def predict(output_key, mode):
config_path = data_config.output_path(output_key, ALL, CONFIG)
config_data = yaml.load(open(config_path))
nn_config = NNConfig(config_data)
vocab_id_mapping = json.load(
open(data_config.output_path(output_key, ALL, VOCAB_ID_MAPPING), 'r'))
dataset = load_dataset(mode=mode,
vocab_id_mapping=vocab_id_mapping,
max_seq_len=nn_config.seq_len,
sampling=False,
with_label=False)
index_iterator = SimpleIndexIterator.from_dataset(dataset)
n_sample = index_iterator.n_sample()
with tf.Session() as sess:
prefix_checkpoint = tf.train.latest_checkpoint(
data_config.model_path(key=output_key))
saver = tf.train.import_meta_graph('{}.meta'.format(prefix_checkpoint))
saver.restore(sess, prefix_checkpoint)
nn = BaseNNModel(config=None)
nn.set_graph(tf.get_default_graph())
fetches = {_key: nn.var(_key) for _key in [LABEL_PREDICT]}
labels_predict = list()
for batch_index in index_iterator.iterate(nn_config.batch_size,
shuffle=False):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches, feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_predict = labels_predict[:n_sample]
# 导出预测的label
with open(data_config.output_path(output_key, mode, LABEL_PREDICT),
'w') as file_obj:
for _label in labels_predict:
file_obj.write('{}\n'.format(_label))
def live_test(output_key):
config_path = data_config.output_path(output_key, ALL, CONFIG)
config_data = yaml.load(open(config_path))
nn_config = NNConfig(config_data)
vocab_id_mapping = json.load(
open(data_config.output_path(output_key, ALL, VOCAB_ID_MAPPING), 'r'))
with tf.Session() as sess:
prefix_checkpoint = tf.train.latest_checkpoint(
data_config.model_path(key=output_key))
saver = tf.train.import_meta_graph('{}.meta'.format(prefix_checkpoint))
saver.restore(sess, prefix_checkpoint)
nn = BaseNNModel(config=None)
nn.set_graph(tf.get_default_graph())
fetches = {
_key: nn.var(_key)
for _key in [LABEL_PREDICT, PROB_PREDICT]
}
while True:
res = input('input: ')
if res == 'quit':
break
turns = res.strip().split('|')
if len(turns) != 3:
print('invalid turns')
continue
tokens_list = list()
for turn in turns:
tokens = re.sub('\s+', ' ', turn.strip()).split(' ')
tokens_list.append(tokens)
placeholder = [[]] * (nn_config.batch_size - 1)
tid_list_0 = tokenized_to_tid_list([
tokens_list[0],
] + placeholder, vocab_id_mapping)
tid_list_1 = tokenized_to_tid_list([
tokens_list[1],
] + placeholder, vocab_id_mapping)
tid_list_2 = tokenized_to_tid_list([
tokens_list[2],
] + placeholder, vocab_id_mapping)
tid_0 = np.asarray(zero_pad_seq_list(tid_list_0,
nn_config.seq_len))
tid_1 = np.asarray(zero_pad_seq_list(tid_list_1,
nn_config.seq_len))
tid_2 = np.asarray(zero_pad_seq_list(tid_list_2,
nn_config.seq_len))
feed_dict = {
nn.var(TID_0): tid_0,
nn.var(TID_1): tid_1,
nn.var(TID_2): tid_2,
nn.var(TEST_MODE): 1
}
res = sess.run(fetches=fetches, feed_dict=feed_dict)
label = res[LABEL_PREDICT][0]
prob = res[PROB_PREDICT][0]
print('label: {}'.format(label))
print('prob: {}'.format(prob))
def train(text_version='ek',
label_version=None,
config_path='config93_naive.yaml'):
"""
python -m algo.main93_v2 train
python3 -m algo.main93_v2 train -c config_ntua93.yaml
:param text_version: string
:param label_version: string
:param config_path: string
:return:
"""
config_data = yaml.load(open(config_path))
output_key = '{}_{}_{}'.format(NNModel.name, text_version,
int(time.time()))
if label_version is not None:
output_key = '{}_{}'.format(label_version, output_key)
print('OUTPUT_KEY: {}'.format(output_key))
