- 修复processor的bug - 更新core/的若干组件, 去除batch的冗余参数 - CRF有个打字错误?已修复 - 更新pos tag 训练脚本tags/v0.2.0
@@ -1,14 +1,18 @@ | |||||
import torch | import torch | ||||
from fastNLP.core.dataset import DataSet | |||||
from fastNLP.core.instance import Instance | |||||
from fastNLP.core.predictor import Predictor | |||||
class API: | class API: | ||||
def __init__(self): | def __init__(self): | ||||
self.pipeline = None | self.pipeline = None | ||||
self.model = None | self.model = None | ||||
def predict(self): | |||||
pass | |||||
def predict(self, *args, **kwargs): | |||||
raise NotImplementedError | |||||
def load(self, name): | def load(self, name): | ||||
_dict = torch.load(name) | _dict = torch.load(name) | ||||
@@ -19,3 +23,47 @@ class API: | |||||
_dict = {'pipeline': self.pipeline, | _dict = {'pipeline': self.pipeline, | ||||
'model': self.model} | 'model': self.model} | ||||
torch.save(_dict, path) | torch.save(_dict, path) | ||||
class POS_tagger(API): | |||||
"""FastNLP API for Part-Of-Speech tagging. | |||||
""" | |||||
def __init__(self): | |||||
super(POS_tagger, self).__init__() | |||||
def predict(self, query): | |||||
""" | |||||
:param query: list of list of str. Each string is a token(word). | |||||
:return answer: list of list of str. Each string is a tag. | |||||
""" | |||||
self.load("/home/zyfeng/fastnlp_0.2.0/reproduction/pos_tag_model/model_pp.pkl") | |||||
data = DataSet() | |||||
for example in query: | |||||
data.append(Instance(words=example)) | |||||
data = self.pipeline(data) | |||||
predictor = Predictor() | |||||
outputs = predictor.predict(self.model, data) | |||||
answers = [] | |||||
for out in outputs: | |||||
out = out.numpy() | |||||
for sent in out: | |||||
answers.append([self.tag_vocab.to_word(tag) for tag in sent]) | |||||
return answers | |||||
def load(self, name): | |||||
_dict = torch.load(name) | |||||
self.pipeline = _dict['pipeline'] | |||||
self.model = _dict['model'] | |||||
self.tag_vocab = _dict["tag_vocab"] | |||||
if __name__ == "__main__": | |||||
tagger = POS_tagger() | |||||
print(tagger.predict([["我", "是", "学生", "。"], ["我", "是", "学生", "。"]])) |
@@ -11,7 +11,7 @@ class Pipeline: | |||||
self.pipeline = [] | self.pipeline = [] | ||||
if isinstance(processors, list): | if isinstance(processors, list): | ||||
for proc in processors: | for proc in processors: | ||||
assert isinstance(proc, Processor), "Must be a Processor, not {}.".format(type(processor)) | |||||
assert isinstance(proc, Processor), "Must be a Processor, not {}.".format(type(proc)) | |||||
self.pipeline = processors | self.pipeline = processors | ||||
def add_processor(self, processor): | def add_processor(self, processor): | ||||
@@ -21,7 +21,7 @@ class Pipeline: | |||||
def process(self, dataset): | def process(self, dataset): | ||||
assert len(self.pipeline) != 0, "You need to add some processor first." | assert len(self.pipeline) != 0, "You need to add some processor first." | ||||
for proc_name, proc in self.pipeline: | |||||
for proc in self.pipeline: | |||||
dataset = proc(dataset) | dataset = proc(dataset) | ||||
return dataset | return dataset | ||||
@@ -9,7 +9,7 @@ class Batch(object): | |||||
""" | """ | ||||
def __init__(self, dataset, batch_size, sampler, use_cuda, sort_in_batch=False, sort_key=None): | |||||
def __init__(self, dataset, batch_size, sampler, use_cuda): | |||||
""" | """ | ||||
:param dataset: a DataSet object | :param dataset: a DataSet object | ||||
@@ -22,8 +22,6 @@ class Batch(object): | |||||
