sldkfj

data.py

@@ -65,10 +65,11 @@ def read_config(config_file):
 
 	#[pretraining]
 	config.asr_path=parser.get("pretraining", "asr_path")
-	config.pretraining_type=int(parser.get("pretraining", "pretraining_type")) # 0 - no pre-training, 1 - phoneme pre-training, 2 - word pre-training, 3 - embedding-constrained word pre-training
+	config.pretraining_type=int(parser.get("pretraining", "pretraining_type")) # 0 - no pre-training, 1 - phoneme pre-training, 2 - phoneme + word pre-training, 3 - word pre-training
 	if config.pretraining_type == 0: config.starting_unfreezing_index = 1 + len(config.word_rnn_num_hidden) + len(config.phone_rnn_num_hidden) + len(config.cnn_N_filt)
 	if config.pretraining_type == 1: config.starting_unfreezing_index = 1 + len(config.word_rnn_num_hidden)
 	if config.pretraining_type == 2: config.starting_unfreezing_index = 1
+	if config.pretraining_type == 3: config.starting_unfreezing_index = 1
 	config.pretraining_lr=float(parser.get("pretraining", "pretraining_lr"))
 	config.pretraining_batch_size=int(parser.get("pretraining", "pretraining_batch_size"))
 	config.pretraining_num_epochs=int(parser.get("pretraining", "pretraining_num_epochs"))
@@ -82,6 +83,10 @@ def read_config(config_file):
 	config.training_batch_size=int(parser.get("training", "training_batch_size"))
 	config.training_num_epochs=int(parser.get("training", "training_num_epochs"))
 	config.dataset_subset_percentage=float(parser.get("training", "dataset_subset_percentage"))
+	config.train_wording_path=parser.get("training", "train_wording_path")
+	if config.train_wording_path=="None": config.train_wording_path = None
+	config.test_wording_path=parser.get("training", "test_wording_path")
+	if config.test_wording_path=="None": config.test_wording_path = None
 
 	# compute downsample factor (divide T by this number)
 	config.phone_downsample_factor = 1

main.py

@@ -11,20 +11,14 @@
 parser.add_argument('--train', action='store_true', help='run SLU training')
 parser.add_argument('--restart', action='store_true', help='load checkpoint from a previous run')
 parser.add_argument('--config_path', type=str, help='path to config file with hyperparameters, etc.')
-parser.add_argument('--train_wording_path', type=str, default=None, help='path to .txt file containing wordings that will be allowed during training')
-parser.add_argument('--test_wording_path', type=str, default=None, help='path to .txt file containing wordings that will be allowed during validation/testing')
 args = parser.parse_args()
 pretrain = args.pretrain
 train = args.train
 restart = args.restart
 config_path = args.config_path
-train_wording_path = args.train_wording_path
-test_wording_path = args.test_wording_path
 
 # Read config file
 config = read_config(config_path)
-config.train_wording_path = train_wording_path
-config.test_wording_path = test_wording_path
 torch.manual_seed(config.seed); np.random.seed(config.seed)
 
 if pretrain:

training.py

@@ -58,7 +58,8 @@ def train(self, dataset, print_interval=100):
 				num_examples += batch_size
 				phoneme_loss, word_loss, phoneme_acc, word_acc = self.model(x,y_phoneme,y_word)
 				if self.config.pretraining_type == 1: loss = phoneme_loss
-				else: loss = phoneme_loss + word_loss
+				if self.config.pretraining_type == 2: loss = phoneme_loss + word_loss
+				if self.config.pretraining_type == 3: loss = word_loss
 				self.optimizer.zero_grad()
 				loss.backward()
 				self.optimizer.step()