conv2D(32, (3, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Conv2D(64, (3, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "# m = MaxPooling2D()(m)\n",
    "m = Conv2D(128, (3, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "# m = MaxPooling2D()(m)\n",
    "m = Conv2D(256, (3, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "# m = MaxPooling2D()(m)\n",
    "m = Flatten()(m)\n",
    "m = Dense(512, activation=\"elu\")(m)\n",
    "m = Dropout(0.5)(m)\n",
    "o = Dense(len(langs), activation=\"softmax\")(m)\n",
    "\n",
    "githubModel = Model(inputs=i, outputs=o)\n",
    "\n",
    "gitOpt = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3)\n",
    "githubModel.compile(\n",
    "    loss=\"categorical_crossentropy\", optimizer=gitOpt, metrics=[\"accuracy\"]\n",
    ")\n",
    "githubModel.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# THIS MODEL USED\n",
    "# architecture heavily borrowed from\n",
    "# https://github.com/pietz/language-recognition/blob/master/Language%20Classifier.ipynb\n",
    "IMAGE_DIMENSIONS = (13, 250, 1)\n",
    "i = Input(shape=IMAGE_DIMENSIONS)\n",
    "m = Conv2D(16, (3, 3), activation=\"elu\", padding=\"same\")(i)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Conv2D(32, (3, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Conv2D(64, (2, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Conv2D(128, (1, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = Flatten()(m)\n",
    "m = Dense(512, activation=\"elu\")(m)\n",
    "m = Flatten()(m)\n",
    "m = Dropout(0.5)(m)\n",
    "o = Dense(len(langs), activation=\"softmax\")(m)\n",
    "\n",
    "cnn = Model(inputs=i, outputs=o)\n",
    "\n",
    "cnopt = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3)\n",
    "cnn.compile(loss=\"categorical_crossentropy\", optimizer=cnopt, metrics=[\"accuracy\"])\n",
    "cnn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_4\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_10 (InputLayer)       [(None, 20, 250, 1)]      0         \n",
      "                                                                 \n",
      " conv2d_49 (Conv2D)          (None, 20, 250, 16)       64        \n",
      "                                                                 \n",
      " max_pooling2d_39 (MaxPooli  (None, 20, 125, 16)       0         \n",
      " ng2D)                                                           \n",
      "                                                                 \n",
      " conv2d_50 (Conv2D)          (None, 20, 125, 32)       1568      \n",
      "                                                                 \n",
      " max_pooling2d_40 (MaxPooli  (None, 20, 62, 32)        0         \n",
      " ng2D)                                                           \n",
      "                                                                 \n",
      " conv2d_51 (Conv2D)          (None, 20, 62, 64)        6208      \n",
      "                                                                 \n",
      " max_pooling2d_41 (MaxPooli  (None, 10, 31, 64)        0         \n",
      " ng2D)                                                           \n",
      "                                                                 \n",
      " conv2d_52 (Conv2D)          (None, 10, 31, 128)       24704     \n",
      "                                                                 \n",
      " max_pooling2d_42 (MaxPooli  (None, 5, 15, 128)        0         \n",
      " ng2D)                                                           \n",
      "                                                                 \n",
      " conv2d_53 (Conv2D)          (None, 5, 15, 256)        98560     \n",
      "                                                                 \n",
      " max_pooling2d_43 (MaxPooli  (None, 2, 7, 256)         0         \n",
      " ng2D)                                                           \n",
      "                                                                 \n",
      " flatten_6 (Flatten)         (None, 3584)              0         \n",
      "                                                                 \n",
      " dense_10 (Dense)            (None, 512)               1835520   \n",
      "                                                                 \n",
