def init_model():
train_data, dev_data, test_data, vocab = readdata()
model = CNN((len(vocab), 128),
num_classes=target_len,
padding=2,
dropout=0.1)
#model = torch.load("rnnmodel/best_RNN_accuracy_2019-05-22-17-18-46")
trainer = Trainer(model=model,
train_data=train_data,
dev_data=dev_data,
device=0,
save_path='cnnmodel',
loss=loss,
metrics=metrics,
callbacks=[FitlogCallback(test_data)])
tester = Tester(test_data, model, metrics=AccuracyMetric())
print(2)
model2 = RNN(embed_num=len(vocab),
input_size=256,
hidden_size=256,
target_size=target_len)
#model2 = torch.load("rnnmodel/best_RNN_accuracy_2019-05-22-17-18-46")
trainer2 = Trainer(model=model2,
train_data=train_data,
dev_data=dev_data,
loss=loss,
metrics=metrics,
save_path='rnnmodel',
batch_size=32,
n_epochs=20,
device=0)
tester2 = Tester(test_data, model, metrics=AccuracyMetric())
return trainer, trainer2, tester, tester2
def test_CheckPointCallback(self):
from fastNLP import CheckPointCallback, Callback
from fastNLP import Tester
class RaiseCallback(Callback):
def __init__(self, stop_step=10):
super().__init__()
self.stop_step = stop_step
def on_backward_begin(self, loss):
if self.step > self.stop_step:
raise RuntimeError()
data_set, model = prepare_env()
tester = Tester(data=data_set, model=model, metrics=AccuracyMetric(pred="predict", target="y"))
import fitlog
fitlog.set_log_dir(self.tempdir, new_log=True)
tempfile_path = os.path.join(self.tempdir, 'chkt.pt')
callbacks = [CheckPointCallback(tempfile_path)]
fitlog_callback = FitlogCallback(data_set, tester)
callbacks.append(fitlog_callback)
callbacks.append(RaiseCallback(100))
try:
trainer = Trainer(data_set, model, optimizer=SGD(lr=0.1), loss=BCELoss(pred="predict", target="y"),
batch_size=32, n_epochs=5, print_every=50, dev_data=data_set,
metrics=AccuracyMetric(pred="predict", target="y"), use_tqdm=True,
callbacks=callbacks, check_code_level=2)
trainer.train()
except:
pass
# 用下面的代码模拟重新运行
data_set, model = prepare_env()
callbacks = [CheckPointCallback(tempfile_path)]
tester = Tester(data=data_set, model=model, metrics=AccuracyMetric(pred="predict", target="y"))
fitlog_callback = FitlogCallback(data_set, tester)
callbacks.append(fitlog_callback)
trainer = Trainer(data_set, model, optimizer=SGD(lr=0.1), loss=BCELoss(pred="predict", target="y"),
batch_size=32, n_epochs=5, print_every=50, dev_data=data_set,
metrics=AccuracyMetric(pred="predict", target="y"), use_tqdm=True,
callbacks=callbacks, check_code_level=2)
trainer.train()
def test_fitlog_callback(self):
import fitlog
fitlog.set_log_dir(self.tempdir, new_log=True)
data_set, model = prepare_env()
from fastNLP import Tester
tester = Tester(data=data_set, model=model, metrics=AccuracyMetric(pred="predict", target="y"))
fitlog_callback = FitlogCallback(data_set, tester)
trainer = Trainer(data_set, model, optimizer=SGD(lr=0.1), loss=BCELoss(pred="predict", target="y"),
batch_size=32, n_epochs=5, print_every=50, dev_data=data_set,
metrics=AccuracyMetric(pred="predict", target="y"), use_tqdm=True,
callbacks=fitlog_callback, check_code_level=2)
trainer.train()
data, char_embed, word_embed = cache()
print(data)
model = CNNBiLSTMCRF(word_embed,
char_embed,
hidden_size=1200,
num_layers=1,
tag_vocab=data.vocabs[Const.TARGET],
encoding_type=encoding_type,
dropout=dropout)
callbacks = [
GradientClipCallback(clip_value=5, clip_type='value'),
FitlogCallback(data.datasets['test'], verbose=1)
]
optimizer = SGD(model.parameters(), lr=lr, momentum=0.9)
scheduler = LRScheduler(
LambdaLR(optimizer, lr_lambda=lambda epoch: 1 / (1 + 0.05 * epoch)))
