Fix: migrate to torch0.4

src/engine.py

@@ -26,13 +26,10 @@ def train_single_batch(self, users, items, ratings):
  users, items, ratings = users.cuda(), items.cuda(), ratings.cuda()
  self.opt.zero_grad()
  ratings_pred = self.model(users, items)
- loss = self.crit(ratings_pred, ratings)
+ loss = self.crit(ratings_pred.view(-1), ratings)
  loss.backward()
  self.opt.step()
- if self.config['use_cuda'] is True:
- loss = loss.data.cpu().numpy()[0]
- else:
- loss = loss.data.numpy()[0]
+ loss = loss.item()
  return loss
 
  def train_an_epoch(self, train_loader, epoch_id):

src/train.py

@@ -5,9 +5,14 @@
 from neumf import NeuMFEngine
 from data import SampleGenerator
 
-gmf_config = {'alias': 'gmf_factor8neg4',
+gmf_config = {'alias': 'gmf_factor8neg4-adam-decayneg4-torch4',
  'num_epoch': 200,
  'batch_size': 1024,
+ # 'optimizer': 'sgd',
+ # 'optimizer': 'rmsprop',
+ # 'rmsprop_lr': 1e-3,
+ # 'rmsprop_alpha': 0.99,
+ # 'rmsprop_momentum': 0,
  'optimizer': 'adam',
  'adam_lr': 1e-3,
  'num_users': 6040,
@@ -16,7 +21,7 @@
  'num_negative': 4,
  'l2_regularization': 0.01,
  'use_cuda': True,
- 'device_id': 7,
+ 'device_id': 0,
  'model_dir':'checkpoints/{}_Epoch{}_HR{:.4f}_NDCG{:.4f}.model'}
 
 mlp_config = {'alias': 'mlp_factor8neg4_bz256_166432168_pretrain_reg_0.0000001',
@@ -73,10 +78,10 @@
 sample_generator = SampleGenerator(ratings=ml1m_rating)
 evaluate_data = sample_generator.evaluate_data
 # Specify the exact model
-# config = gmf_config
-# engine = GMFEngine(config)
-config = mlp_config
-engine = MLPEngine(config)
+config = gmf_config
+engine = GMFEngine(config)
+# config = mlp_config
+# engine = MLPEngine(config)
 # config = neumf_config
 # engine = NeuMFEngine(config)
 for epoch in range(config['num_epoch']):

src/utils.py

@@ -23,11 +23,13 @@ def use_cuda(enabled, device_id=0):
 
 
 def use_optimizer(network, params):
- if params['optimizer'] == 'adam':
+ if params['optimizer'] == 'sgd':
+ optimizer = torch.optim.SGD(network.parameters(), lr=params['adam_lr'])
+ elif params['optimizer'] == 'adam':
  optimizer = torch.optim.Adam(network.parameters(), lr=params['adam_lr'], weight_decay=params['l2_regularization'])
  elif params['optimizer'] == 'rmsprop':
  optimizer = torch.optim.RMSprop(network.parameters(),
- lr=params['rmsprop_lr'],
- alpha=params['rmsprop_alpha'],
- momentum=params['rmsprop_momentum'])
+ lr=params['rmsprop_lr'])
+ # alpha=params['rmsprop_alpha'],
+ # momentum=params['rmsprop_momentum'])
  return optimizer