      " dropout_5 (Dropout)         (None, 512)               0         \n",
      "                                                                 \n",
      " dense_11 (Dense)            (None, 2)                 1026      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1967650 (7.51 MB)\n",
      "Trainable params: 1967650 (7.51 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# v2 test\n",
    "# modifications of my own design\n",
    "IMAGE_DIMENSIONS = (20, 250, 1)\n",
    "i = Input(shape=IMAGE_DIMENSIONS)\n",
    "m = Conv2D(16, (1, 3), activation=\"elu\", padding=\"same\")(i)\n",
    "m = MaxPooling2D((1, 2))(m)\n",
    "m = Conv2D(32, (1, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = MaxPooling2D((1, 2))(m)\n",
    "m = Conv2D(64, (1, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Conv2D(128, (1, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Conv2D(256, (1, 3), activation=\"elu\", padding=\"same\")(m)\n",
    "m = MaxPooling2D()(m)\n",
    "m = Flatten()(m)\n",
    "m = Dense(512, activation=\"elu\")(m)\n",
    "m = Dropout(0.5)(m)\n",
    "o = Dense(len(langs), activation=\"softmax\")(m)\n",
    "\n",
    "cnn_v2_test = Model(inputs=i, outputs=o)\n",
    "\n",
    "cnopt_v2_test = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3)\n",
    "cnn_v2_test.compile(\n",
    "    loss=\"categorical_crossentropy\", optimizer=cnopt_v2_test, metrics=[\"accuracy\"]\n",
    ")\n",
    "cnn_v2_test.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " conv2d_13 (Conv2D)          (None, 18, 248, 32)       320       \n",
      "                                                                 \n",
      " max_pooling2d_5 (MaxPoolin  (None, 9, 124, 32)        0         \n",
      " g2D)                                                            \n",
      "                                                                 \n",
      " dropout_3 (Dropout)         (None, 9, 124, 32)        0         \n",
      "                                                                 \n",
      " time_distributed (TimeDist  (None, 9, 3968)           0         \n",
      " ributed)                                                        \n",
      "                                                                 \n",
      " gru (GRU)                   (None, 128)               1573632   \n",
      "                                                                 \n",
      " dropout_4 (Dropout)         (None, 128)               0         \n",
      "                                                                 \n",
      " dense_6 (Dense)             (None, 2)                 258       \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1574210 (6.01 MB)\n",
      "Trainable params: 1574210 (6.01 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Define the spectrogram shape at the input\n",
    "# input_shape = IMAGE_DIMENSIONS  # 13 - height, 1000 - width, 1 - grayscale\n",
    "\n",
    "# Create a Sequential model\n",
    "model = Sequential()\n",
    "\n",
    "# Add 2D convolution layer\n",
    "model.add(\n",
    "    Conv2D(32, kernel_size=(3, 3), activation=\"relu\", input_shape=IMAGE_DIMENSIONS)\n",
    ")\n",
    "\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "# dropout layer for regularization\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(TimeDistributed(Flatten()))\n",
    "\n",
    "# add GRU layer\n",
    "model.add(GRU(128, return_sequences=False))\n",
    "\n",
    "model.add(Dropout(0.5))\n",
    "\n",
    "model.add(Dense(len(langs), activation=\"softmax\"))\n",
    "opt = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3) # last tested on 1e-4, got like 1 for loss\n",
    "model.compile(loss=\"categorical_crossentropy\", optimizer=opt, metrics=[\"accuracy\"])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# taken directly from\n",
    "# https://github.com/sergeyvilov/MCV-spoken-language-recognition/blob/master/audio_models.py\n",
    "# not used\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras import layers, models\n",
    "from torch import dropout\n",
    "\n",
    "\n",
    "class CRNN(models.Model):\n",
    "    def __init__(\n",
    "        self,\n",
    "        n_languages=len(langs),\n",
    "        input_height=20,\n",