callbacks.append(scheduler)
trainer = Trainer(train_data=data.datasets['dev'][:100],
model=model,
optimizer=optimizer,
sampler=None,
device=0,
dev_data=data.datasets['dev'][:100],
batch_size=batch_size,
metrics=SpanFPreRecMetric(
dataset.set_input(Const.INPUT, Const.INPUT_LEN)
dataset.set_target(Const.TARGET)
testset.rename_field('words', Const.INPUT)
testset.rename_field('target', Const.TARGET)
testset.rename_field('seq_len', Const.INPUT_LEN)
testset.set_input(Const.INPUT, Const.INPUT_LEN)
testset.set_target(Const.TARGET)
train_data, dev_data = dataset.split(0.1)
loss = CrossEntropyLoss(pred=Const.OUTPUT, target=Const.TARGET)
metrics = AccuracyMetric(pred=Const.OUTPUT, target=Const.TARGET)
trainer = Trainer(model=model,
train_data=train_data,
dev_data=dev_data,
loss=loss,
batch_size=16,
metrics=metrics,
n_epochs=20,
callbacks=[FitlogCallback(dataset)])
trainer.train()
tester = Tester(data=testset, model=model, metrics=metrics)
tester.test()
tester = Tester(data=train_data, model=model, metrics=metrics)
tester.test()
fitlog.finish()
def train(config, task_name):
train_data = pickle.load(
open(os.path.join(config.bert_data_path, config.train_name), "rb"))
print(train_data[0])
# debug
if config.debug:
train_data = train_data[0:30]
dev_data = pickle.load(
open(os.path.join(config.bert_data_path, config.dev_name), "rb"))
print(dev_data[0])
# test_data = pickle.load(open(os.path.join(config.bert_data_path, config.test_name), "rb"))
schemas = get_schemas(config.source_path)
state_dict = torch.load(config.bert_path)
# print(state_dict)
text_model = BertForMultiLabelSequenceClassification.from_pretrained(
config.bert_folder, state_dict=state_dict, num_labels=len(schemas))
# optimizer
param_optimizer = list(text_model.named_parameters())
no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
optimizer_grouped_parameters = [{
'params':
[p for n, p in param_optimizer if not any(nd in n for nd in no_decay)],
'weight_decay':
0.01
}, {
'params':
[p for n, p in param_optimizer if any(nd in n for nd in no_decay)],
'weight_decay':
0.0
}]
num_train_optimization_steps = int(
len(train_data) / config.batch_size /
config.update_every) * config.epoch
if config.local_rank != -1:
num_train_optimization_steps = num_train_optimization_steps // torch.distributed.get_world_size(
)
optimizer = BertAdam(
lr=config.lr,
warmup=config.warmup_proportion,
t_total=num_train_optimization_steps).construct_from_pytorch(
optimizer_grouped_parameters)
timing = TimingCallback()
early_stop = EarlyStopCallback(config.patience)
logs = FitlogCallback(dev_data)
f1 = F1_score(pred='output', target='label_id')
trainer = Trainer(train_data=train_data,
model=text_model,
loss=BCEWithLogitsLoss(),
batch_size=config.batch_size,
check_code_level=-1,
metrics=f1,
metric_key='f1',
n_epochs=int(config.epoch),
dev_data=dev_data,
save_path=config.save_path,
print_every=config.print_every,
validate_every=config.validate_every,
update_every=config.update_every,
optimizer=optimizer,
use_tqdm=False,
device=config.device,
callbacks=[timing, early_stop, logs])
trainer.train()
# test result
tester = Tester(
dev_data,
text_model,
metrics=f1,
device=config.device,
batch_size=config.batch_size,
)
tester.test()
target='target',
seq_len='seq_len',
encoding_type=encoding_type)
acc_metric = AccuracyMetric(pred='pred', target='target', seq_len='seq_len')
metrics = [f1_metric, acc_metric]
if args.optim == 'adam':
optimizer = optim.Adam(model.parameters(), lr=args.lr)
elif args.optim == 'sgd':
optimizer = optim.SGD(model.parameters(),
lr=args.lr,
momentum=args.momentum)