# 准备输出路径的文件夹
data_config.prepare_output_folder(output_key=output_key)
data_config.prepare_model_folder(output_key=output_key)
shutil.copy(config_path, data_config.output_path(output_key, ALL, CONFIG))
w2v_key = '{}_{}'.format(config_data['word']['w2v_version'], text_version)
w2v_model_path = data_config.path(ALL, WORD2VEC, w2v_key)
vocab_train_path = data_config.path(TRAIN, VOCAB, text_version)
# 加载字典集
# 在模型中会采用所有模型中支持的词向量, 并为有足够出现次数的单词随机生成词向量
vocab_meta_list = load_vocab_list(vocab_train_path)
vocabs = [
_meta['t'] for _meta in vocab_meta_list
if _meta['tf'] >= config_data['word']['min_tf']
]
# 加载词向量与相关数据
lookup_table, vocab_id_mapping, embedding_dim = load_lookup_table2(
w2v_model_path=w2v_model_path, vocabs=vocabs)
json.dump(
vocab_id_mapping,
open(data_config.output_path(output_key, ALL, VOCAB_ID_MAPPING), 'w'))
# 加载配置
nn_config = NNConfig(config_data)
train_config = TrainConfig(config_data['train'])
early_stop_metric = train_config.early_stop_metric
# 加载训练数据
datasets, output_dim = load_dataset(vocab_id_mapping=vocab_id_mapping,
max_seq_len=nn_config.seq_len,
with_label=True,
label_version=label_version)
# 初始化数据集的检索
index_iterators = {
mode: IndexIterator(datasets[mode][LABEL_GOLD])
for mode in [TRAIN, TEST]
}
# 按配置将训练数据切割成训练集和验证集
index_iterators[TRAIN].split_train_valid(train_config.valid_rate)
# 计算各个类的权重
if train_config.use_class_weights:
label_weight = {
# 参考 sklearn 中 class_weight='balanced'的公式, 实验显示效果显着
_label: float(index_iterators[TRAIN].n_sample()) /
(index_iterators[TRAIN].dim * len(_index))
for _label, _index in index_iterators[TRAIN].label_index.items()
}
else:
label_weight = {
_label: 1.
for _label in range(index_iterators[TRAIN].dim)
}
# 基于加载的数据更新配置
nn_config.set_embedding_dim(embedding_dim)
nn_config.set_output_dim(output_dim)
# 搭建神经网络
nn = NNModel(config=nn_config)
nn.build_neural_network(lookup_table=lookup_table)
batch_size = train_config.batch_size
fetches = {
mode: {_key: nn.var(_key)
for _key in fetch_key[mode]}
for mode in [TRAIN, TEST]
}
last_eval = {TRAIN: None, VALID: None, TEST: None}
model_output_prefix = data_config.model_path(key=output_key) + '/model'
best_res = {mode: None for mode in [TRAIN, VALID]}
no_update_count = {mode: 0 for mode in [TRAIN, VALID]}
max_no_update_count = 10
eval_history = {TRAIN: list(), DEV: list(), TEST: list()}
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver(tf.global_variables())
dataset = datasets[TRAIN]
index_iterator = index_iterators[TRAIN]
# 训练开始 ##########################################################################
for epoch in range(train_config.epoch):
print('== epoch {} = {} ='.format(epoch, output_key))
# 利用训练集进行训练
print('TRAIN')
n_sample = index_iterator.n_sample(TRAIN)
labels_predict = list()
labels_gold = list()
for batch_index in index_iterator.iterate(batch_size,
mode=TRAIN,
shuffle=True):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TRAIN]
}
feed_dict[nn.var(SAMPLE_WEIGHTS)] = list(
map(label_weight.get, feed_dict[nn.var(LABEL_GOLD)]))
feed_dict[nn.var(TEST_MODE)] = 0
if train_config.input_dropout_keep_prob < 1.:
for _key in [TID_0, TID_1, TID_2]:
var = nn.var(_key)
_tids = feed_dict[var]
feed_dict[var] = tid_dropout(
_tids, train_config.input_dropout_keep_prob)
res = sess.run(fetches=fetches[TRAIN], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
n_sample], labels_gold[:
n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
print_evaluation(res)
eval_history[TRAIN].append(res)
global_step = tf.train.global_step(sess, nn.var(GLOBAL_STEP))