    "        lstm_layers=2,\n",
    "        dropout=0.25,\n",
    "        **kwargs\n",
    "    ):\n",
    "        super().__init__(**kwargs)\n",
    "\n",
    "        self.image_height = input_height\n",
    "        self.lstm_layers = lstm_layers\n",
    "        self.lstm_hout = 128\n",
    "        self.dropout = dropout\n",
    "        self.n_languages = n_languages\n",
    "\n",
    "        # Convolutional layers, note that 'padding=same' ensures the output has the same dimension as the input.\n",
    "        self.conv1 = layers.Conv2D(10, kernel_size=(1, 5), padding=\"same\")\n",
    "        self.bn1 = layers.BatchNormalization()\n",
    "\n",
    "        self.conv2 = layers.Conv2D(1, kernel_size=(1, 5), padding=\"same\")\n",
    "        self.bn2 = layers.BatchNormalization()\n",
    "\n",
    "        # LSTM Layer, note that return_sequences is set to True to return the full sequence.\n",
    "        self.rnn = layers.Bidirectional(\n",
    "            layers.LSTM(self.lstm_hout, return_sequences=True, dropout=dropout),\n",
    "            merge_mode=\"concat\",\n",
    "        )\n",
    "\n",
    "        # Fully connected layers\n",
    "        self.fc1 = layers.Dense(32)\n",
    "        self.fc3 = layers.Dense(32)\n",
    "        self.fc4 = layers.Dense(self.n_languages)\n",
    "\n",
    "    def build(self, input_shape):\n",
    "        # Define the RNN and FC layers now that we have the input_shape\n",
    "        self.input_shape = input_shape[1:]  # Exclude the batch dimension\n",
    "        # Compute the number of features after Conv2D layers\n",
    "        lstm_input_dim = self.input_shape[0] * self.input_shape[2]\n",
    "\n",
    "        # LSTM Layer\n",
    "        self.rnn = layers.Bidirectional(\n",
    "            layers.LSTM(\n",
    "                self.lstm_hout,\n",
    "                return_sequences=True,\n",
    "                dropout=dropout,\n",
    "            ),\n",
    "            merge_mode=\"concat\",\n",
    "        )\n",
    "\n",
    "        # Fully connected layers\n",
    "        self.fc1 = layers.Dense(32)\n",
    "        self.fc3 = layers.Dense(32)\n",
    "        self.fc4 = layers.Dense(self.n_languages)\n",
    "\n",
    "        super().build(input_shape)\n",
    "\n",
    "    def call(self, x):\n",
    "        # Add a dimension equivalent to channels at the start (N_batch, Channels, Height, Width)\n",
    "        x = tf.expand_dims(x, axis=-1)\n",
    "\n",
    "        # Apply convolutional layers\n",
    "        out = self.conv1(x)\n",
    "        out = self.bn1(out)\n",
    "        out = tf.nn.relu(out)\n",
    "\n",
    "        out = self.conv2(out)\n",
    "        out = self.bn2(out)\n",
    "        out = tf.nn.relu(out)\n",
    "\n",
    "        # Reshaping to (batch_size, sequence, features) before applying RNN\n",
    "        out = tf.reshape(out, (tf.shape(out)[0], tf.shape(out)[2], -1))  # type: ignore\n",
    "\n",
    "        # Apply RNN layer\n",
    "        out = self.rnn(out)\n",
    "\n",
    "        # Flattening is not always necessary before a Dense layer, because the Dense layer can handle multi-dimensional inputs\n",
    "\n",
    "        # Apply fully connected layers\n",
    "        out = self.fc1(out)\n",
    "        out = tf.nn.relu(out)\n",
    "        out = tf.nn.dropout(out, rate=self.dropout)\n",
    "\n",
    "        out = self.fc3(out)\n",
    "        out = tf.nn.relu(out)\n",
    "        out = tf.nn.dropout(out, rate=self.dropout)\n",
    "\n",
    "        out = self.fc4(out)\n",
    "\n",
    "        return out\n",
    "\n",
    "\n",
    "# Usage\n",
    "crnn_model = CRNN(n_languages=len(langs), input_height=13, lstm_layers=2, dropout=0.25)\n",
    "crnn_opt = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3)\n",
    "crnn_model.compile(\n",
    "    optimizer=crnn_opt, loss=\"categorical_crossentropy\", metrics=[\"accuracy\"]\n",
    ")\n",
    "print(IMAGE_DIMENSIONS)\n",
    "crnn_model.build((32, 1, IMAGE_DIMENSIONS[2], IMAGE_DIMENSIONS[1]))\n",
    "crnn_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(x_train_arr.shape)\n",
    "crnn_model.fit(\n",
    "    x_train_arr,\n",
    "    y_train_encoded,\n",
    "    validation_data=(x_test_arr, y_test_encoded),\n",
    "    batch_size=32,\n",