callbacks = [
FitlogCallback({
'test': datasets['test'],
'train': datasets['train']
}),
LRScheduler(
lr_scheduler=LambdaLR(optimizer, lambda ep: 1 / (1 + 0.03)**ep))
]
print('label_vocab:{}\n{}'.format(len(vocabs['label']),
vocabs['label'].idx2word))
trainer = Trainer(datasets['train'],
model,
optimizer=optimizer,
loss=loss,
metrics=metrics,
dev_data=datasets['dev'],
device=device,
batch_size=args.batch,
n_epochs=args.epoch,
pred = pred.argmax(dim=-1).tolist()
self.p.extend(pred)
def get_metric(self, reset=True):
f1_score = m.f1_score(self.l, self.p, average="macro")
if reset:
self.l = []
self.p = []
return {"f1:": f1_score}
optimizer = Adam(lr=args.lr, weight_decay=0)
acc = AccuracyMetric()
f1 = f1metric()
loss = CrossEntropyLoss()
trainer = Trainer(
trainset,
cla,
optimizer=optimizer,
loss=loss,
batch_size=args.batch_size,
n_epochs=args.num_epoch,
dev_data=testset,
metrics=[acc, f1],
save_path=args.save_dir,
callbacks=[FitlogCallback(log_loss_every=5)],
)
trainer.train()
fitlog.finish() # finish the logging
rnn_text_model = RNN.RNN_Text(vocab_size=m, input_size=50, hidden_layer_size=128, target_size=k, dropout=0.1)
cnn_text_model = CNN.CNN_Text(vocab_size=m, input_size=50, target_size=k, dropout=0.05)
model = rnn_text_model
# ModelLoader.load_pytorch(model, "model_ckpt_large_CNN.pkl")
trainer = Trainer(
train_data=train_set,
model=model,
loss=CrossEntropyLoss(pred='pred', target='label'),
n_epochs=50,
batch_size=16,
metrics=AccuracyMetric(pred='pred', target='label'),
dev_data=dev_set,
optimizer=Adam(lr=1e-3),
callbacks=[FitlogCallback(data=test_set)]
)
trainer.train()
# saver = ModelSaver("model_ckpt_large_RNN.pkl")
# saver.save_pytorch(model)
tester = Tester(
data=train_set,
model=model,
metrics=AccuracyMetric(pred='pred', target='label'),
batch_size=16,
)
tester.test()
tester = Tester(
# elmo_embed = ElmoEmbedding(vocab=data.vocabs['cap_words'],
# model_dir_or_name='.',
# requires_grad=True, layers='mix')
# char_embed = StackEmbedding([elmo_embed, char_embed])
model = CNNBiLSTMCRF(word_embed,
char_embed,
hidden_size=200,
num_layers=1,
tag_vocab=data.vocabs[Const.TARGET],
encoding_type=encoding_type)
callbacks = [
GradientClipCallback(clip_type='value', clip_value=5),
FitlogCallback({'test': data.datasets['test']}, verbose=1),
# SaveModelCallback('save_models/', top=3, only_param=False, save_on_exception=True)
]
# optimizer = Adam(model.parameters(), lr=0.001)
# optimizer = SWATS(model.parameters(), verbose=True)
optimizer = SGD(model.parameters(), lr=0.01, momentum=0.9)
scheduler = LRScheduler(
LambdaLR(optimizer, lr_lambda=lambda epoch: 1 / (1 + 0.05 * epoch)))
callbacks.append(scheduler)
trainer = Trainer(train_data=data.datasets['train'],
model=model,
optimizer=optimizer,
sampler=BucketSampler(batch_size=20),
device=1,
dev_data=data.datasets['dev'],
def train(args):
text_data = TextData()
with open(os.path.join(args.vocab_dir, args.vocab_data), 'rb') as fin:
text_data = pickle.load(fin)
vocab_size = text_data.vocab_size
class_num = text_data.class_num
# class_num = 1
seq_len = text_data.max_seq_len
print("(vocab_size,class_num,seq_len):({0},{1},{2})".format(
vocab_size, class_num, seq_len))
train_data = text_data.train_set
val_data = text_data.val_set
test_data = text_data.test_set
train_data.set_input('words', 'seq_len')
train_data.set_target('target')
val_data.set_input('words', 'seq_len')
val_data.set_target('target')
test_data.set_input('words', 'seq_len')
test_data.set_target('target')
init_embeds = None
if args.pretrain_model == "None":
print("No pretrained model with be used.")
