Add training

cnn.py

@@ -52,7 +52,7 @@ class CNNetwork(nn.Module):
             nn.MaxPool2d(kernel_size=2)
         )
         self.flatten = nn.Flatten()
-        self.linear = nn.Linear(128 * 5 * 4, 10)
+        self.linear = nn.Linear(128 * 5 * 11, 10)
         self.softmax = nn.Softmax(dim=1)
 
     def forward(self, input_data):

dataset.py

@@ -18,6 +18,7 @@ class VoiceDataset(Dataset):
         # file processing
         self._data_path = os.path.join(data_directory)
         self._labels = os.listdir(self._data_path)
+        self.label_mapping = {label: i for i, label in enumerate(self._labels)}
         self.audio_files_labels = self._join_audio_files()
 
         self.device = device
@@ -50,7 +51,9 @@ class VoiceDataset(Dataset):
         
         # apply transformation
         wav = self.transformation(wav)
-        return wav, label
+
+        # return wav and integer representation of the label
+        return wav, self.label_mapping[label]
 
 
     def _join_audio_files(self):

notebooks/playground.ipynb

@@ -3,7 +3,7 @@
   {
    "cell_type": "code",
    "execution_count": 8,
-   "id": "46dbbffd",
+   "id": "55895dc1",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -14,7 +14,7 @@
   {
    "cell_type": "code",
    "execution_count": 10,
-   "id": "56056453",
+   "id": "d1e95c40",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -25,7 +25,7 @@
   {
    "cell_type": "code",
    "execution_count": 86,
-   "id": "b5cbac9d",
+   "id": "0ae6ce32",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -38,7 +38,7 @@
   {
    "cell_type": "code",
    "execution_count": 85,
-   "id": "1970ad63",
+   "id": "4200acc4",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -49,7 +49,7 @@
   {
    "cell_type": "code",
    "execution_count": 78,
-   "id": "c28b0c5e",
+   "id": "b98d408a",
    "metadata": {},
    "outputs": [
     {
@@ -70,8 +70,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 80,
-   "id": "69025839",
+   "execution_count": 97,
+   "id": "f26723ab",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -81,13 +81,13 @@
     "        hop_length=512,\n",
     "        n_mels=64\n",
     "    )\n",
-    "dataset = VoiceDataset('../data', mel_spectrogram, 16000, device)"
+    "dataset = VoiceDataset('../data/train', mel_spectrogram, 16000, device)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 81,
-   "id": "8dfbb1b4",
+   "execution_count": 93,
+   "id": "7664a918",
    "metadata": {},
    "outputs": [
     {
@@ -96,7 +96,7 @@
        "5718"
       ]
      },
-     "execution_count": 81,
+     "execution_count": 93,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -108,7 +108,7 @@
   {
    "cell_type": "code",
    "execution_count": 82,
-   "id": "1071e53d",
+   "id": "0adfe082",
    "metadata": {},
    "outputs": [
     {
@@ -136,7 +136,7 @@
   {
    "cell_type": "code",
    "execution_count": 83,
-   "id": "7a6d8133",
+   "id": "6f095274",
    "metadata": {},
    "outputs": [
     {
@@ -157,7 +157,7 @@
   {
    "cell_type": "code",
    "execution_count": 87,
-   "id": "4b8f75a0",
+   "id": "362d6f74",
    "metadata": {},
    "outputs": [
     {
@@ -200,13 +200,104 @@
     "summary(cnn, (1, 64, 44))"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 91,
+   "id": "d2da6515",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "tensor(0)"
+      ]
+     },
+     "execution_count": 91,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "torch.tensor(0)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 95,
+   "id": "8a10cc8c",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "{'aman': 0, 'imran': 1, 'labib': 2}"
+      ]
+     },
+     "execution_count": 95,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "dataset.label_mapping"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 98,
+   "id": "e65a95c3",
+   "metadata": {},
+   "outputs": [
+    {
+     "ename": "TypeError",
+     "evalue": "join() argument must be str, bytes, or os.PathLike object, not 'int'",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
+      "Cell \u001b[0;32mIn[98], line 1\u001b[0m\n\u001b[0;32m----> 1\u001b[0m \u001b[43mdataset\u001b[49m\u001b[43m[\u001b[49m\u001b[38;5;241;43m0\u001b[39;49m\u001b[43m]\u001b[49m\n",
+      "File \u001b[0;32m~/ml-sandbox/VoID/notebooks/../dataset.py:41\u001b[0m, in \u001b[0;36mVoiceDataset.__getitem__\u001b[0;34m(self, index)\u001b[0m\n\u001b[1;32m     38\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m__getitem__\u001b[39m(\u001b[38;5;28mself\u001b[39m, index):\n\u001b[1;32m     39\u001b[0m     \u001b[38;5;66;03m# get file\u001b[39;00m\n\u001b[1;32m     40\u001b[0m     file, label \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39maudio_files_labels[index]\n\u001b[0;32m---> 41\u001b[0m     filepath \u001b[38;5;241m=\u001b[39m \u001b[43mos\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mpath\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mjoin\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_data_path\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mlabel\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mfile\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     43\u001b[0m     \u001b[38;5;66;03m# load wav\u001b[39;00m\n\u001b[1;32m     44\u001b[0m     wav, sr \u001b[38;5;241m=\u001b[39m torchaudio\u001b[38;5;241m.\u001b[39mload(filepath, normalize\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mTrue\u001b[39;00m)\n",
+      "File \u001b[0;32m~/anaconda3/envs/void/lib/python3.9/posixpath.py:90\u001b[0m, in \u001b[0;36mjoin\u001b[0;34m(a, *p)\u001b[0m\n\u001b[1;32m     88\u001b[0m             path \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m sep \u001b[38;5;241m+\u001b[39m b\n\u001b[1;32m     89\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m (\u001b[38;5;167;01mTypeError\u001b[39;00m, \u001b[38;5;167;01mAttributeError\u001b[39;00m, \u001b[38;5;167;01mBytesWarning\u001b[39;00m):\n\u001b[0;32m---> 90\u001b[0m     \u001b[43mgenericpath\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_check_arg_types\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mjoin\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43ma\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mp\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     91\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m\n\u001b[1;32m     92\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m path\n",
+      "File \u001b[0;32m~/anaconda3/envs/void/lib/python3.9/genericpath.py:152\u001b[0m, in \u001b[0;36m_check_arg_types\u001b[0;34m(funcname, *args)\u001b[0m\n\u001b[1;32m    150\u001b[0m         hasbytes \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mTrue\u001b[39;00m\n\u001b[1;32m    151\u001b[0m     \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m--> 152\u001b[0m         \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mTypeError\u001b[39;00m(\u001b[38;5;124mf\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mfuncname\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m() argument must be str, bytes, or \u001b[39m\u001b[38;5;124m'\u001b[39m\n\u001b[1;32m    153\u001b[0m                         \u001b[38;5;124mf\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mos.PathLike object, not \u001b[39m\u001b[38;5;132;01m{\u001b[39;00ms\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__class__\u001b[39m\u001b[38;5;241m.\u001b[39m\u001b[38;5;18m__name__\u001b[39m\u001b[38;5;132;01m!r}\u001b[39;00m\u001b[38;5;124m'\u001b[39m) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[1;32m    154\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m hasstr \u001b[38;5;129;01mand\u001b[39;00m hasbytes:\n\u001b[1;32m    155\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mTypeError\u001b[39;00m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mCan\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mt mix strings and bytes in path components\u001b[39m\u001b[38;5;124m\"\u001b[39m) \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n",
+      "\u001b[0;31mTypeError\u001b[0m: join() argument must be str, bytes, or os.PathLike object, not 'int'"
+     ]
+    }
+   ],
+   "source": [
+    "dataset[0]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 104,
+   "id": "a1357e5b",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'2023-05-12 22:28:06.556207'"
+      ]
+     },
+     "execution_count": 104,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "from datetime import datetime\n",
+    "now = datetime.now()"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "888726ed",
+   "id": "190c8d4b",
    "metadata": {},
    "outputs": [],
-   "source": []
+   "source": [
+    "now.strftime(\"%Y%m%d-%H%M%S\")"
+   ]
   }
  ],
  "metadata": {