self.batch_size = batch_size | self.batch_size = batch_size | ||||
self.sampler = sampler | self.sampler = sampler | ||||
self.use_cuda = use_cuda | self.use_cuda = use_cuda | ||||
self.sort_in_batch = sort_in_batch | |||||
self.sort_key = sort_key if sort_key is not None else 'word_seq' | |||||
self.idx_list = None | self.idx_list = None | ||||
self.curidx = 0 | self.curidx = 0 | ||||
@@ -119,7 +119,7 @@ class DataSet(object): | |||||
assert isinstance(val, bool) | assert isinstance(val, bool) | ||||
self.field_arrays[name].is_target = val | self.field_arrays[name].is_target = val | ||||
else: | else: | ||||
raise KeyError | |||||
raise KeyError("{} is not a valid field name.".format(name)) | |||||
return self | return self | ||||
def set_need_tensor(self, **kwargs): | def set_need_tensor(self, **kwargs): | ||||
@@ -43,12 +43,11 @@ class SeqLabelEvaluator(Evaluator): | |||||
:return accuracy: | :return accuracy: | ||||
""" | """ | ||||
truth = [item["truth"] for item in truth] | truth = [item["truth"] for item in truth] | ||||
total_correct, total_count= 0., 0. | |||||
total_correct, total_count = 0., 0. | |||||
for x, y in zip(predict, truth): | for x, y in zip(predict, truth): | ||||
x = torch.Tensor(x) | |||||
x = torch.tensor(x) | |||||
y = y.to(x) # make sure they are in the same device | y = y.to(x) # make sure they are in the same device | ||||
mask = x.ge(1).float() | |||||
# correct = torch.sum(x * mask.float() == (y * mask.long()).float()) | |||||
mask = x.ge(1).long() | |||||
correct = torch.sum(x * mask == y * mask) | correct = torch.sum(x * mask == y * mask) | ||||
correct -= torch.sum(x.le(0)) | correct -= torch.sum(x.le(0)) | ||||
total_correct += float(correct) | total_correct += float(correct) | ||||
@@ -74,7 +74,7 @@ class Tester(object): | |||||
output_list = [] | output_list = [] | ||||
truth_list = [] | truth_list = [] | ||||
data_iterator = Batch(dev_data, self.batch_size, sampler=RandomSampler(), use_cuda=self.use_cuda, sort_in_batch=True, sort_key='word_seq') | |||||
data_iterator = Batch(dev_data, self.batch_size, sampler=RandomSampler(), use_cuda=self.use_cuda) | |||||
with torch.no_grad(): | with torch.no_grad(): | ||||
for batch_x, batch_y in data_iterator: | for batch_x, batch_y in data_iterator: | ||||
@@ -11,6 +11,7 @@ from fastNLP.core.metrics import Evaluator | |||||
from fastNLP.core.optimizer import Optimizer | from fastNLP.core.optimizer import Optimizer | ||||
from fastNLP.core.sampler import RandomSampler | from fastNLP.core.sampler import RandomSampler | ||||
from fastNLP.core.tester import SeqLabelTester, ClassificationTester, SNLITester | from fastNLP.core.tester import SeqLabelTester, ClassificationTester, SNLITester | ||||
from fastNLP.core.tester import Tester | |||||
from fastNLP.saver.logger import create_logger | from fastNLP.saver.logger import create_logger | ||||
from fastNLP.saver.model_saver import ModelSaver | from fastNLP.saver.model_saver import ModelSaver | ||||
@@ -144,7 +145,7 @@ class Trainer(object): | |||||
# prepare mini-batch iterator | # prepare mini-batch iterator | ||||
data_iterator = Batch(train_data, batch_size=self.batch_size, sampler=RandomSampler(), | data_iterator = Batch(train_data, batch_size=self.batch_size, sampler=RandomSampler(), | ||||
use_cuda=self.use_cuda, sort_in_batch=True, sort_key='word_seq') | |||||
use_cuda=self.use_cuda) | |||||
logger.info("prepared data iterator") | logger.info("prepared data iterator") | ||||