print("vocabsize:{0}".format(vocab_size))
init_embeds = (vocab_size, args.embed_size)
elif args.pretrain_model == "word2vec":
embeds_path = os.path.join(args.prepare_dir, 'w2v_embeds.pkl')
print("Loading Word2Vec pretrained embedding from {0}.".format(
embeds_path))
with open(embeds_path, 'rb') as fin:
init_embeds = pickle.load(fin)
elif args.pretrain_model == 'glove':
embeds_path = os.path.join(args.prepare_dir, 'glove_embeds.pkl')
print(
"Loading Glove pretrained embedding from {0}.".format(embeds_path))
with open(embeds_path, 'rb') as fin:
init_embeds = pickle.load(fin)
elif args.pretrain_model == 'glove2wv':
embeds_path = os.path.join(args.prepare_dir, 'glove2wv_embeds.pkl')
print(
"Loading Glove pretrained embedding from {0}.".format(embeds_path))
with open(embeds_path, 'rb') as fin:
init_embeds = pickle.load(fin)
else:
init_embeds = (vocab_size, args.embed_size)
if args.model == "CNNText":
print("Using CNN Model.")
model = CNNText(init_embeds,
num_classes=class_num,
padding=2,
dropout=args.dropout)
elif args.model == "StarTransformer":
print("Using StarTransformer Model.")
model = STSeqCls(init_embeds,
num_cls=class_num,
hidden_size=args.hidden_size)
elif args.model == "MyCNNText":
model = MyCNNText(init_embeds=init_embeds,
num_classes=class_num,
padding=2,
dropout=args.dropout)
print("Using user defined CNNText")
elif args.model == "LSTMText":
print("Using LSTM Model.")
model = LSTMText(init_embeds=init_embeds,
output_dim=class_num,
hidden_dim=args.hidden_size,
num_layers=args.num_layers,
dropout=args.dropout)
elif args.model == "Bert":
print("Using Bert Model.")
else:
print("Using default model: CNNText.")
model = CNNText((vocab_size, args.embed_size),
num_classes=class_num,
padding=2,
dropout=0.1)
print(model)
if args.cuda:
device = torch.device('cuda')
else:
device = None
print("train_size:{0} ; val_size:{1} ; test_size:{2}".format(
train_data.get_length(), val_data.get_length(),
test_data.get_length()))
if args.optim == "Adam":
print("Using Adam as optimizer.")
optimizer = fastnlp_optim.Adam(lr=0.001,
weight_decay=args.weight_decay)
if (args.model_suffix == "default"):
args.model_suffix == args.optim
else:
print("No Optimizer will be used.")
