AllenNLP 共指消解的多 GPU 训练
Multi-GPU training of AllenNLP coreference resolution
我正在尝试复制（或接近）End-to-end Neural Coreference Resolution paper on the CoNLL-2012 shared task. I intend to do some enhancements on top of this, so I decided to use AllenNLP's CoreferenceResolver 获得的结果。这就是我初始化和训练模型的方式：
import torch
from allennlp.common import Params
from allennlp.data import Vocabulary
from allennlp.data.dataset_readers import ConllCorefReader
from allennlp.data.dataset_readers.dataset_utils import Ontonotes
from allennlp.data.iterators import BasicIterator, MultiprocessIterator
from allennlp.data.token_indexers import SingleIdTokenIndexer, TokenCharactersIndexer
from allennlp.models import CoreferenceResolver
from allennlp.modules import Embedding, FeedForward
from allennlp.modules.seq2seq_encoders import PytorchSeq2SeqWrapper
from allennlp.modules.seq2vec_encoders import CnnEncoder
from allennlp.modules.text_field_embedders import BasicTextFieldEmbedder
from allennlp.modules.token_embedders import TokenCharactersEncoder
from allennlp.training import Trainer
from allennlp.training.learning_rate_schedulers import LearningRateScheduler
from torch.nn import LSTM, ReLU
from torch.optim import Adam
def read_data(directory_path):
data = []
for file_path in Ontonotes().dataset_path_iterator(directory_path):
data += dataset_reader.read(file_path)
return data
INPUT_FILE_PATH_TEMPLATE = "data/CoNLL-2012/v4/data/%s"
dataset_reader = ConllCorefReader(10, {"tokens": SingleIdTokenIndexer(),
"token_characters": TokenCharactersIndexer()})
training_data = read_data(INPUT_FILE_PATH_TEMPLATE % "train")
validation_data = read_data(INPUT_FILE_PATH_TEMPLATE % "development")
vocabulary = Vocabulary.from_instances(training_data + validation_data)
model = CoreferenceResolver(vocab=vocabulary,
text_field_embedder=BasicTextFieldEmbedder({"tokens": Embedding.from_params(vocabulary, Params({"embedding_dim": embeddings_dimension, "pretrained_file": "glove.840B.300d.txt"})),
"token_characters": TokenCharactersEncoder(embedding=Embedding(num_embeddings=vocabulary.get_vocab_size("token_characters"), embedding_dim=8, vocab_namespace="token_characters"),
encoder=CnnEncoder(embedding_dim=8, num_filters=50, ngram_filter_sizes=(3, 4, 5), output_dim=100))}),
context_layer=PytorchSeq2SeqWrapper(LSTM(input_size=400, hidden_size=200, num_layers=1, dropout=0.2, bidirectional=True, batch_first=True)),
mention_feedforward=FeedForward(input_dim=1220, num_layers=2, hidden_dims=[150, 150], activations=[ReLU(), ReLU()], dropout=[0.2, 0.2]),
antecedent_feedforward=FeedForward(input_dim=3680, num_layers=2, hidden_dims=[150, 150], activations=[ReLU(), ReLU()], dropout=[0.2, 0.2]),
feature_size=20,
max_span_width=10,
spans_per_word=0.4,
max_antecedents=250,
lexical_dropout=0.5)
if torch.cuda.is_available():
cuda_device = 0
model = model.cuda(cuda_device)
else:
cuda_device = -1
iterator = BasicIterator(batch_size=1)
iterator.index_with(vocabulary)
optimiser = Adam(model.parameters(), weight_decay=0.1)
Trainer(model=model,
train_dataset=training_data,
validation_dataset=validation_data,
optimizer=optimiser,
learning_rate_scheduler=LearningRateScheduler.from_params(optimiser, Params({"type": "step", "step_size": 100})),
iterator=iterator,
num_epochs=150,
patience=1,
cuda_device=cuda_device).train()