    "    epochs=10,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " gru_1 (GRU)                 (None, 13, 256)           390144    \n",
      "                                                                 \n",
      " gru_2 (GRU)                 (None, 13, 256)           394752    \n",
      "                                                                 \n",
      " gru_3 (GRU)                 (None, 256)               394752    \n",
      "                                                                 \n",
      " dense_7 (Dense)             (None, 256)               65792     \n",
      "                                                                 \n",
      " dense_8 (Dense)             (None, 256)               65792     \n",
      "                                                                 \n",
      " dense_9 (Dense)             (None, 5)                 1285      \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1312517 (5.01 MB)\n",
      "Trainable params: 1312517 (5.01 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "rnn = Sequential()\n",
    "\n",
    "rnn.add(\n",
    "    GRU(\n",
    "        256,\n",
    "        return_sequences=True,\n",
    "        input_shape=(IMAGE_DIMENSIONS[0], IMAGE_DIMENSIONS[1]),\n",
    "    )\n",
    ")\n",
    "rnn.add(GRU(256, return_sequences=True))\n",
    "rnn.add(GRU(256))\n",
    "\n",
    "# Add a fully connected layer\n",
    "rnn.add(Dense(256, activation=\"relu\"))\n",
    "rnn.add(Dense(256, activation=\"relu\"))\n",
    "\n",
    "# Output Layer - softmax activation function for multi-class classification\n",
    "rnn.add(Dense(len(langs), activation=\"softmax\"))\n",
    "\n",
    "# Compile the model\n",
    "rnn.compile(\n",
    "    loss=tf.keras.losses.categorical_crossentropy,\n",
    "    optimizer=tf.keras.optimizers.legacy.Adam(),\n",
    "    metrics=[\"accuracy\"],\n",
    ")\n",
    "\n",
    "rnn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_2\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                Output Shape                 Param #   Connected to                  \n",
      "==================================================================================================\n",
      " input (InputLayer)          [(None, 13, 250, 1)]         0         []                            \n",
      "                                                                                                  \n",
      " batch_normalization (Batch  (None, 13, 250, 1)           4         ['input[0][0]']               \n",
      " Normalization)                                                                                   \n",
      "                                                                                                  \n",
      " conv2d_14 (Conv2D)          (None, 13, 250, 10)          60        ['batch_normalization[0][0]'] \n",
      "                                                                                                  \n",
      " batch_normalization_1 (Bat  (None, 13, 250, 10)          40        ['conv2d_14[0][0]']           \n",
      " chNormalization)                                                                                 \n",
      "                                                                                                  \n",
      " conv2d_15 (Conv2D)          (None, 13, 250, 1)           51        ['batch_normalization_1[0][0]'\n",
      "                                                                    ]                             \n",
      "                                                                                                  \n",
      " batch_normalization_2 (Bat  (None, 13, 250, 1)           4         ['conv2d_15[0][0]']           \n",
      " chNormalization)                                                                                 \n",
      "                                                                                                  \n",
      " squeeze_last_dim (Lambda)   (None, 13, 250)              0         ['batch_normalization_2[0][0]'\n",
      "                                                                    ]                             \n",
      "                                                                                                  \n",
