Refactor configuration to yaml

.gitignore

@@ -1,3 +1,4 @@
+train_config.yaml
 notebooks/*.json
 notebooks/*.state
 FullyConnected

train.py

@@ -7,35 +7,32 @@
 import torch.nn as nn
 import torch.nn.functional as F
 import torchvision
+import yaml
 from cuda import *
 from data import *
 from net import *
 from torch.autograd import Variable
 from torch.utils.data import DataLoader
 from tqdm import tqdm
 
+config_path = Path('train_config.yaml')
+config = None
+with config_path.open('r') as f:
+    config = yaml.load(f)
 H = {}  # Training history and statistics
 CUDA, device = get_cuda_if_available()
 H['cuda'] = CUDA
 
-train_path = Path(
-    '/home/arccha/.kaggle/competitions/digit-recognizer/train.csv')
-if not train_path.exists():
-    train_path = '../train.csv'
-else:
-    train_path = str(train_path)
-DATA_NUM = 42000  # 42000 - max
-H['data_num'] = DATA_NUM
-VALIDATION_NUM = 4200
-H['validation_num'] = VALIDATION_NUM
-BATCH_SIZE = 32
-H['batch_size'] = BATCH_SIZE
+train_path = config['train_path']
+H['data_num'] = config['data_num']
+H['validation_num'] = config['validation_num']
+H['batch_size'] = config['batch_size']
 train_dataset, validation_dataset = train_validation_split(
-    train_path, max_rows=DATA_NUM, validation_num=VALIDATION_NUM, pretransform=True)
+    train_path, max_rows=config['data_num'], validation_num=config['validation_num'], pretransform=True)
 train_loader = DataLoader(dataset=train_dataset,
-                          batch_size=BATCH_SIZE, shuffle=True, num_workers=4, pin_memory=True)
+                          batch_size=config['batch_size'], shuffle=True, num_workers=4, pin_memory=True)
 validation_loader = DataLoader(
-    dataset=validation_dataset, batch_size=BATCH_SIZE, shuffle=False, num_workers=1, pin_memory=True)
+    dataset=validation_dataset, batch_size=config['batch_size'], shuffle=False, num_workers=1, pin_memory=True)
 
 validation_classes = np.zeros(10)
 for x, y in tqdm(validation_loader, desc='Validation stats'):
@@ -48,19 +45,17 @@
 net_dir.mkdir(parents=True, exist_ok=True)
 net.to(device)
 
-LR = 0.001
-optimizer = torch.optim.Adam(net.parameters(), lr=LR)
+optimizer = torch.optim.Adam(net.parameters(), lr=config['learning_rate'])
 H['optimizer'] = str(optimizer)
 criterion = nn.CrossEntropyLoss()
 H['criterion'] = str(criterion)
 
-EPOCH_NUM = 50
-H['epoch_num'] = EPOCH_NUM
+H['epoch_num'] = config['epoch_num']
 H['loss'] = []
 H['train_acc'] = []
 H['test_acc'] = []
 start = time.process_time()
-for epoch in tqdm(range(EPOCH_NUM), desc='Total'):
+for epoch in tqdm(range(config['epoch_num']), desc='Total'):
     running_loss = 0.0
     for x, y in tqdm(train_loader, desc='Epoch ' + str(epoch)):
         x = x.to(device)
@@ -71,22 +66,23 @@
         loss.backward()
         optimizer.step()
         running_loss += loss.item()
-    H['loss'].append(running_loss / (DATA_NUM / BATCH_SIZE))
+    H['loss'].append(
+        running_loss / (config['data_num'] / config['batch_size']))
     acc = 0
     for x, y_true in tqdm(train_loader, desc='Train acc ' + str(epoch)):
         x = x.to(device)
         y_true = y_true.to(device)
         y_pred = net(x).argmax(dim=1)
         acc += y_true.eq(y_pred).sum()
-    acc = float(acc) / (len(train_loader) * BATCH_SIZE)
+    acc = float(acc) / (len(train_loader) * config['batch_size'])
     H['train_acc'].append(acc)
     acc = 0
     for x, y_true in tqdm(validation_loader, desc='Validation ' + str(epoch)):
         x = x.to(device)
         y_true = y_true.to(device)
         y_pred = net(x).argmax(dim=1)
         acc += y_true.eq(y_pred).sum()
-    acc = float(acc) / VALIDATION_NUM
+    acc = float(acc) / config['validation_num']
     H['test_acc'].append(acc)
     net_state_path = net_dir.joinpath('net' + str(epoch) + '.state')
     net_state_path.touch(exist_ok=True)

train_config.yaml.sample

@@ -0,0 +1,6 @@
+train_path: '/home/arccha/.kaggle/competitions/digit-recognizer/train.csv'
+data_num: 42000
+validation_num: 4200
+batch_size: 32
+learning_rate: 0.001
+epoch_num: 50