class Hyperparamters:
# Train parameters
num_train_epochs = 20
print_step = 100
batch_size = 8#64
summary_step = 10
num_saved_per_epoch = 3
max_to_keep = 100
logdir = 'logdir/CML_Denses'
file_save_model = 'model/CML_Denses'
inference_model = 'CML_Denses'
# Train/Test data
data_dir = os.path.join(pwd,'data')
train_data = 'train_onehot.csv'
test_data = 'test_onehot.csv'
# Load vocabulcary dict
dict_id2label,dict_label2id = load_vocabulary(os.path.join(pwd,'data','vocabulary_label.txt') )
label_vocabulary = list(dict_id2label.values())
# Optimization parameters
warmup_proportion = 0.1
use_tpu = None
do_lower_case = True
learning_rate = 5e-5
# Sequence and Label
sequence_length = 60
num_labels = len(list(dict_id2label))
# ALBERT
model = 'albert_small_zh_google'
bert_path = os.path.join(pwd,model)
vocab_file = os.path.join(pwd,model,'vocab_chinese.txt')
init_checkpoint = os.path.join(pwd,model,'albert_model.ckpt')
saved_model_path = os.path.join(pwd,'model')
class Hyperparamters:
# Train parameters
num_train_epochs = 400
print_step = 100
batch_size = 64
summary_step = 10
num_saved_per_epoch = 1
max_to_keep = 100
logdir = 'logdir/model_01'
file_save_model = 'model/model_01'
# Train/Test data
data_dir = os.path.join(pwd, 'data')
train_data = 'train_onehot.csv'
test_data = 'test_onehot.csv'
# Predict model file
file_model = 'model/saved_01'
# Load vocabulcary dict
dict_id2label, dict_label2id = load_vocabulary(
os.path.join(pwd, 'data', 'vocabulary_label.txt'))
print(dict_id2label)
label_vocabulary = list(dict_id2label.values())
print(dict_label2id)
# Optimization parameters
warmup_proportion = 0.1
use_tpu = None
do_lower_case = True
learning_rate = 5e-5
# TextCNN parameters
num_filters = 128
filter_sizes = [2, 3, 4, 5, 6, 7]
embedding_size = 384
keep_prob = 0.5
# Sequence and Label
sequence_length = 60
num_labels = len(list(dict_id2label))
# ALBERT
model = 'albert_small_zh_google'
bert_path = os.path.join(pwd, model)
vocab_file = os.path.join(pwd, model, 'vocab_chinese.txt')
init_checkpoint = os.path.join(pwd, model, 'albert_model.ckpt')
saved_model_path = os.path.join(pwd, 'model')
class Hyperparamters:
# Train Parameters
print_step = 100
summary_step = 10
batch_size = 64
num_saved_per_epoch = 3
logdir = 'logdir/model_02'
# Load dict
dict_id2label, dict_label2id = load_vocabulary(
os.path.join(pwd, 'data', 'vocabulary_label.txt'))
label_vocabulary = list(dict_id2label.values())
# Optimization parameters
use_tpu = None
num_train_epochs = 20
warmup_proportion = 0.1
do_lower_case = True
learning_rate = 5e-5
# TextCNN parameters
num_filters = 128
filter_sizes = [2, 3, 4, 5, 6, 7]
embedding_size = 384
keep_prob = 0.5
# Sequence and Label
sequence_length = 60
num_labels = len(list(dict_id2label))
# BERT model
model = 'albert_small_zh_google'
bert_path = os.path.join(pwd, model)
data_dir = os.path.join(pwd, 'data')
vocab_file = os.path.join(pwd, model, 'vocab_chinese.txt')
init_checkpoint = os.path.join(pwd, model, 'albert_model.ckpt')
saved_model_path = os.path.join(pwd, 'model')