      " bidirectional (Bidirection  (None, 13, 128)              161280    ['squeeze_last_dim[0][0]']    \n",
      " al)                                                                                              \n",
      "                                                                                                  \n",
      " bidirectional_1 (Bidirecti  (None, 13, 128)              98816     ['bidirectional[0][0]']       \n",
      " onal)                                                                                            \n",
      "                                                                                                  \n",
      " lambda (Lambda)             (None, 128)                  0         ['bidirectional_1[0][0]']     \n",
      "                                                                                                  \n",
      " dense_10 (Dense)            (None, 128)                  16512     ['lambda[0][0]']              \n",
      "                                                                                                  \n",
      " dot (Dot)                   (None, 13)                   0         ['dense_10[0][0]',            \n",
      "                                                                     'bidirectional_1[0][0]']     \n",
      "                                                                                                  \n",
      " attSoftmax (Softmax)        (None, 13)                   0         ['dot[0][0]']                 \n",
      "                                                                                                  \n",
      " dot_1 (Dot)                 (None, 128)                  0         ['attSoftmax[0][0]',          \n",
      "                                                                     'bidirectional_1[0][0]']     \n",
      "                                                                                                  \n",
      " dense_11 (Dense)            (None, 64)                   8256      ['dot_1[0][0]']               \n",
      "                                                                                                  \n",
      " dense_12 (Dense)            (None, 32)                   2080      ['dense_11[0][0]']            \n",
      "                                                                                                  \n",
      " output (Dense)              (None, 5)                    165       ['dense_12[0][0]']            \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 287268 (1.10 MB)\n",
      "Trainable params: 287244 (1.10 MB)\n",
      "Non-trainable params: 24 (96.00 Byte)\n",
      "__________________________________________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "\"\"\"\n",
    ":author:\n",
    "Paul Bethge (bethge@zkm.de)\n",
    "2021\n",
    "\n",
    ":License:\n",
    "This package is published under Simplified BSD License.\n",
    "\"\"\"\n",
    "\n",
    "# https://github.com/douglas125/SpeechCmdRecognition/blob/master/SpeechModels.py\n",
    "\n",
    "from tensorflow.keras.layers import Dropout, BatchNormalization\n",
    "from tensorflow.keras.layers import Input, Dense, LSTM, Conv2D, Bidirectional\n",
    "from tensorflow.keras.layers import Lambda, Dot, Softmax\n",
    "from tensorflow.keras.models import Model\n",
    "\n",
    "import tensorflow.keras.backend as K\n",
    "\n",
    "# from src.utils.training_utils import get_feature_layer\n",
    "\n",
    "\n",
    "def create_model():\n",
    "    \"\"\"\n",
    "    This code is heavily borrowed from <https://github.com/douglas125/SpeechCmdRecognition>.\n",
    "    \"\"\"\n",
    "    # inputs = Input((IMAGE_DIMENSIONS[1], IMAGE_DIMENSIONS[0], 1), name=\"input\")\n",
    "    # height, length, channels\n",
    "    i = Input(shape=IMAGE_DIMENSIONS, name=\"input\")\n",
    "    x = BatchNormalization()(i)\n",
    "\n",
    "    x = Conv2D(10, (5, 1), activation=\"relu\", padding=\"same\")(x)\n",
    "    x = BatchNormalization()(x)\n",
    "    x = Conv2D(1, (5, 1), activation=\"relu\", padding=\"same\")(x)\n",
    "    x = BatchNormalization()(x)\n",
    "\n",
    "    x = Lambda(lambda q: K.squeeze(q, -1), name=\"squeeze_last_dim\")(x)\n",
    "\n",