if train_config.valid_rate == 0.:
if best_res[TRAIN] is None or res[
early_stop_metric] > best_res[TRAIN][early_stop_metric]:
best_res[TRAIN] = res
no_update_count[TRAIN] = 0
saver.save(sess,
save_path=model_output_prefix,
global_step=global_step)
else:
no_update_count[TRAIN] += 1
else:
if best_res[TRAIN] is None or res[
early_stop_metric] > best_res[TRAIN][early_stop_metric]:
best_res[TRAIN] = res
no_update_count[TRAIN] = 0
else:
no_update_count[TRAIN] += 1
# 计算在验证集上的表现, 不更新模型参数
print('VALID')
n_sample = index_iterator.n_sample(VALID)
labels_predict = list()
labels_gold = list()
for batch_index in index_iterator.iterate(batch_size,
mode=VALID,
shuffle=False):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
n_sample], labels_gold[:
n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
eval_history[DEV].append(res)
print_evaluation(res)
# Early Stop
if best_res[VALID] is None or res[
early_stop_metric] > best_res[VALID][early_stop_metric]:
saver.save(sess,
save_path=model_output_prefix,
global_step=global_step)
best_res[VALID] = res
no_update_count[VALID] = 0
else:
no_update_count[VALID] += 1
# eval test
_mode = TEST
_dataset = datasets[_mode]
_index_iterator = index_iterators[_mode]
_n_sample = _index_iterator.n_sample()
labels_predict = list()
labels_gold = list()
for batch_index in _index_iterator.iterate(batch_size,
shuffle=False):
feed_dict = {
nn.var(_key): _dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += _dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
_n_sample], labels_gold[:
_n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
eval_history[TEST].append(res)
if no_update_count[TRAIN] >= max_no_update_count:
break
# 训练结束 ##########################################################################
# 确保输出文件夹存在
print(
'========================= BEST ROUND EVALUATION ========================='
)
json.dump(eval_history,
open(data_config.output_path(output_key, 'eval', 'json'), 'w'))
with tf.Session() as sess:
prefix_checkpoint = tf.train.latest_checkpoint(
data_config.model_path(key=output_key))
saver = tf.train.import_meta_graph('{}.meta'.format(prefix_checkpoint))
saver.restore(sess, prefix_checkpoint)
nn = BaseNNModel(config=None)
nn.set_graph(tf.get_default_graph())
for mode in [TRAIN, TEST]:
dataset = datasets[mode]
index_iterator = index_iterators[mode]
n_sample = index_iterator.n_sample()
prob_predict = list()
labels_predict = list()
labels_gold = list()
hidden_feats = list()
for batch_index in index_iterator.iterate(batch_size,
shuffle=False):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
prob_predict += res[PROB_PREDICT].tolist()
labels_predict += res[LABEL_PREDICT].tolist()
hidden_feats += res[HIDDEN_FEAT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
prob_predict = prob_predict[:n_sample]
labels_predict = labels_predict[:n_sample]
labels_gold = labels_gold[:n_sample]
hidden_feats = hidden_feats[:n_sample]
if mode == TEST:
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
best_res[TEST] = res
# 导出隐藏层
with open(data_config.output_path(output_key, mode, HIDDEN_FEAT),
'w') as file_obj:
for _feat in hidden_feats:
file_obj.write('\t'.join(map(str, _feat)) + '\n')
# 导出预测的label
with open(data_config.output_path(output_key, mode, LABEL_PREDICT),
'w') as file_obj:
for _label in labels_predict:
file_obj.write('{}\n'.format(_label))
with open(data_config.output_path(output_key, mode, PROB_PREDICT),
'w') as file_obj:
for _prob in prob_predict:
file_obj.write('\t'.join(map(str, _prob)) + '\n')