# one forward and backward pass | # one forward and backward pass | ||||
@@ -230,7 +231,6 @@ class Trainer(object): | |||||
def update(self): | def update(self): | ||||
"""Perform weight update on a model. | """Perform weight update on a model. | ||||
For PyTorch, just call optimizer to update. | |||||
""" | """ | ||||
self._optimizer.step() | self._optimizer.step() | ||||
@@ -319,7 +319,7 @@ class Trainer(object): | |||||
ModelSaver(os.path.join(self.pickle_path, model_name)).save_pytorch(network) | ModelSaver(os.path.join(self.pickle_path, model_name)).save_pytorch(network) | ||||
def _create_validator(self, valid_args): | def _create_validator(self, valid_args): | ||||
raise NotImplementedError | |||||
return Tester(**valid_args) | |||||
def set_validator(self, validor): | def set_validator(self, validor): | ||||
self.validator = validor | self.validator = validor | ||||
@@ -116,11 +116,11 @@ class AdvSeqLabel(SeqLabeling): | |||||
num_classes = args["num_classes"] | num_classes = args["num_classes"] | ||||
self.Embedding = encoder.embedding.Embedding(vocab_size, word_emb_dim, init_emb=emb) | self.Embedding = encoder.embedding.Embedding(vocab_size, word_emb_dim, init_emb=emb) | ||||
self.Rnn = encoder.lstm.LSTM(word_emb_dim, hidden_dim, num_layers=3, dropout=0.3, bidirectional=True) | |||||
self.Rnn = encoder.lstm.LSTM(word_emb_dim, hidden_dim, num_layers=3, dropout=0.5, bidirectional=True) | |||||
self.Linear1 = encoder.Linear(hidden_dim * 2, hidden_dim * 2 // 3) | self.Linear1 = encoder.Linear(hidden_dim * 2, hidden_dim * 2 // 3) | ||||
self.batch_norm = torch.nn.BatchNorm1d(hidden_dim * 2 // 3) | self.batch_norm = torch.nn.BatchNorm1d(hidden_dim * 2 // 3) | ||||
self.relu = torch.nn.ReLU() | self.relu = torch.nn.ReLU() | ||||
self.drop = torch.nn.Dropout(0.3) | |||||
self.drop = torch.nn.Dropout(0.5) | |||||
self.Linear2 = encoder.Linear(hidden_dim * 2 // 3, num_classes) | self.Linear2 = encoder.Linear(hidden_dim * 2 // 3, num_classes) | ||||
self.Crf = decoder.CRF.ConditionalRandomField(num_classes) | self.Crf = decoder.CRF.ConditionalRandomField(num_classes) | ||||
@@ -135,7 +135,7 @@ class AdvSeqLabel(SeqLabeling): | |||||
""" | """ | ||||
word_seq = word_seq.long() | word_seq = word_seq.long() | ||||
word_seq_origin_len = word_seq_origin_len.long() | word_seq_origin_len = word_seq_origin_len.long() | ||||
truth = truth.long() | |||||
truth = truth.long() if truth is not None else None | |||||
self.mask = self.make_mask(word_seq, word_seq_origin_len) | self.mask = self.make_mask(word_seq, word_seq_origin_len) | ||||
batch_size = word_seq.size(0) | batch_size = word_seq.size(0) | ||||
@@ -128,7 +128,7 @@ class ConditionalRandomField(nn.Module): | |||||
vpath = data.new_zeros((seq_len, batch_size, n_tags), dtype=torch.long) | vpath = data.new_zeros((seq_len, batch_size, n_tags), dtype=torch.long) | ||||
vscore = data[0] | vscore = data[0] | ||||
if self.include_start_end_trans: | if self.include_start_end_trans: | ||||
vscore += self.start_scores.view(1. -1) | |||||
vscore += self.start_scores.view(1, -1) | |||||
for i in range(1, seq_len): | for i in range(1, seq_len): | ||||
prev_score = vscore.view(batch_size, n_tags, 1) | prev_score = vscore.view(batch_size, n_tags, 1) | ||||
cur_score = data[i].view(batch_size, 1, n_tags) | cur_score = data[i].view(batch_size, 1, n_tags) | ||||
@@ -1,6 +1,6 @@ | |||||
[train] | [train] | ||||
epochs = 20 | |||||
batch_size = 32 | |||||
epochs = 5 | |||||
batch_size = 64 | |||||
pickle_path = "./save/" | pickle_path = "./save/" | ||||