    "    x = Bidirectional(LSTM(64, return_sequences=True))(x)  # [b_s, seq_len, vec_dim]\n",
    "    x = Bidirectional(LSTM(64, return_sequences=True))(x)  # [b_s, seq_len, vec_dim]\n",
    "\n",
    "    xFirst = Lambda(lambda q: q[:, -1])(x)  # [b_s, vec_dim]\n",
    "    query = Dense(128)(xFirst)\n",
    "\n",
    "    # dot product attention\n",
    "    attScores = Dot(axes=[1, 2])([query, x])\n",
    "    attScores = Softmax(name=\"attSoftmax\")(attScores)  # [b_s, seq_len]\n",
    "\n",
    "    # rescale sequence\n",
    "    attVector = Dot(axes=[1, 1])([attScores, x])  # [b_s, vec_dim]\n",
    "\n",
    "    x = Dense(64, activation=\"relu\")(attVector)\n",
    "    # x = Dropout(0.25)(x)\n",
    "    x = Dense(32)(x)\n",
    "    # x = Dropout(0.25)(x)\n",
    "\n",
    "    output = Dense(len(langs), activation=\"softmax\", name=\"output\")(x)\n",
    "\n",
    "    model = Model(inputs=[i], outputs=[output])\n",
    "\n",
    "    return model\n",
    "\n",
    "\n",
    "attrnn = create_model()\n",
    "attrnn_opt = tf.keras.optimizers.legacy.Adam(learning_rate=1e-3)\n",
    "attrnn.compile(\n",
    "    optimizer=attrnn_opt,\n",
    "    loss=\"categorical_crossentropy\",\n",
    "    metrics=[\"accuracy\"],\n",
    ")\n",
    "print(attrnn.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#https://github.com/sergeyvilov/MCV-spoken-language-recognition/blob/master/audio_models.py\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "\n",
    "\n",
    "class CRNN(nn.Module):\n",
    "    def __init__(\n",
    "        self,\n",
    "        device,\n",
    "        n_languages=5,\n",
    "        input_height=20,\n",
    "        lstm_layers=2,\n",
    "        dropout=0.25,\n",
    "        **kwargs\n",
    "    ):\n",
    "        \"\"\"\n",
    "        Convolutional RNN with the same number of parameters as above attention RNN\n",
    "        \"\"\"\n",
    "\n",
    "        super().__init__()\n",
    "\n",
    "        self.device = device\n",
    "\n",
    "        self.image_height = input_height\n",
    "\n",
    "        self.lstm_layers = lstm_layers\n",
    "        self.lstm_hout = 128\n",
    "\n",
    "        # input size: N_batch x 1 x image_height x L\n",
    "\n",
    "        self.conv1 = nn.Conv2d(\n",
    "            1, 10, kernel_size=(1, 5), padding=\"same\"\n",
    "        )  # conv only in time dimension\n",
    "        self.bn1 = nn.BatchNorm2d(10)\n",
    "\n",
    "        self.conv2 = nn.Conv2d(\n",
    "            10, 1, kernel_size=(1, 5), padding=\"same\"\n",
    "        )  # conv only in time dimension\n",
    "        self.bn2 = nn.BatchNorm2d(1)\n",
    "\n",
    "        # convert dimensions to N_batch x L x image_height here!\n",
    "\n",
    "        self.rnn = nn.LSTM(\n",
    "            self.image_height,\n",
    "            self.lstm_hout,\n",
    "            self.lstm_layers,\n",
    "            bidirectional=True,\n",
    "            batch_first=True,\n",
    "        )  # Hin, Hout\n",
    "\n",
    "        self.relu = nn.ReLU()\n",
    "\n",
    "        self.flatten = nn.Flatten()\n",
    "\n",
    "        self.dropout = nn.Dropout(dropout)\n",
    "\n",
    "        self.fc1 = nn.LazyLinear(32)\n",
    "        self.fc3 = nn.Linear(32, 32)\n",
    "        self.fc4 = nn.Linear(32, n_languages)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = torch.unsqueeze(x, 1)\n",
    "\n",
    "        n_batch = x.shape[0]\n",
    "\n",
    "        h0 = torch.zeros(2 * self.lstm_layers, n_batch, self.lstm_hout).to(self.device)\n",
    "        c0 = torch.zeros(2 * self.lstm_layers, n_batch, self.lstm_hout).to(self.device)\n",
    "\n",
    "        out = self.conv1(x)\n",
    "        out = self.bn1(out)\n",
    "\n",
    "        out = self.conv2(out)\n",
    "        out = self.bn2(out)\n",
    "\n",
    "        out = torch.permute(\n",
    "            out, (0, 3, 2, 1)\n",
    "        ).squeeze()  # remove last singleton dimension\n",
    "\n",
    "        out, (hn, cn) = self.rnn(out, (h0, c0))\n",
    "\n",
    "        out = self.flatten(out)\n",
    "\n",
    "        out = self.fc1(out)\n",
    "        out = self.relu(out)\n",
    "\n",
    "        out = self.dropout(out)\n",
    "\n",
    "        out = self.fc3(out)\n",
    "        out = self.relu(out)\n",