for mode in [TRAIN, VALID, TEST]:
if mode == VALID and train_config.valid_rate == 0.:
continue
res = best_res[mode]
print(mode)
print_evaluation(res)
json.dump(
res,
open(data_config.output_path(output_key, mode, EVALUATION),
'w'))
print()
test_score_list = map(lambda _item: _item['f1'], eval_history[TEST])
print('best test f1 reached: {}'.format(max(test_score_list)))
print('OUTPUT_KEY: {}'.format(output_key))
def train(dataset_key,
text_version,
label_version=None,
config_path='config.yaml'):
"""
python -m algo.main93 train semeval2019_task3_dev -t ek
python3 -m algo.main93 train semeval2019_task3_dev -t ek -c config_ntua93.yaml
:param dataset_key: string
:param text_version: string
:param label_version: string
:param config_path: string
:return:
"""
config_data = yaml.load(open(config_path))
output_key = '{}_{}_{}'.format(config_data['module'].rsplit('.', 1)[1],
text_version, int(time.time()))
if label_version is not None:
output_key = '{}_{}'.format(label_version, output_key)
print('OUTPUT_KEY: {}'.format(output_key))
# 准备输出路径的文件夹
data_config.prepare_output_folder(output_key=output_key)
data_config.prepare_model_folder(output_key=output_key)
shutil.copy(config_path, data_config.output_path(output_key, ALL, CONFIG))
# 根据配置加载模块
module_relative_path = config_data['module']
NNModel = getattr(importlib.import_module(module_relative_path), 'NNModel')
NNConfig = getattr(importlib.import_module(module_relative_path),
'NNConfig')
if config_data['analyzer'] == WORD:
w2v_key = '{}_{}'.format(config_data['word']['w2v_version'],
text_version)
w2v_model_path = data_config.path(ALL, WORD2VEC, w2v_key)
vocab_train_path = data_config.path(TRAIN, VOCAB, text_version)
# 加载字典集
# 在模型中会采用所有模型中支持的词向量, 并为有足够出现次数的单词随机生成词向量
vocab_meta_list = load_vocab_list(vocab_train_path)
# vocab_meta_list += load_vocab_list(semeval2018_task3_date_config.path(TRAIN, VOCAB, text_version))
vocabs = [
_meta['t'] for _meta in vocab_meta_list
if _meta['tf'] >= config_data[WORD]['min_tf']
]
# 加载词向量与相关数据
lookup_table, vocab_id_mapping, embedding_dim = load_lookup_table(
w2v_model_path=w2v_model_path, vocabs=vocabs)
json.dump(
vocab_id_mapping,
open(data_config.output_path(output_key, ALL, VOCAB_ID_MAPPING),
'w'))
max_seq_len = MAX_WORD_SEQ_LEN
elif config_data['analyzer'] == CHAR:
texts = load_text_list(data_config.path(TRAIN, TEXT))
char_set = set()
for text in texts:
char_set |= set(text)
lookup_table, vocab_id_mapping, embedding_dim = build_random_lookup_table(
vocabs=char_set, dim=config_data['char']['embedding_dim'])
max_seq_len = MAX_CHAR_SEQ_LEN
else:
raise ValueError('invalid analyzer: {}'.format(
config_data['analyzer']))
# 加载训练数据
datasets, output_dim = load_dataset(data_config=data_config,
analyzer=config_data['analyzer'],
vocab_id_mapping=vocab_id_mapping,
seq_len=max_seq_len,
with_label=True,
label_version=label_version,
text_version=text_version)
# 加载配置
nn_config = NNConfig(config_data)
train_config = TrainConfig(config_data['train'])
# 初始化数据集的检索
index_iterators = {
mode: IndexIterator(datasets[mode][LABEL_GOLD])
for mode in [TRAIN, TEST]
}
# 按配置将训练数据切割成训练集和验证集
index_iterators[TRAIN].split_train_valid(train_config.valid_rate)
# 计算各个类的权重
if train_config.use_class_weights:
label_weight = {
# 参考 sklearn 中 class_weight='balanced'的公式, 实验显示效果显着
_label: float(index_iterators[TRAIN].n_sample()) /
(index_iterators[TRAIN].dim * len(_index))
for _label, _index in index_iterators[TRAIN].label_index.items()
}
else:
label_weight = {
_label: 1.
