Permalink
376 lines (320 sloc) 14.6 KB
"""
This is the loadable seq2seq trainer library that is
in charge of training details, loss compute, and statistics.
See train.py for a use case of this library.
Note: To make this a general library, we implement *only*
mechanism things here(i.e. what to do), and leave the strategy
things to users(i.e. how to do it). Also see train.py(one of the
users of this library) for the strategy things we do.
"""
from __future__ import division
import onmt.inputters as inputters
import onmt.utils
from onmt.utils.logging import logger
def build_trainer(opt, device_id, model, fields,
optim, data_type, model_saver=None):
"""
Simplify `Trainer` creation based on user `opt`s*
Args:
opt (:obj:`Namespace`): user options (usually from argument parsing)
model (:obj:`onmt.models.NMTModel`): the model to train
fields (dict): dict of fields
optim (:obj:`onmt.utils.Optimizer`): optimizer used during training
data_type (str): string describing the type of data
e.g. "text", "img", "audio"
model_saver(:obj:`onmt.models.ModelSaverBase`): the utility object
used to save the model
"""
train_loss = onmt.utils.loss.build_loss_compute(
model, fields["tgt"].vocab, opt)
valid_loss = onmt.utils.loss.build_loss_compute(
model, fields["tgt"].vocab, opt, train=False)
trunc_size = opt.truncated_decoder # Badly named...
shard_size = opt.max_generator_batches
norm_method = opt.normalization
grad_accum_count = opt.accum_count
n_gpu = opt.world_size
if device_id >= 0:
gpu_rank = opt.gpu_ranks[device_id]
else:
gpu_rank = 0
n_gpu = 0
gpu_verbose_level = opt.gpu_verbose_level
report_manager = onmt.utils.build_report_manager(opt)
trainer = onmt.Trainer(model, train_loss, valid_loss, optim, trunc_size,
shard_size, data_type, norm_method,
grad_accum_count, n_gpu, gpu_rank,
gpu_verbose_level, report_manager,
model_saver=model_saver)
return trainer
class Trainer(object):
"""
Class that controls the training process.
Args:
model(:py:class:`onmt.models.model.NMTModel`): translation model
to train
train_loss(:obj:`onmt.utils.loss.LossComputeBase`):
training loss computation
valid_loss(:obj:`onmt.utils.loss.LossComputeBase`):
training loss computation
optim(:obj:`onmt.utils.optimizers.Optimizer`):
the optimizer responsible for update
trunc_size(int): length of truncated back propagation through time
shard_size(int): compute loss in shards of this size for efficiency
data_type(string): type of the source input: [text|img|audio]
norm_method(string): normalization methods: [sents|tokens]
grad_accum_count(int): accumulate gradients this many times.
report_manager(:obj:`onmt.utils.ReportMgrBase`):
the object that creates reports, or None
model_saver(:obj:`onmt.models.ModelSaverBase`): the saver is
used to save a checkpoint.
Thus nothing will be saved if this parameter is None
"""
def __init__(self, model, train_loss, valid_loss, optim,
trunc_size=0, shard_size=32, data_type='text',
norm_method="sents", grad_accum_count=1, n_gpu=1, gpu_rank=1,
gpu_verbose_level=0, report_manager=None, model_saver=None):
# Basic attributes.
self.model = model
self.train_loss = train_loss
self.valid_loss = valid_loss
self.optim = optim
self.trunc_size = trunc_size
self.shard_size = shard_size
self.data_type = data_type
self.norm_method = norm_method
self.grad_accum_count = grad_accum_count
self.n_gpu = n_gpu
self.gpu_rank = gpu_rank
self.gpu_verbose_level = gpu_verbose_level
self.report_manager = report_manager
self.model_saver = model_saver
assert grad_accum_count > 0
if grad_accum_count > 1:
assert(self.trunc_size == 0), \
"""To enable accumulated gradients,
you must disable target sequence truncating."""
# Set model in training mode.
self.model.train()
def train(self, train_iter_fct, valid_iter_fct, train_steps, valid_steps):
"""
The main training loops.
by iterating over training data (i.e. `train_iter_fct`)
and running validation (i.e. iterating over `valid_iter_fct`
Args:
train_iter_fct(function): a function that returns the train
iterator. e.g. something like
train_iter_fct = lambda: generator(*args, **kwargs)
valid_iter_fct(function): same as train_iter_fct, for valid data
train_steps(int):
valid_steps(int):
save_checkpoint_steps(int):
Return:
None
"""
logger.info('Start training...')