    "\n",
    "        out = self.dropout(out)\n",
    "\n",
    "        out = self.fc4(out)\n",
    "\n",
    "        return out\n",
    "\n",
    "# set device, use mps if available\n",
    "if torch.backends.mps.is_available():\n",
    "    device = torch.device(\"mps\")\n",
    "elif torch.cuda.is_available():\n",
    "    device = torch.device(\"cuda\")\n",
    "else:\n",
    "    device = torch.device(\"cpu\")\n",
    "\n",
    "model = CRNN(image_height=IMAGE_DIMENSIONS[0], device=device).to(device)\n",
    "\n",
    "# input = torch.randn(3, 20, 13).to(device) #N_batch x 1 x image_height x L\n",
    "\n",
    "model().shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.callbacks import EarlyStopping\n",
    "\n",
    "earlyStopping = EarlyStopping(\n",
    "    monitor=\"val_loss\", patience=4, verbose=0, mode=\"min\", restore_best_weights=True\n",
    ")\n",
    "# specify models to train\n",
    "seq = {\n",
    "    \"model\": sequentialModel,\n",
    "    \"name\": \"sequential\",\n",
    "    \"epochs\": 10,\n",
    "    \"batch_size\": 8,\n",
    "}\n",
    "git = {\n",
    "    \"model\": githubModel,\n",
    "    \"name\": \"github\",\n",
    "    \"epochs\": 10,\n",
    "    \"batch_size\": 16,\n",
    "}\n",
    "crnn = {\n",
    "    \"model\": model,\n",
    "    \"name\": \"crnn\",\n",
    "    \"epochs\": 15,\n",
    "    \"batch_size\": 16,\n",
    "}\n",
    "cnn = {\n",
    "    \"model\": cnn,\n",
    "    \"name\": \"cnn\",\n",
    "    \"epochs\": 10,\n",
    "    \"batch_size\": 32,\n",
    "}\n",
    "rnn = {\n",
    "    \"model\": rnn,\n",
    "    \"name\": \"rnn\",\n",
    "    \"epochs\": 100,\n",
    "    \"batch_size\": 64,\n",
    "}\n",
    "attrnn = {\n",
    "    \"model\": attrnn,\n",
    "    \"name\": \"attrnn\",\n",
    "    \"epochs\": 100,\n",
    "    \"batch_size\": 64,\n",
    "}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train Models\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "2899/2899 [==============================] - 80s 27ms/step - loss: 0.6354 - accuracy: 0.7638 - val_loss: 0.6801 - val_accuracy: 0.7484\n",
      "Epoch 2/10\n",
      "2899/2899 [==============================] - 95s 33ms/step - loss: 0.6173 - accuracy: 0.7707 - val_loss: 0.6978 - val_accuracy: 0.7452\n",
      "Epoch 3/10\n",
      "2899/2899 [==============================] - 75s 26ms/step - loss: 0.6029 - accuracy: 0.7768 - val_loss: 0.7401 - val_accuracy: 0.7440\n",
      "Epoch 4/10\n",
      "2899/2899 [==============================] - 96s 33ms/step - loss: 0.5809 - accuracy: 0.7845 - val_loss: 0.6900 - val_accuracy: 0.7546\n",
      "Epoch 5/10\n",
      "2899/2899 [==============================] - 94s 33ms/step - loss: 0.5654 - accuracy: 0.7912 - val_loss: 0.7134 - val_accuracy: 0.7455\n"
     ]
    }
   ],
   "source": [
    "# models = [cnn, rnn, crnn, seq, git]\n",
    "models = [cnn]\n",
    "histories = []\n",
    "\n",
    "history = cnn[\"model\"].fit(\n",
    "    x_train_arr,\n",
    "    y_train_encoded,\n",
    "    validation_data=(x_test_arr, y_test_encoded),\n",
    "    batch_size=cnn[\"batch_size\"],\n",
    "    epochs=cnn[\"epochs\"],\n",
    "    verbose=1,\n",
    "    callbacks=earlyStopping,\n",
    ")\n",
    "histories.append((cnn[\"name\"], history))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/thabnir/maisproj/.venv/lib/python3.10/site-packages/keras/src/engine/training.py:3079: UserWarning: You are saving your model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')`.\n",
      "  saving_api.save_model(\n"
     ]
    }
   ],
   "source": [
    "cnn[\"model\"].save(\"cnn.h5\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.6816518902778625, 0.7530515193939209]"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loaded_model = tf.keras.models.load_model(\"cnn.h5\")\n",
    "# test\n",
    "print(\"test\")\n",
    "loaded_model.test_on_batch(x_test_arr, y_test_encoded)\n",
    "\n",
    "# for model_dict in models:\n",
    "#     print(f\"Training {model_dict['name']}\")\n",
    "#     history = model_dict[\"model\"].fit(\n",
    "#         x_train_arr,\n",
    "#         y_train_encoded,\n",
    "#         validation_data=(x_test_arr, y_test_encoded),\n",
    "#         batch_size=model_dict[\"batch_size\"],\n",
    "#         epochs=model_dict[\"epochs\"],\n",