optimizer = None
criterion = CrossEntropyLoss()
metric = AccuracyMetric()
model_save_path = os.path.join(args.model_dir, args.model,
args.model_suffix)
earlystop = EarlyStopCallback(args.patience)
fitlog_back = FitlogCallback({"val": val_data, "train": train_data})
trainer = Trainer(train_data=train_data,
model=model,
save_path=model_save_path,
device=device,
n_epochs=args.epochs,
optimizer=optimizer,
dev_data=val_data,
loss=criterion,
batch_size=args.batch_size,
metrics=metric,
callbacks=[fitlog_back, earlystop])
trainer.train()
print("Train Done.")
tester = Tester(data=val_data,
model=model,
metrics=metric,
batch_size=args.batch_size,
device=device)
tester.test()
print("Test Done.")
print("Predict the answer with best model...")
acc = 0.0
output = []
data_iterator = Batch(test_data, batch_size=args.batch_size)
for data_x, batch_y in data_iterator:
i_data = Variable(data_x['words']).cuda()
pred = model(i_data)[C.OUTPUT]
pred = pred.sigmoid()
# print(pred.shape)
output.append(pred.cpu().data)
output = torch.cat(output, 0).numpy()
print(output.shape)
print("Predict Done. {} records".format(len(output)))
result_save_path = os.path.join(args.result_dir,
args.model + "_" + args.model_suffix)
with open(result_save_path + ".pkl", 'wb') as f:
pickle.dump(output, f)
output = output.squeeze()[:, 1].tolist()
projectid = text_data.test_projectid.values
answers = []
count = 0
for i in range(len(output)):
if output[i] > 0.5:
count += 1
print("true sample count:{}".format(count))
add_count = 0
for i in range(len(projectid) - len(output)):
output.append([0.13])
add_count += 1
print("Add {} default result in predict.".format(add_count))
df = pd.DataFrame()
df['projectid'] = projectid
df['y'] = output
df.to_csv(result_save_path + ".csv", index=False)
print("Predict Done, results saved to {}".format(result_save_path))
fitlog.finish()
def train(config, task_name):
train_data = pickle.load(
open(os.path.join(config.data_path, config.train_name), "rb"))
# debug
if config.debug:
train_data = train_data[0:100]
dev_data = pickle.load(
open(os.path.join(config.data_path, config.dev_name), "rb"))
print(len(train_data), len(dev_data))
# test_data = pickle.load(open(os.path.join(config.data_path, config.test_name), "rb"))
# load w2v data
# weight = pickle.load(open(os.path.join(config.data_path, config.weight_name), "rb"))
word_vocab = pickle.load(
open(os.path.join(config.data_path, config.word_vocab_name), "rb"))
char_vocab = pickle.load(
open(os.path.join(config.data_path, config.char_vocab_name), "rb"))
pos_vocab = pickle.load(
open(os.path.join(config.data_path, config.pos_vocab_name), "rb"))
# spo_vocab = pickle.load(open(os.path.join(config.data_path, config.spo_vocab_name), "rb"))
tag_vocab = pickle.load(
open(os.path.join(config.data_path, config.tag_vocab_name), "rb"))
print('word vocab', len(word_vocab))
print('char vocab', len(char_vocab))
print('pos vocab', len(pos_vocab))
# print('spo vocab', len(spo_vocab))
print('tag vocab', len(tag_vocab))
schema = get_schemas(config.source_path)
if task_name == 'bilstm_crf':
model = AdvSeqLabel(
char_init_embed=(len(char_vocab), config.char_embed_dim),
word_init_embed=(len(word_vocab), config.word_embed_dim),
pos_init_embed=(len(pos_vocab), config.pos_embed_dim),
spo_embed_dim=len(schema),
sentence_length=config.sentence_length,
hidden_size=config.hidden_dim,
num_classes=len(tag_vocab),
dropout=config.dropout,
id2words=tag_vocab.idx2word,
encoding_type=config.encoding_type)
elif task_name == 'trans_crf':
model = TransformerSeqLabel(