validate = false | validate = false | ||||
save_best_dev = true | save_best_dev = true | ||||
@@ -1,3 +1,4 @@ | |||||
import copy | |||||
import os | import os | ||||
import torch | import torch | ||||
@@ -6,6 +7,7 @@ from fastNLP.api.pipeline import Pipeline | |||||
from fastNLP.api.processor import VocabProcessor, IndexerProcessor, SeqLenProcessor | from fastNLP.api.processor import VocabProcessor, IndexerProcessor, SeqLenProcessor | ||||
from fastNLP.core.dataset import DataSet | from fastNLP.core.dataset import DataSet | ||||
from fastNLP.core.instance import Instance | from fastNLP.core.instance import Instance | ||||
from fastNLP.core.metrics import SeqLabelEvaluator | |||||
from fastNLP.core.optimizer import Optimizer | from fastNLP.core.optimizer import Optimizer | ||||
from fastNLP.core.trainer import Trainer | from fastNLP.core.trainer import Trainer | ||||
from fastNLP.loader.config_loader import ConfigLoader, ConfigSection | from fastNLP.loader.config_loader import ConfigLoader, ConfigSection | ||||
@@ -13,9 +15,12 @@ from fastNLP.loader.dataset_loader import PeopleDailyCorpusLoader | |||||
from fastNLP.models.sequence_modeling import AdvSeqLabel | from fastNLP.models.sequence_modeling import AdvSeqLabel | ||||
cfgfile = './pos_tag.cfg' | cfgfile = './pos_tag.cfg' | ||||
# datadir = "/home/zyfeng/data/" | |||||
# data_name = "POS_PD_1998.txt" | |||||
datadir = "/home/zyfeng/fastnlp_0.2.0/test/data_for_tests/" | datadir = "/home/zyfeng/fastnlp_0.2.0/test/data_for_tests/" | ||||
data_name = "people_daily_raw.txt" | data_name = "people_daily_raw.txt" | ||||
pos_tag_data_path = os.path.join(datadir, data_name) | pos_tag_data_path = os.path.join(datadir, data_name) | ||||
pickle_path = "save" | pickle_path = "save" | ||||
data_infer_path = os.path.join(datadir, "infer.utf8") | data_infer_path = os.path.join(datadir, "infer.utf8") | ||||
@@ -54,6 +59,9 @@ def train(): | |||||
seq_len_proc = SeqLenProcessor("word_seq", "word_seq_origin_len") | seq_len_proc = SeqLenProcessor("word_seq", "word_seq_origin_len") | ||||
seq_len_proc(dataset) | seq_len_proc(dataset) | ||||
dev_set = copy.deepcopy(dataset) | |||||
dev_set.set_is_target(truth=True) | |||||
print("processors defined") | print("processors defined") | ||||
# dataset.set_is_target(tag_ids=True) | # dataset.set_is_target(tag_ids=True) | ||||
model_param["vocab_size"] = len(word_vocab_proc.get_vocab()) | model_param["vocab_size"] = len(word_vocab_proc.get_vocab()) | ||||
@@ -66,14 +74,15 @@ def train(): | |||||
# call trainer to train | # call trainer to train | ||||
trainer = Trainer(epochs=train_param["epochs"], | trainer = Trainer(epochs=train_param["epochs"], | ||||
batch_size=train_param["batch_size"], | batch_size=train_param["batch_size"], | ||||
validate=False, | |||||
validate=True, | |||||
optimizer=Optimizer("SGD", lr=0.01, momentum=0.9), | optimizer=Optimizer("SGD", lr=0.01, momentum=0.9), | ||||
evaluator=SeqLabelEvaluator() | |||||
) | ) | ||||
trainer.train(model, dataset) | |||||
trainer.train(model, dataset, dev_set) | |||||
# save model & pipeline | # save model & pipeline | ||||
pp = Pipeline([word_vocab_proc, word_indexer, seq_len_proc]) | |||||
save_dict = {"pipeline": pp, "model": model} | |||||
pp = Pipeline([word_indexer, seq_len_proc]) | |||||
save_dict = {"pipeline": pp, "model": model, "tag_vocab": tag_vocab_proc.get_vocab()} | |||||
torch.save(save_dict, "model_pp.pkl") | torch.save(save_dict, "model_pp.pkl") | ||||