pretrained_model_name=args.pretrained_bert_model, factory=nf)
hidden_size = pretrained_bert_model.local_parameters["hidden_size"]
tokenizer = BertTokenizer.from_pretrained(args.pretrained_bert_model)
data_desc = JointIntentSlotDataDesc(args.data_dir, args.do_lower_case,
args.dataset_name, args.none_slot_label,
args.pad_label)
# Create sentence classification loss on top
classifier = nemo_nlp.JointIntentSlotClassifier(
hidden_size=hidden_size,
num_intents=data_desc.num_intents,
num_slots=data_desc.num_slots,
dropout=args.fc_dropout)
loss_fn = nemo_nlp.JointIntentSlotLoss(num_slots=data_desc.num_slots)
def create_pipeline(num_samples=-1,
batch_size=32,
num_gpus=1,
local_rank=0,
mode='train'):
nf.logger.info(f"Loading {mode} data...")
data_file = f'{data_desc.data_dir}/{mode}.tsv'
slot_file = f'{data_desc.data_dir}/{mode}_slots.tsv'
shuffle = args.shuffle_data if mode == 'train' else False
data_layer = nemo_nlp.BertJointIntentSlotDataLayer(
input_file=data_file,
slot_file=slot_file,
args.pad_label)
# Create sentence classification loss on top
classifier = nemo_nlp.JointIntentSlotClassifier(
hidden_size=hidden_size,
num_intents=data_desc.num_intents,
num_slots=data_desc.num_slots,
dropout=args.fc_dropout)
if args.class_balancing == 'weighted_loss':
# Using weighted loss will enable weighted loss for both intents and slots
# Use the intent_loss_weight hyperparameter to adjust intent loss to
# prevent overfitting or underfitting.
loss_fn = nemo_nlp.JointIntentSlotLoss(
num_slots=data_desc.num_slots,
slot_classes_loss_weights=data_desc.slot_weights,
intent_classes_loss_weights=data_desc.intent_weights,
intent_loss_weight=args.intent_loss_weight)
else:
loss_fn = nemo_nlp.JointIntentSlotLoss(num_slots=data_desc.num_slots)
def create_pipeline(num_samples=-1,
batch_size=32,
num_gpus=1,
local_rank=0,
mode='train'):
nemo.logging.info(f"Loading {mode} data...")
data_file = f'{data_desc.data_dir}/{mode}.tsv'
slot_file = f'{data_desc.data_dir}/{mode}_slots.tsv'
shuffle = args.shuffle_data if mode == 'train' else False
num_slots = 12
pad_label = num_slots - 1
else:
nf.logger.info("Looks like you pass in the name of dataset that isn't "
"already supported by NeMo. Please make sure that you "
"build the preprocessing method for it.")
# Create sentence classification loss on top
hidden_size = pretrained_bert_model.local_parameters["hidden_size"]
classifier = nemo_nlp.JointIntentSlotClassifier(hidden_size=hidden_size,
num_intents=num_intents,
num_slots=num_slots,
dropout=args.fc_dropout)
loss_fn = nemo_nlp.JointIntentSlotLoss(num_slots=num_slots)
tokenizer = BertTokenizer.from_pretrained(args.pretrained_bert_model)
def create_pipeline(data_file,
slot_file,
max_seq_length,
batch_size=32,
num_samples=-1,
shuffle=True,
num_gpus=1,
local_rank=0,
mode='train'):
nf.logger.info(f"Loading {mode} data...")
data_layer = nemo_nlp.BertJointIntentSlotDataLayer(