char_init_embed=(len(char_vocab), config.char_embed_dim),
word_init_embed=(len(word_vocab), config.word_embed_dim),
pos_init_embed=(len(pos_vocab), config.pos_embed_dim),
spo_embed_dim=len(schema),
num_classes=len(tag_vocab),
id2words=tag_vocab.idx2word,
encoding_type=config.encoding_type,
num_layers=config.num_layers,
inner_size=config.inner_size,
key_size=config.key_size,
value_size=config.value_size,
num_head=config.num_head,
dropout=config.dropout)
optimizer = Adam(lr=config.lr, weight_decay=config.weight_decay)
timing = TimingCallback()
early_stop = EarlyStopCallback(config.patience)
# loss = NLLLoss()
logs = FitlogCallback(dev_data)
metrics = SpanFPreRecMetric(tag_vocab,
pred='pred',
seq_len='seq_len',
target='tag')
train_data.set_input('tag')
dev_data.set_input('tag')
dev_data.set_target('seq_len')
#print(train_data.get_field_names())
trainer = Trainer(
train_data=train_data,
model=model,
# loss=loss,
metrics=metrics,
metric_key='f',
batch_size=config.batch_size,
n_epochs=config.epoch,
dev_data=dev_data,
save_path=config.save_path,
check_code_level=-1,
print_every=config.print_every,
validate_every=config.validate_every,
optimizer=optimizer,
use_tqdm=False,
device=config.device,
callbacks=[timing, early_stop, logs])
trainer.train()
# test result
tester = Tester(dev_data,
model,
metrics=metrics,
device=config.device,
batch_size=config.batch_size)
tester.test()
def train(args):
text_data = TextData()
with open(os.path.join(args.vocab_dir, args.vocab_data), 'rb') as fin:
text_data = pickle.load(fin)
vocab_size = text_data.vocab_size
class_num = text_data.class_num
seq_len = text_data.max_seq_len
print("(vocab_size,class_num,seq_len):({0},{1},{2})".format(
vocab_size, class_num, seq_len))
train_data = text_data.train_set
test_dev_data = text_data.test_set
train_data.set_input('words', 'seq_len')
train_data.set_target('target')
test_dev_data.set_input('words', 'seq_len')
test_dev_data.set_target('target')
test_data, dev_data = test_dev_data.split(0.2)
test_data = test_dev_data
init_embeds = None
if args.pretrain_model == "None":
print("No pretrained model with be used.")
print("vocabsize:{0}".format(vocab_size))
init_embeds = (vocab_size, args.embed_size)
elif args.pretrain_model == "word2vec":
embeds_path = os.path.join(args.prepare_dir, 'w2v_embeds.pkl')
print("Loading Word2Vec pretrained embedding from {0}.".format(
embeds_path))
with open(embeds_path, 'rb') as fin:
init_embeds = pickle.load(fin)
elif args.pretrain_model == 'glove':
embeds_path = os.path.join(args.prepare_dir, 'glove_embeds.pkl')
print(
"Loading Glove pretrained embedding from {0}.".format(embeds_path))
with open(embeds_path, 'rb') as fin:
init_embeds = pickle.load(fin)
elif args.pretrain_model == 'glove2wv':
embeds_path = os.path.join(args.prepare_dir, 'glove2wv_embeds.pkl')
print(
"Loading Glove pretrained embedding from {0}.".format(embeds_path))
with open(embeds_path, 'rb') as fin:
init_embeds = pickle.load(fin)
else:
init_embeds = (vocab_size, args.embed_size)
if args.model == "CNNText":
print("Using CNN Model.")
model = CNNText(init_embeds,
num_classes=class_num,
padding=2,
dropout=args.dropout)
elif args.model == "StarTransformer":
print("Using StarTransformer Model.")
model = STSeqCls(init_embeds,
num_cls=class_num,
hidden_size=args.hidden_size)
elif args.model == "MyCNNText":
model = MyCNNText(init_embeds=init_embeds,
num_classes=class_num,
padding=2,
dropout=args.dropout)
print("Using user defined CNNText")
elif args.model == "LSTMText":
print("Using LSTM Model.")