读取数据后，我训练了模型，但 运行 GPU 内存不足：RuntimeError: CUDA out of memory. Tried to allocate 4.43 GiB (GPU 0; 11.17 GiB total capacity; 3.96 GiB already allocated; 3.40 GiB free; 3.47 GiB cached)。因此，我尝试使用多个 GPU 来训练这个模型。我正在使用 Tesla K80s（具有 12GiB 内存）。
我试过使用 AllenNLP 的 MultiprocessIterator, by itialising the iterator as MultiprocessIterator(BasicIterator(batch_size=1), num_workers=torch.cuda.device_count()). However, only 1 GPU is being used (by monitoring the memory usage through the nvidia-smi command) & got the error below. I also tried fiddling with its parameters (increasing num_workers or decreasing output_queue_size) & the ulimit (as mentioned by this PyTorch issue) 但无济于事。
Process Process-3:
Traceback (most recent call last):
File "/usr/lib/python3.6/multiprocessing/process.py", line 258, in _bootstrap
self.run()
Traceback (most recent call last):
File "/usr/lib/python3.6/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/usr/lib/python3.6/multiprocessing/process.py", line 93, in run
self._target(*self._args, **self._kwargs)
File "/home/user/.local/lib/python3.6/site-packages/allennlp/data/iterators/multiprocess_iterator.py", line 32, in _create_tensor_dicts
output_queue.put(tensor_dict)
File "/usr/lib/python3.6/multiprocessing/process.py", line 93, in run
self._target(*self._args, **self._kwargs)
File "/home/user/.local/lib/python3.6/site-packages/allennlp/data/iterators/multiprocess_iterator.py", line 32, in _create_tensor_dicts
output_queue.put(tensor_dict)
File "<string>", line 2, in put
File "<string>", line 2, in put
File "/usr/lib/python3.6/multiprocessing/managers.py", line 772, in _callmethod
raise convert_to_error(kind, result)
File "/usr/lib/python3.6/multiprocessing/managers.py", line 772, in _callmethod
raise convert_to_error(kind, result)
multiprocessing.managers.RemoteError:
---------------------------------------------------------------------------
Traceback (most recent call last):
File "/usr/lib/python3.6/multiprocessing/managers.py", line 228, in serve_client
request = recv()
File "/usr/lib/python3.6/multiprocessing/connection.py", line 251, in recv
return _ForkingPickler.loads(buf.getbuffer())
File "/home/user/.local/lib/python3.6/site-packages/torch/multiprocessing/reductions.py", line 276, in rebuild_storage_fd
fd = df.detach()
File "/usr/lib/python3.6/multiprocessing/resource_sharer.py", line 58, in detach
return reduction.recv_handle(conn)
File "/usr/lib/python3.6/multiprocessing/reduction.py", line 182, in recv_handle
return recvfds(s, 1)[0]
File "/usr/lib/python3.6/multiprocessing/reduction.py", line 161, in recvfds
len(ancdata))
RuntimeError: received 0 items of ancdata
---------------------------------------------------------------------------
我还尝试通过 PyTorch's DataParallel 实现这一点，方法是将模型的 context_layer、mention_feedforward、antecedent_feedforward 与自定义 DataParallelWrapper 包装起来（以提供兼容性使用 AllenNLP 假设的 class 函数）。尽管如此，只使用了 1 个 GPU，它最终还是像以前一样耗尽了内存。
class DataParallelWrapper(DataParallel):
def __init__(self, module):
super().__init__(module)
def get_output_dim(self):
return self.module.get_output_dim()
def get_input_dim(self):
return self.module.get_input_dim()
def forward(self, *inputs):
return self.module.forward(inputs)
在深入研究代码后，我发现 AllenNLP 直接通过其 Trainer 在幕后执行此操作。 cuda_device 可以是单个 int（在单处理的情况下）或 list 的 ints（在多处理的情况下）：
cuda_device : Union[int, List[int]], optional (default = -1)
An integer or list of integers specifying the CUDA device(s) to use. If -1, the CPU is used.