for _label in range(index_iterators[TRAIN].dim)
}
# 基于加载的数据更新配置
nn_config.set_embedding_dim(embedding_dim)
nn_config.set_output_dim(output_dim)
nn_config.set_seq_len(max_seq_len)
# 搭建神经网络
nn = NNModel(config=nn_config)
nn.build_neural_network(lookup_table=lookup_table)
batch_size = train_config.batch_size
fetches = {
mode: {_key: nn.var(_key)
for _key in fetch_key[mode]}
for mode in [TRAIN, TEST]
}
last_eval = {TRAIN: None, VALID: None, TEST: None}
model_output_prefix = data_config.model_path(key=output_key) + '/model'
best_res = {mode: None for mode in [TRAIN, VALID]}
no_update_count = {mode: 0 for mode in [TRAIN, VALID]}
max_no_update_count = 10
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver(tf.global_variables())
dataset = datasets[TRAIN]
index_iterator = index_iterators[TRAIN]
# 训练开始 ##########################################################################
for epoch in range(train_config.epoch):
print('== epoch {} =='.format(epoch))
# 利用训练集进行训练
print('TRAIN')
n_sample = index_iterator.n_sample(TRAIN)
labels_predict = list()
labels_gold = list()
for batch_index in index_iterator.iterate(batch_size,
mode=TRAIN,
shuffle=True):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TRAIN]
}
feed_dict[nn.var(SAMPLE_WEIGHTS)] = list(
map(label_weight.get, feed_dict[nn.var(LABEL_GOLD)]))
feed_dict[nn.var(TEST_MODE)] = 0
res = sess.run(fetches=fetches[TRAIN], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
n_sample], labels_gold[:
n_sample]
labels_predict, labels_gold = labels_predict[:
n_sample], labels_gold[:
n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
last_eval[TRAIN] = res
print_evaluation(res)
global_step = tf.train.global_step(sess, nn.var(GLOBAL_STEP))
if train_config.valid_rate == 0.:
if best_res[TRAIN] is None or res[F1_SCORE] > best_res[TRAIN][
F1_SCORE]:
best_res[TRAIN] = res
no_update_count[TRAIN] = 0
saver.save(sess,
save_path=model_output_prefix,
global_step=global_step)
else:
no_update_count[TRAIN] += 1
else:
if best_res[TRAIN] is None or res[F1_SCORE] > best_res[TRAIN][
F1_SCORE]:
best_res[TRAIN] = res
no_update_count[TRAIN] = 0
else:
no_update_count[TRAIN] += 1
# 计算在验证集上的表现, 不更新模型参数
print('VALID')
n_sample = index_iterator.n_sample(VALID)
labels_predict = list()
labels_gold = list()
for batch_index in index_iterator.iterate(batch_size,
mode=VALID,
shuffle=False):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
n_sample], labels_gold[:
n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
last_eval[VALID] = res
print_evaluation(res)
# Early Stop
if best_res[VALID] is None or res[F1_SCORE] > best_res[VALID][
F1_SCORE]:
saver.save(sess,
save_path=model_output_prefix,
global_step=global_step)
best_res[VALID] = res
no_update_count[VALID] = 0
else:
no_update_count[VALID] += 1
if no_update_count[TRAIN] >= max_no_update_count:
break
# 训练结束 ##########################################################################
# 确保输出文件夹存在
print(
'========================= BEST ROUND EVALUATION ========================='
)
with tf.Session() as sess:
prefix_checkpoint = tf.train.latest_checkpoint(
data_config.model_path(key=output_key))
saver = tf.train.import_meta_graph('{}.meta'.format(prefix_checkpoint))
saver.restore(sess, prefix_checkpoint)
nn = BaseNNModel(config=None)
nn.set_graph(tf.get_default_graph())
for mode in [TRAIN, TEST]:
dataset = datasets[mode]
index_iterator = index_iterators[mode]
n_sample = index_iterator.n_sample()
prob_predict = list()
labels_predict = list()
labels_gold = list()
hidden_feats = list()
for batch_index in index_iterator.iterate(batch_size,
shuffle=False):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
prob_predict += res[PROB_PREDICT].tolist()
labels_predict += res[LABEL_PREDICT].tolist()
