Updated find_lr advances #26

find_learning_rates.py

@@ -1,5 +1,6 @@
 import torch
 import yaml
+import argparse
 from datetime import datetime
 from torch.utils.data import DataLoader, random_split
 from datasets.dataset import get_dataset
@@ -24,7 +25,7 @@ def main(config_path, optimizer_type, optimizer_params, batch_size):
     # Split data
     total_size = len(data)
     test_size = int(total_size * config['data']['test_size'])
-    val_size = int((total_size - test_size) * config['data']['val_size'])
+    val_size = int(total_size * config['data']['val_size'])
     train_size = total_size - test_size - val_size
 
     data_train, _ = random_split(data, [train_size + val_size, test_size], generator=torch.Generator().manual_seed(config['random_seed']))
@@ -36,7 +37,6 @@ def main(config_path, optimizer_type, optimizer_params, batch_size):
     # Model setup
     model_factory = ModelFactory()
     model = model_factory.create(config['model']['type'], **config['model']['parameters']).to(device)
-    print(model)
 
     # Loss setup
     loss_factory = LossFactory()
@@ -55,6 +55,11 @@ def main(config_path, optimizer_type, optimizer_params, batch_size):
     plot_lr_vs_loss(log_lrs, losses, path.join(config['paths']['plot_path'], plot_filename))
 
 if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description='Process configuration file, optimizer types, and batch sizes.')
+    parser.add_argument('config_filename', type=str, help='Filename of the configuration file within the "config" directory')
+
+    args = parser.parse_args()
+
     batch_sizes = [8, 16, 32, 64]
     optimizer_types = ["SGD", "Adam"]
     adam_params = {
@@ -71,7 +76,8 @@ def main(config_path, optimizer_type, optimizer_params, batch_size):
         "nesterov": False
     }
 
-    config_path = "config/fine_tuning_config.yaml"
+    # Build the path to the configuration file within the 'config' directory
+    config_path = f"config/{args.config_filename}"
 
     for optimizer_type in optimizer_types:
         for batch_size in batch_sizes:

utils/training.py

@@ -1,7 +1,7 @@
 import math
 import torch
 
-def find_lr(model, train_loader, criterion, optimizer_class, optimizer_params, init_value=1e-8, final_value=10, beta=0.98, device=None):
+def find_lr(model, train_loader, criterion, optimizer_class, optimizer_params, init_value=1e-8, final_value=1e-1, beta=0.98, device=None):
     if device is None:
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     model.to(device)