train.py

@@ -0,0 +1,82 @@
+from datetime import datetime
+from tqdm import tqdm
+
+# torch
+import torch
+import torchaudio
+from torch import nn
+from torch.utils.data import DataLoader
+
+# internal
+from dataset import VoiceDataset
+from cnn import CNNetwork
+
+BATCH_SIZE = 128
+EPOCHS = 100
+LEARNING_RATE = 0.001
+
+TRAIN_FILE="data/train"
+SAMPLE_RATE=16000
+
+def train(model, dataloader, loss_fn, optimizer, device, epochs):
+  for i in tqdm(range(epochs), "Training model..."):
+    print(f"Epoch {i + 1}")
+
+    train_epoch(model, dataloader, loss_fn, optimizer, device)
+
+    print (f"----------------------------------- \n")
+  
+  print("---- Finished Training ----")
+  
+
+def train_epoch(model, dataloader, loss_fn, optimizer, device):
+  for x, y in dataloader:
+    x, y = x.to(device), y.to(device)
+
+    # calculate loss
+    pred = model(x)
+    loss = loss_fn(pred, y)
+
+    # backprop and update weights
+    optimizer.zero_grad()
+    loss.backward()
+    optimizer.step()
+  
+  print(f"Loss: {loss.item()}") 
+
+if __name__ == "__main__":
+    if torch.cuda.is_available():
+        device = "cuda"
+    else:
+        device = "cpu"
+    print(f"Using {device} device.")
+
+    # instantiating our dataset object and create data loader
+    mel_spectrogram = torchaudio.transforms.MelSpectrogram(
+        sample_rate=SAMPLE_RATE,
+        n_fft=1024,
+        hop_length=512,
+        n_mels=64
+    )
+
+    train_dataset = VoiceDataset(TRAIN_FILE, mel_spectrogram, SAMPLE_RATE, device)
+    train_dataloader = DataLoader(train_dataset, batch_size=BATCH_SIZE, shuffle=True)
+
+    # construct model
+    model = CNNetwork().to(device)
+    print(model)
+
+    # init loss function and optimizer
+    loss_fn = nn.CrossEntropyLoss()
+    optimizer = torch.optim.Adam(model.parameters(), lr=LEARNING_RATE)
+
+
+    # train model
+    train(model, train_dataloader, loss_fn, optimizer, device, EPOCHS)
+
+    # save model
+    now = datetime.now()
+    now = now.strftime("%Y%m%d_%H%M%S")
+    model_filename = f"models/void_{now}.pth"
+    torch.save(model.state_dict(), model_filename)
+    print(f"Trained feed forward net saved at {model_filename}")