hidden_feats += res[HIDDEN_FEAT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
prob_predict = prob_predict[:n_sample]
labels_predict = labels_predict[:n_sample]
labels_gold = labels_gold[:n_sample]
hidden_feats = hidden_feats[:n_sample]
if mode == TEST:
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
best_res[TEST] = res
# 导出隐藏层
with open(data_config.output_path(output_key, mode, HIDDEN_FEAT),
'w') as file_obj:
for _feat in hidden_feats:
file_obj.write('\t'.join(map(str, _feat)) + '\n')
# 导出预测的label
with open(data_config.output_path(output_key, mode, LABEL_PREDICT),
'w') as file_obj:
for _label in labels_predict:
file_obj.write('{}\n'.format(_label))
with open(data_config.output_path(output_key, mode, PROB_PREDICT),
'w') as file_obj:
for _prob in prob_predict:
file_obj.write('\t'.join(map(str, _prob)) + '\n')
for mode in [TRAIN, VALID, TEST]:
if mode == VALID and train_config.valid_rate == 0.:
continue
res = best_res[mode]
print(mode)
print_evaluation(res)
json.dump(
res,
open(data_config.output_path(output_key, mode, EVALUATION),
'w'))
print()
print('OUTPUT_KEY: {}'.format(output_key))
def train(model_name,
label_version=None,
label_key=None,
config_path='c93f.yaml',
check=False):
"""
python -m algo.main93_v2 train
python3 -m algo.main93_v2 train -c config_ntua93.yaml
:param model_name: string
:param label_version: string
:param config_path: string
:return:
"""
text_version = 'ek'
config_data = yaml.load(open(config_path))
NNModel = getattr(
importlib.import_module('algo.m93.{}'.format(model_name)), 'NNModel')
output_key = '{}_{}_{}'.format(model_name, text_version, int(time.time()))
if label_key is not None:
output_key = '{}_{}'.format(label_key, output_key)
output_key = 'f_{}'.format(output_key)
print('OUTPUT_KEY: {}'.format(output_key))
# 准备输出路径的文件夹
data_config.prepare_output_folder(output_key=output_key)
data_config.prepare_model_folder(output_key=output_key)
shutil.copy(config_path, data_config.output_path(output_key, ALL, CONFIG))
w2v_key = '{}_{}'.format(config_data['word']['w2v_version'], text_version)
w2v_model_path = data_config.path(ALL, WORD2VEC, w2v_key)
vocab_train_path = data_config.path(TRAIN, VOCAB, text_version)
# 加载字典集
# 在模型中会采用所有模型中支持的词向量, 并为有足够出现次数的单词随机生成词向量
vocab_meta_list = load_vocab_list(vocab_train_path)
vocabs = [
_meta['t'] for _meta in vocab_meta_list
if _meta['tf'] >= config_data['word']['min_tf']
]
# 加载词向量与相关数据
lookup_table, vocab_id_mapping, embedding_dim = load_lookup_table2(
w2v_model_path=w2v_model_path, vocabs=vocabs)
json.dump(
vocab_id_mapping,
open(data_config.output_path(output_key, ALL, VOCAB_ID_MAPPING), 'w'))
# 加载配置
nn_config = NNConfig(config_data)
train_config = TrainConfig(config_data['train'])
early_stop_metric = train_config.early_stop_metric
# 加载训练数据
datasets = dict()
datasets[TRAIN], output_dim = load_dataset(
mode=[TRAIN, TEST],
vocab_id_mapping=vocab_id_mapping,
max_seq_len=nn_config.seq_len,
label_version=label_version,
sampling=train_config.train_sampling,
label_map=train_config.label_map(label_key))
datasets[TEST], _ = load_dataset(
mode=FINAL,
vocab_id_mapping=vocab_id_mapping,
max_seq_len=nn_config.seq_len,
label_version=label_version,
sampling=False,
label_map=train_config.label_map(label_key))
# 初始化数据集的检索
index_iterators = {
TRAIN: IndexIterator.from_dataset(datasets[TRAIN]),
TEST: IndexIterator.from_dataset(datasets[TEST])
}
# 按配置将训练数据切割成训练集和验证集
index_iterators[TRAIN].split_train_valid(train_config.valid_rate)
# 计算各个类的权重
if train_config.use_class_weights:
label_weight = {
# 参考 sklearn 中 class_weight='balanced'的公式, 实验显示效果显着
_label: float(index_iterators[TRAIN].n_sample()) /
(index_iterators[TRAIN].dim * len(_index))
for _label, _index in index_iterators[TRAIN].label_index.items()
}
else:
label_weight = {
_label: 1.