step = self.optim._step + 1
true_batchs = []
accum = 0
normalization = 0
train_iter = train_iter_fct()
total_stats = onmt.utils.Statistics()
report_stats = onmt.utils.Statistics()
self._start_report_manager(start_time=total_stats.start_time)
while step <= train_steps:
reduce_counter = 0
for i, batch in enumerate(train_iter):
if self.n_gpu == 0 or (i % self.n_gpu == self.gpu_rank):
if self.gpu_verbose_level > 1:
logger.info("GpuRank %d: index: %d accum: %d"
% (self.gpu_rank, i, accum))
true_batchs.append(batch)
if self.norm_method == "tokens":
num_tokens = batch.tgt[1:].ne(
self.train_loss.padding_idx).sum()
normalization += num_tokens.item()
else:
normalization += batch.batch_size
accum += 1
if accum == self.grad_accum_count:
reduce_counter += 1
if self.gpu_verbose_level > 0:
logger.info("GpuRank %d: reduce_counter: %d \
n_minibatch %d"
% (self.gpu_rank, reduce_counter,
len(true_batchs)))
if self.n_gpu > 1:
normalization = sum(onmt.utils.distributed
.all_gather_list
(normalization))
self._gradient_accumulation(
true_batchs, normalization, total_stats,
report_stats)
report_stats = self._maybe_report_training(
step, train_steps,
self.optim.learning_rate,
report_stats)
true_batchs = []
accum = 0
normalization = 0
if (step % valid_steps == 0):
if self.gpu_verbose_level > 0:
logger.info('GpuRank %d: validate step %d'
% (self.gpu_rank, step))
valid_iter = valid_iter_fct()
valid_stats = self.validate(valid_iter)
if self.gpu_verbose_level > 0:
logger.info('GpuRank %d: gather valid stat \
step %d' % (self.gpu_rank, step))
valid_stats = self._maybe_gather_stats(valid_stats)
if self.gpu_verbose_level > 0:
logger.info('GpuRank %d: report stat step %d'
% (self.gpu_rank, step))
self._report_step(self.optim.learning_rate,
step, valid_stats=valid_stats)
if self.gpu_rank == 0:
self._maybe_save(step)
step += 1
if step > train_steps:
break
if self.gpu_verbose_level > 0:
logger.info('GpuRank %d: we completed an epoch \
at step %d' % (self.gpu_rank, step))
train_iter = train_iter_fct()
return total_stats
def validate(self, valid_iter):
""" Validate model.
valid_iter: validate data iterator
Returns:
:obj:`nmt.Statistics`: validation loss statistics
"""
# Set model in validating mode.
self.model.eval()
stats = onmt.utils.Statistics()
for batch in valid_iter:
src = inputters.make_features(batch, 'src', self.data_type)
if self.data_type == 'text':
_, src_lengths = batch.src
elif self.data_type == 'audio':
src_lengths = batch.src_lengths
else:
src_lengths = None
tgt = inputters.make_features(batch, 'tgt')
# F-prop through the model.
outputs, attns = self.model(src, tgt, src_lengths)
# Compute loss.
batch_stats = self.valid_loss.monolithic_compute_loss(
batch, outputs, attns)
# Update statistics.
stats.update(batch_stats)
# Set model back to training mode.
self.model.train()
return stats
def _gradient_accumulation(self, true_batchs, normalization, total_stats,
report_stats):
if self.grad_accum_count > 1:
self.model.zero_grad()
for batch in true_batchs:
target_size = batch.tgt.size(0)
# Truncated BPTT: reminder not compatible with accum > 1
if self.trunc_size:
trunc_size = self.trunc_size
else:
trunc_size = target_size
# dec_state = None
src = inputters.make_features(batch, 'src', self.data_type)
if self.data_type == 'text':
_, src_lengths = batch.src
report_stats.n_src_words += src_lengths.sum().item()
elif self.data_type == 'audio':
src_lengths = batch.src_lengths
else:
src_lengths = None
tgt_outer = inputters.make_features(batch, 'tgt')
for j in range(0, target_size-1, trunc_size):
# 1. Create truncated target.
tgt = tgt_outer[j: j + trunc_size]
# 2. F-prop all but generator.
if self.grad_accum_count == 1:
self.model.zero_grad()
outputs, attns = \
self.model(src, tgt, src_lengths)
# 3. Compute loss in shards for memory efficiency.
batch_stats = self.train_loss.sharded_compute_loss(
batch, outputs, attns, j,
trunc_size, self.shard_size, normalization)
total_stats.update(batch_stats)
report_stats.update(batch_stats)
# 4. Update the parameters and statistics.
if self.grad_accum_count == 1:
# Multi GPU gradient gather
if self.n_gpu > 1:
grads = [p.grad.data for p in self.model.parameters()
if p.requires_grad
and p.grad is not None]
onmt.utils.distributed.all_reduce_and_rescale_tensors(
grads, float(1))
self.optim.step()
# If truncated, don't backprop fully.
# TO CHECK
# if dec_state is not None:
# dec_state.detach()
if self.model.decoder.state is not None:
self.model.decoder.detach_state()
# in case of multi step gradient accumulation,
# update only after accum batches
if self.grad_accum_count > 1:
if self.n_gpu > 1:
grads = [p.grad.data for p in self.model.parameters()
if p.requires_grad
and p.grad is not None]
onmt.utils.distributed.all_reduce_and_rescale_tensors(
grads, float(1))
self.optim.step()
def _start_report_manager(self, start_time=None):
"""
Simple function to start report manager (if any)
"""
if self.report_manager is not None:
if start_time is None:
self.report_manager.start()
else:
self.report_manager.start_time = start_time
def _maybe_gather_stats(self, stat):
"""
Gather statistics in multi-processes cases
Args:
stat(:obj:onmt.utils.Statistics): a Statistics object to gather
or None (it returns None in this case)
Returns:
stat: the updated (or unchanged) stat object
"""
if stat is not None and self.n_gpu > 1:
return onmt.utils.Statistics.all_gather_stats(stat)
return stat
def _maybe_report_training(self, step, num_steps, learning_rate,
report_stats):
"""
Simple function to report training stats (if report_manager is set)
see `onmt.utils.ReportManagerBase.report_training` for doc
"""
if self.report_manager is not None:
return self.report_manager.report_training(
step, num_steps, learning_rate, report_stats,
multigpu=self.n_gpu > 1)
def _report_step(self, learning_rate, step, train_stats=None,
valid_stats=None):
"""
Simple function to report stats (if report_manager is set)
see `onmt.utils.ReportManagerBase.report_step` for doc
"""
if self.report_manager is not None:
return self.report_manager.report_step(
learning_rate, step, train_stats=train_stats,
valid_stats=valid_stats)
def _maybe_save(self, step):
"""
Save the model if a model saver is set
"""
if self.model_saver is not None:
self.model_saver.maybe_save(step)