    "#         verbose=1,\n",
    "#         callbacks=earlyStopping,\n",
    "#     )\n",
    "#     histories.append((model_dict[\"name\"], history))\n",
    "#     print(f\"Finished training {model_dict['name']}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Keras Results\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_accuracy(name, history):\n",
    "    acc = history.history[\"accuracy\"]\n",
    "    val_acc = history.history[\"val_accuracy\"]\n",
    "    epochs = range(1, len(acc) + 1)\n",
    "    plt.plot(epochs, acc, \"-\", label=f\"{name} Training Accuracy\")\n",
    "    plt.plot(epochs, val_acc, \":\", label=f\"{name} Validation Accuracy\")\n",
    "    plt.title(f\"{name} Training and Validation Accuracy\")\n",
    "    plt.xlabel(\"Epoch\")\n",
    "    plt.ylabel(\"Accuracy\")\n",
    "    plt.legend(loc=\"upper left\")\n",
    "    plt.plot()\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "def plot_loss(name, history):\n",
    "    loss = history.history[\"loss\"]\n",
    "    val_loss = history.history[\"val_loss\"]\n",
    "    epochs = range(1, len(loss) + 1)\n",
    "\n",
    "    plt.plot(epochs, loss, \"-\", label=f\"{name} Training Loss\")\n",
    "    plt.plot(epochs, val_loss, \":\", label=f\"{name} Validation Loss\")\n",
    "    plt.title(f\"{name} Training and Validation Loss\")\n",
    "    plt.xlabel(\"Epoch\")\n",
    "    plt.ylabel(\"Loss\")\n",
    "    plt.legend(loc=\"upper left\")\n",
    "    plt.plot()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for history in histories:\n",
    "    plot_accuracy(name=history[0], history=history[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for history in histories:\n",
    "    plot_loss(name=history[0], history=history[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving cnn to cnn_lang_classifier.keras\n"
     ]
    }
   ],
   "source": [
    "for model in models:\n",
    "    print(f\"Saving {model['name']} to {model['name']}_lang_classifier.keras\")\n",
    "    model[\"model\"].save(f\"{model['name']}_lang_classifier.keras\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluating cnn\n",
      "2899/2899 [==============================] - 20s 7ms/step - loss: 0.6674 - accuracy: 0.7601\n",
      "cnn Loss: 0.6673745512962341\n",
      "cnn Accuracy: 76.01%\n",
      "Random Accuracy: 20.00%\n",
      "Difference: 56.01%\n"
     ]
    }
   ],
   "source": [
    "loss_acc = []\n",
    "for model in models:\n",
    "    print(f\"Evaluating {model['name']}\")\n",
    "    [loss, accuracy] = model[\"model\"].evaluate(\n",
    "        x_test_arr, y_test_encoded, verbose=1\n",
    "    )\n",
    "    loss_acc.append((model[\"name\"], loss, accuracy))\n",
    "    print(f\"{model['name']} Loss: {loss}\")\n",
    "    print(f\"{model['name']} Accuracy: {accuracy * 100:.2f}%\")\n",
    "\n",
    "    # compare to random:\n",
    "    random_accuracy = 1 / len(langs)\n",
    "    print(f\"Random Accuracy: {random_accuracy * 100:.2f}%\")\n",
    "    print(f\"Difference: {(accuracy - random_accuracy) * 100:.2f}%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2899/2899 [==============================] - 20s 7ms/step\n"
     ]
    }
   ],
   "source": [
    "sns.set()\n",
    "label_encoder = LabelEncoder()\n",
    "y_preds = []\n",
    "for model in models:\n",
    "    y_preds.append(model[\"model\"].predict(x_test_arr))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for y_predicted in y_preds:\n",
    "    test = label_encoder.fit_transform(\n",
    "        y_test_encoded.argmax(axis=1)\n",
    "    )  # Replace y_test with your actual labels\n",
    "    pred = label_encoder.transform(y_predicted.argmax(axis=1))\n",
    "\n",
    "    mat = confusion_matrix(test, pred, normalize=\"true\")\n",
    "\n",
    "    sns.heatmap(\n",
    "        mat,\n",
    "        square=True,\n",
    "        annot=True,\n",
    "        fmt=\".2f\",\n",
    "        cbar=False,\n",
    "        cmap=\"Blues\",\n",
    "        xticklabels=langs,\n",
    "        yticklabels=langs,\n",
    "    )\n",
    "\n",
    "    plt.xlabel(\"Predicted label\")\n",
    "    plt.ylabel(\"Actual label\")\n",
    "    plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.10.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}