for _label in range(index_iterators[TRAIN].dim)
}
# 基于加载的数据更新配置
nn_config.set_embedding_dim(embedding_dim)
nn_config.set_output_dim(output_dim)
# 搭建神经网络
nn = NNModel(config=nn_config)
nn.build_neural_network(lookup_table=lookup_table)
batch_size = train_config.batch_size
fetches = {
mode: {_key: nn.var(_key)
for _key in fetch_key[mode]}
for mode in [TRAIN, TEST]
}
model_output_prefix = data_config.model_path(key=output_key) + '/model'
best_res = {mode: None for mode in [TRAIN, VALID, TEST]}
no_update_count = {mode: 0 for mode in [TRAIN, VALID]}
max_no_update_count = 10
eval_history = {TRAIN: list(), VALID: list(), TEST: list()}
best_epoch = -1
best_epoch_test = -1
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver(tf.global_variables())
dataset = datasets[TRAIN]
index_iterator = index_iterators[TRAIN]
# 训练开始 ##########################################################################
for epoch in range(train_config.epoch):
print('== epoch {} = {} ='.format(epoch, output_key))
# 利用训练集进行训练
print('TRAIN')
n_sample = index_iterator.n_sample(TRAIN)
labels_predict = list()
labels_gold = list()
for batch_index in index_iterator.iterate(batch_size,
mode=TRAIN,
shuffle=True):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TRAIN]
}
feed_dict[nn.var(SAMPLE_WEIGHTS)] = list(
map(label_weight.get, feed_dict[nn.var(LABEL_GOLD)]))
feed_dict[nn.var(TEST_MODE)] = 0
res = sess.run(fetches=fetches[TRAIN], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
n_sample], labels_gold[:
n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
print_evaluation(res)
eval_history[TRAIN].append(res)
global_step = tf.train.global_step(sess, nn.var(GLOBAL_STEP))
if train_config.valid_rate == 0.:
if best_epoch <= 10 or (best_res[TRAIN] is None
or res[early_stop_metric] >
best_res[TRAIN][early_stop_metric]):
best_epoch = epoch
best_res[TRAIN] = res
no_update_count[TRAIN] = 0
saver.save(sess,
save_path=model_output_prefix,
global_step=global_step)
else:
no_update_count[TRAIN] += 1
else:
if best_res[TRAIN] is None or res[
early_stop_metric] > best_res[TRAIN][early_stop_metric]:
best_res[TRAIN] = res
no_update_count[TRAIN] = 0
else:
no_update_count[TRAIN] += 1
# 计算在验证集上的表现, 不更新模型参数
print('VALID')
n_sample = index_iterator.n_sample(VALID)
labels_predict = list()
labels_gold = list()
for batch_index in index_iterator.iterate(batch_size,
mode=VALID,
shuffle=False):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
n_sample], labels_gold[:
n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
eval_history[VALID].append(res)
print_evaluation(res)
# Early Stop
if best_epoch <= 10 or (best_res[VALID] is None
or res[early_stop_metric] >
best_res[VALID][early_stop_metric]):
best_epoch = epoch
saver.save(sess,
save_path=model_output_prefix,
global_step=global_step)
best_res[VALID] = res
no_update_count[VALID] = 0
else:
no_update_count[VALID] += 1
# eval test
_mode = TEST
_dataset = datasets[_mode]
_index_iterator = SimpleIndexIterator.from_dataset(_dataset)
_n_sample = _index_iterator.n_sample()
labels_predict = list()
labels_gold = list()
for batch_index in _index_iterator.iterate(batch_size,
shuffle=False):
feed_dict = {
nn.var(_key): _dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
labels_predict += res[LABEL_PREDICT].tolist()
labels_gold += _dataset[LABEL_GOLD][batch_index].tolist()
labels_predict, labels_gold = labels_predict[:
_n_sample], labels_gold[:
_n_sample]