所以应该传递所有需要的 GPU 设备：
if torch.cuda.is_available():
cuda_device = list(range(torch.cuda.device_count()))
model = model.cuda(cuda_device[0])
else:
cuda_device = -1
请注意，model 仍然需要手动移动到 GPU（通过 model.cuda(...)），否则它会尝试使用多个 CPU。
我正在尝试复制（或接近）End-to-end Neural Coreference Resolution paper on the CoNLL-2012 shared task. I intend to do some enhancements on top of this, so I decided to use AllenNLP's CoreferenceResolver 获得的结果。这就是我初始化和训练模型的方式：
import torch
from allennlp.common import Params
from allennlp.data import Vocabulary
from allennlp.data.dataset_readers import ConllCorefReader
from allennlp.data.dataset_readers.dataset_utils import Ontonotes
from allennlp.data.iterators import BasicIterator, MultiprocessIterator
from allennlp.data.token_indexers import SingleIdTokenIndexer, TokenCharactersIndexer
from allennlp.models import CoreferenceResolver
from allennlp.modules import Embedding, FeedForward
from allennlp.modules.seq2seq_encoders import PytorchSeq2SeqWrapper
from allennlp.modules.seq2vec_encoders import CnnEncoder
from allennlp.modules.text_field_embedders import BasicTextFieldEmbedder
from allennlp.modules.token_embedders import TokenCharactersEncoder
from allennlp.training import Trainer
from allennlp.training.learning_rate_schedulers import LearningRateScheduler
from torch.nn import LSTM, ReLU
from torch.optim import Adam
def read_data(directory_path):
data = []
for file_path in Ontonotes().dataset_path_iterator(directory_path):
data += dataset_reader.read(file_path)
return data
INPUT_FILE_PATH_TEMPLATE = "data/CoNLL-2012/v4/data/%s"
dataset_reader = ConllCorefReader(10, {"tokens": SingleIdTokenIndexer(),
"token_characters": TokenCharactersIndexer()})
training_data = read_data(INPUT_FILE_PATH_TEMPLATE % "train")
validation_data = read_data(INPUT_FILE_PATH_TEMPLATE % "development")
vocabulary = Vocabulary.from_instances(training_data + validation_data)
model = CoreferenceResolver(vocab=vocabulary,
text_field_embedder=BasicTextFieldEmbedder({"tokens": Embedding.from_params(vocabulary, Params({"embedding_dim": embeddings_dimension, "pretrained_file": "glove.840B.300d.txt"})),
"token_characters": TokenCharactersEncoder(embedding=Embedding(num_embeddings=vocabulary.get_vocab_size("token_characters"), embedding_dim=8, vocab_namespace="token_characters"),
encoder=CnnEncoder(embedding_dim=8, num_filters=50, ngram_filter_sizes=(3, 4, 5), output_dim=100))}),
context_layer=PytorchSeq2SeqWrapper(LSTM(input_size=400, hidden_size=200, num_layers=1, dropout=0.2, bidirectional=True, batch_first=True)),
mention_feedforward=FeedForward(input_dim=1220, num_layers=2, hidden_dims=[150, 150], activations=[ReLU(), ReLU()], dropout=[0.2, 0.2]),
antecedent_feedforward=FeedForward(input_dim=3680, num_layers=2, hidden_dims=[150, 150], activations=[ReLU(), ReLU()], dropout=[0.2, 0.2]),
feature_size=20,
max_span_width=10,
spans_per_word=0.4,
max_antecedents=250,
lexical_dropout=0.5)
if torch.cuda.is_available():
cuda_device = 0
model = model.cuda(cuda_device)
else:
cuda_device = -1
iterator = BasicIterator(batch_size=1)
iterator.index_with(vocabulary)
optimiser = Adam(model.parameters(), weight_decay=0.1)
Trainer(model=model,
train_dataset=training_data,
validation_dataset=validation_data,
optimizer=optimiser,
learning_rate_scheduler=LearningRateScheduler.from_params(optimiser, Params({"type": "step", "step_size": 100})),
iterator=iterator,
num_epochs=150,
patience=1,
cuda_device=cuda_device).train()