model = LSTMText(init_embeds=init_embeds,
output_dim=class_num,
hidden_dim=args.hidden_size,
num_layers=args.num_layers,
dropout=args.dropout)
elif args.model == "Bert":
print("Using Bert Model.")
else:
print("Using default model: CNNText.")
model = CNNText((vocab_size, args.embed_size),
num_classes=class_num,
padding=2,
dropout=0.1)
print(model)
if args.cuda:
device = torch.device('cuda')
else:
device = None
print("train_size:{0} ; dev_size:{1} ; test_size:{2}".format(
train_data.get_length(), dev_data.get_length(),
test_data.get_length()))
if args.optim == "Adam":
print("Using Adam as optimizer.")
optimizer = fastnlp_optim.Adam(lr=0.001,
weight_decay=args.weight_decay)
if (args.model_suffix == "default"):
args.model_suffix == args.optim
else:
print("No Optimizer will be used.")
optimizer = None
criterion = CrossEntropyLoss()
metric = AccuracyMetric()
model_save_path = os.path.join(args.model_dir, args.model,
args.model_suffix)
earlystop = EarlyStopCallback(args.patience)
trainer = Trainer(train_data=train_data,
model=model,
save_path=model_save_path,
device=device,
n_epochs=args.epochs,
optimizer=optimizer,
dev_data=test_data,
loss=criterion,
batch_size=args.batch_size,
metrics=metric,
callbacks=[FitlogCallback(test_data), earlystop])
trainer.train()
print("Train Done.")
tester = Tester(data=test_data,
model=model,
metrics=metric,
batch_size=args.batch_size,
device=device)
tester.test()
print("Test Done.")
fitlog.finish()
def train(config, task_name):
train_data = pickle.load(
open(os.path.join(config.data_path, config.train_name), "rb"))
# debug
if config.debug:
train_data = train_data[0:30]
dev_data = pickle.load(
open(os.path.join(config.data_path, config.dev_name), "rb"))
# test_data = pickle.load(open(os.path.join(config.data_path, config.test_name), "rb"))
vocabulary = pickle.load(
open(os.path.join(config.data_path, config.vocabulary_name), "rb"))
# load w2v data
# weight = pickle.load(open(os.path.join(config.data_path, config.weight_name), "rb"))
if task_name == "lstm":
text_model = LSTM(vocab_size=len(vocabulary),
embed_dim=config.embed_dim,
output_dim=config.class_num,
hidden_dim=config.hidden_dim,
num_layers=config.num_layers,
dropout=config.dropout)
elif task_name == "lstm_maxpool":
text_model = LSTM_maxpool(vocab_size=len(vocabulary),
embed_dim=config.embed_dim,
output_dim=config.class_num,
hidden_dim=config.hidden_dim,
num_layers=config.num_layers,
dropout=config.dropout)
elif task_name == "cnn":
text_model = CNN(vocab_size=len(vocabulary),
embed_dim=config.embed_dim,
class_num=config.class_num,
kernel_num=config.kernel_num,
kernel_sizes=config.kernel_sizes,
dropout=config.dropout,
static=config.static,
in_channels=config.in_channels)
elif task_name == "rnn":
text_model = RNN(vocab_size=len(vocabulary),
embed_dim=config.embed_dim,
output_dim=config.class_num,
hidden_dim=config.hidden_dim,
num_layers=config.num_layers,
dropout=config.dropout)
# elif task_name == "cnn_w2v":
# text_model = CNN_w2v(vocab_size=len(vocabulary), embed_dim=config.embed_dim,
# class_num=config.class_num, kernel_num=config.kernel_num,
# kernel_sizes=config.kernel_sizes, dropout=config.dropout,
# static=config.static, in_channels=config.in_channels,
# weight=weight)
elif task_name == "rcnn":
text_model = RCNN(vocab_size=len(vocabulary),
embed_dim=config.embed_dim,
output_dim=config.class_num,
hidden_dim=config.hidden_dim,
num_layers=config.num_layers,
