first version of trainer

src/train_and_evaluate.py

@@ -1,3 +1,71 @@
+# Fernando López Gavilánez, 2023
 
+from torch.utils.tensorboard import SummaryWriter
+from torch.utils.tensorboard.summary import hparams
 
-# Use here the rest of the elements to train the model
+import keyword_spotting.trainer as trainer
+
+from keyword_spotting.model import LeNetAudio
+
+
+def main(args):
+
+    if (not os.path.ispath(args.data_path)) or (not os.path.ispath(args.partition_path)):
+        raise Exception('Non valid data or partition paths!')
+
+    model = LeNetAudio(
+        num_classes,
+        window_size=int(args.window_size*args.sampling_rate)
+        )
+
+    trainer = trainer.AudioTrainer(
+        model,
+        args.seed,
+        args.num_classes,
+        args.cuda,
+        args.window_size,
+        args.sampling_rate
+    )
+
+    trainer.prepare_data(args.data_path)
+
+    trainer.train(
+        args.epochs,
+        args.optimizer,
+        args.lr,
+        args.momentum,
+        args.weight_decay,
+        args.balance
+        )
+
+    # evaluate here
+
+if __name__ == '__main__':
+
+    parser = argparse.ArgumentParser(description="Script to partitions for donateacry dataset")
+
+    # Hyperparameters
+    parser.add_argument('--seed', type=int, default=0, metavar='S', help='random seed')
+    parser.add_argument('--use_sampler', dest='use_batch_sampler', action='store_true', help='use Weighted Random Sampler to deal with data unbalance')
+    parser.add_argument('--batch_size', type=int, default=8, metavar='N', help='Batch size to use')
+    parser.add_argument('--epochs', type=int, default=200, metavar='N', help='number of epochs to train')
+    parser.add_argument('--lr', type=float, default=0.001, metavar='LR', help='learning rate')
+    parser.add_argument('--lr_patience', type=int, default=20, help='number of epochs of no loss improvement before updating lr')
+    parser.add_argument('--patience', type=int, default=60, help='number of epochs of no loss improvement before stop training')
+
+    # Optimizer
+    parser.add_argument('--optimizer', default='adam', help='optimization method: sgd | adam | rmsprop')
+    parser.add_argument('--weight_decay', type=float, default=0.0, metavar='WD', help='weight decay for the optimizer')
+    parser.add_argument('--momentum', type=float, default=0.9, metavar='M', help='SGD momentum')
+
+    # Audio
+    parser.add_argument('--sampling_rate', type=float, default=16000, metavar='SR', help='sampling rate of the audio signal')
+    parser.add_argument('--time_window', type=float, default=1.0, metavar='TW', help='time window covered by every data sample')
+
+    # Data and partitions
+    parser.add_argument('--data_path', default='', help='root path of the audio dataset')
+    parser.add_argument('--partition_path', default='', help='path to the partition folder containing the train, dev and test dataframes')
+
+    args = parser.parse_args()
+
+    main(args)