读取数据后，我训练了模型，但 运行 GPU 内存不足：RuntimeError: CUDA out of memory. Tried to allocate 4.43 GiB (GPU 0; 11.17 GiB total capacity; 3.96 GiB already allocated; 3.40 GiB free; 3.47 GiB cached)。因此，我尝试使用多个 GPU 来训练这个模型。我正在使用 Tesla K80s（具有 12GiB 内存）。
我试过使用 AllenNLP 的 MultiprocessIterator, by itialising the iterator as MultiprocessIterator(BasicIterator(batch_size=1), num_workers=torch.cuda.device_count()). However, only 1 GPU is being used (by monitoring the memory usage through the nvidia-smi command) & got the error below. I also tried fiddling with its parameters (increasing num_workers or decreasing output_queue_size) & the ulimit (as mentioned by this PyTorch issue) 但无济于事。
Process Process-3:
Traceback (most recent call last):
File "/usr/lib/python3.6/multiprocessing/process.py", line 258, in _bootstrap
self.run()
Traceback (most recent call last):
File "/usr/lib/python3.6/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/usr/lib/python3.6/multiprocessing/process.py", line 93, in run
self._target(*self._args, **self._kwargs)
File "/home/user/.local/lib/python3.6/site-packages/allennlp/data/iterators/multiprocess_iterator.py", line 32, in _create_tensor_dicts
output_queue.put(tensor_dict)
File "/usr/lib/python3.6/multiprocessing/process.py", line 93, in run
self._target(*self._args, **self._kwargs)
File "/home/user/.local/lib/python3.6/site-packages/allennlp/data/iterators/multiprocess_iterator.py", line 32, in _create_tensor_dicts
output_queue.put(tensor_dict)
File "<string>", line 2, in put
File "<string>", line 2, in put
File "/usr/lib/python3.6/multiprocessing/managers.py", line 772, in _callmethod
raise convert_to_error(kind, result)
File "/usr/lib/python3.6/multiprocessing/managers.py", line 772, in _callmethod
raise convert_to_error(kind, result)
multiprocessing.managers.RemoteError:
---------------------------------------------------------------------------
Traceback (most recent call last):
File "/usr/lib/python3.6/multiprocessing/managers.py", line 228, in serve_client
request = recv()
File "/usr/lib/python3.6/multiprocessing/connection.py", line 251, in recv
return _ForkingPickler.loads(buf.getbuffer())
File "/home/user/.local/lib/python3.6/site-packages/torch/multiprocessing/reductions.py", line 276, in rebuild_storage_fd
fd = df.detach()
File "/usr/lib/python3.6/multiprocessing/resource_sharer.py", line 58, in detach
return reduction.recv_handle(conn)
File "/usr/lib/python3.6/multiprocessing/reduction.py", line 182, in recv_handle
return recvfds(s, 1)[0]
File "/usr/lib/python3.6/multiprocessing/reduction.py", line 161, in recvfds
len(ancdata))
RuntimeError: received 0 items of ancdata
---------------------------------------------------------------------------
我还尝试通过 PyTorch's DataParallel 实现这一点，方法是将模型的 context_layer、mention_feedforward、antecedent_feedforward 与自定义 DataParallelWrapper 包装起来（以提供兼容性使用 AllenNLP 假设的 class 函数）。尽管如此，只使用了 1 个 GPU，它最终还是像以前一样耗尽了内存。
class DataParallelWrapper(DataParallel):
def __init__(self, module):
super().__init__(module)
def get_output_dim(self):
return self.module.get_output_dim()
def get_input_dim(self):
return self.module.get_input_dim()
def forward(self, *inputs):
return self.module.forward(inputs)
在深入研究代码后，我发现 AllenNLP 直接通过其 Trainer 在幕后执行此操作。 cuda_device 可以是单个 int（在单处理的情况下）或 list 的 ints（在多处理的情况下）：
cuda_device:Union[int, List[int]], optional (default = -1) An integer or list of integers specifying the CUDA device(s) to use. If -1, the CPU is used.
所以应该传递所有需要的 GPU 设备：
if torch.cuda.is_available():
cuda_device = list(range(torch.cuda.device_count()))
model = model.cuda(cuda_device[0])
else:
cuda_device = -1
请注意，model 仍然需要手动移动到 GPU（通过 model.cuda(...)），否则它会尝试使用多个 CPU。