dropout=config.dropout)
#elif task_name == "bert":
# text_model = BertModel.from_pretrained(config.bert_path)
optimizer = Adam(lr=config.lr, weight_decay=config.weight_decay)
timing = TimingCallback()
early_stop = EarlyStopCallback(config.patience)
logs = FitlogCallback(dev_data)
f1 = F1_score(pred='output', target='target')
trainer = Trainer(train_data=train_data,
model=text_model,
loss=BCEWithLogitsLoss(),
batch_size=config.batch_size,
check_code_level=-1,
metrics=f1,
metric_key='f1',
n_epochs=config.epoch,
dev_data=dev_data,
save_path=config.save_path,
print_every=config.print_every,
validate_every=config.validate_every,
optimizer=optimizer,
use_tqdm=False,
device=config.device,
callbacks=[timing, early_stop, logs])
trainer.train()
# test result
tester = Tester(
dev_data,
text_model,
metrics=f1,
device=config.device,
batch_size=config.batch_size,
)
tester.test()
"hidden_size": arg.hidden_size,
"dropout": arg.dropout,
"use_allennlp": False,
},
)
optimizer = Adadelta(lr=arg.lr, params=model.parameters())
scheduler = StepLR(optimizer, step_size=10, gamma=0.5)
callbacks = [
LRScheduler(scheduler),
]
if arg.task in ['snli']:
callbacks.append(
FitlogCallback(data_info.datasets[arg.testset_name], verbose=1))
elif arg.task == 'mnli':
callbacks.append(
FitlogCallback(
{
'dev_matched': data_info.datasets['dev_matched'],
'dev_mismatched': data_info.datasets['dev_mismatched']
},
verbose=1))
trainer = Trainer(train_data=data_info.datasets['train'],
model=model,
optimizer=optimizer,
num_workers=0,
batch_size=arg.batch_size,
n_epochs=arg.n_epochs,
optimizer = optim.AdamW(param_, lr=args.lr, weight_decay=args.weight_decay)
elif args.optim == 'sgd':
# optimizer = optim.SGD(model.parameters(),lr=args.lr,momentum=args.momentum,
# weight_decay=args.weight_decay)
optimizer = optim.SGD(param_,
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
if 'msra' in args.dataset:
datasets['dev'] = datasets['test']
fitlog_evaluate_dataset = {'test': datasets['test']}
if args.test_train:
fitlog_evaluate_dataset['train'] = datasets['train']
evaluate_callback = FitlogCallback(fitlog_evaluate_dataset, verbose=1)
lrschedule_callback = LRScheduler(
lr_scheduler=LambdaLR(optimizer, lambda ep: 1 / (1 + 0.05 * ep)))
clip_callback = GradientClipCallback(clip_type='value', clip_value=5)
# model.state_dict()
class CheckWeightCallback(Callback):
def __init__(self, model):
super().__init__()
self.model_ = model
def on_step_end(self):
print('parameter weight:', flush=True)
print(self.model_.state_dict()['encoder.layer_0.attn.w_q.weight'],
flush=True)
embed_dim=128,
hidden_dim=hidden_units,
output_dim=8)
# mymodel = load_model(mymodel, './model/best_TextModel_acc_2019-06-28-09-07-50')
trainer = Trainer(
train_data=train_data,
model=mymodel,
loss=CrossEntropyLoss(pred='pred', target='target'),
# loss=SkipBudgetLoss(pred='pred', target='target', updated_states='updated_states'),
metrics=[AccuracyMetric(), UsedStepsMetric()],
n_epochs=30,
batch_size=batch_size,
print_every=-1,
validate_every=-1,
dev_data=test_data,
save_path='./model',
optimizer=Adam(lr=learning_rate, weight_decay=0),
check_code_level=0,
device="cuda",
metric_key='acc',
use_tqdm=False,
callbacks=[FitlogCallback(test_data)])
start = time.clock()
trainer.train()
end = time.clock()
training_time = end - start
print('total training time:%fs' % (end - start))
fitlog.add_hyper({'time': training_time})
fitlog.finish()