res = basic_evaluate(gold=labels_gold, pred=labels_predict)
eval_history[TEST].append(res)
print('TEST')
print_evaluation(res)
if best_res[TEST] is None or res[F1_SCORE] > best_res[TEST][
F1_SCORE]:
best_res[TEST] = res
best_epoch_test = epoch
if no_update_count[TRAIN] >= max_no_update_count:
break
# 训练结束 ##########################################################################
# 确保输出文件夹存在
print(
'========================= BEST ROUND EVALUATION ========================='
)
json.dump(eval_history,
open(data_config.output_path(output_key, 'eval', 'json'), 'w'))
labels_predict_final = None
labels_gold_final = load_label_list(data_config.path(FINAL, LABEL))
with tf.Session() as sess:
prefix_checkpoint = tf.train.latest_checkpoint(
data_config.model_path(key=output_key))
saver = tf.train.import_meta_graph('{}.meta'.format(prefix_checkpoint))
saver.restore(sess, prefix_checkpoint)
nn = BaseNNModel(config=None)
nn.set_graph(tf.get_default_graph())
for mode in [TRAIN, TEST, FINAL]:
dataset = load_dataset(mode=mode,
vocab_id_mapping=vocab_id_mapping,
max_seq_len=nn_config.seq_len,
with_label=False)
index_iterator = SimpleIndexIterator.from_dataset(dataset)
n_sample = index_iterator.n_sample()
prob_predict = list()
labels_predict = list()
for batch_index in index_iterator.iterate(batch_size,
shuffle=False):
feed_dict = {
nn.var(_key): dataset[_key][batch_index]
for _key in feed_key[TEST]
}
feed_dict[nn.var(TEST_MODE)] = 1
res = sess.run(fetches=fetches[TEST], feed_dict=feed_dict)
prob_predict += res[PROB_PREDICT].tolist()
labels_predict += res[LABEL_PREDICT].tolist()
prob_predict = prob_predict[:n_sample]
labels_predict = labels_predict[:n_sample]
if mode == FINAL:
labels_predict_final = labels_predict
# 导出预测的label
with open(data_config.output_path(output_key, mode, LABEL_PREDICT),
'w') as file_obj:
for _label in labels_predict:
file_obj.write('{}\n'.format(_label))
with open(data_config.output_path(output_key, mode, PROB_PREDICT),
'w') as file_obj:
for _prob in prob_predict:
file_obj.write('\t'.join(map(str, _prob)) + '\n')
print('====== best epoch test: {} ======'.format(best_epoch_test))
for mode in [TRAIN, VALID, TEST]:
if mode == VALID and train_config.valid_rate == 0.:
continue
print(mode)
res = eval_history[mode][best_epoch_test]
print_evaluation(res)
if mode == TEST:
for col in res[CONFUSION_MATRIX]:
print(','.join(map(str, col)))
print(eval_history[TEST][best_epoch_test])
print()
print('====== best epoch valid: {} ======'.format(best_epoch))
for mode in [TRAIN, VALID, TEST]:
if mode == VALID and train_config.valid_rate == 0.:
continue
print(mode)
res = eval_history[mode][best_epoch]
print_evaluation(res)
if mode == TEST:
for col in res[CONFUSION_MATRIX]:
print(','.join(map(str, col)))
json.dump(
res,
open(data_config.output_path(output_key, mode, EVALUATION), 'w'))
print()
print(eval_history[TEST][best_epoch])
print()
if check:
print('====== label_map check ======')
label_map = train_config.label_map(label_key)
if label_map is not None:
new_gold = list()
new_pred = list()
for g, p in zip(labels_gold_final, labels_predict_final):
if g in label_map:
new_gold.append(label_map[g])
new_pred.append(p)
labels_gold_final = new_gold
labels_predict_final = new_pred
res = basic_evaluate(gold=labels_gold_final, pred=labels_predict_final)
print_evaluation(res)
for col in res[CONFUSION_MATRIX]:
print(','.join(map(str, col)))
print('OUTPUT_KEY: {}'.format(output_key))
