class Trainer(TrainerBase):
def __init__(
self,
model: Model,
optimizer: torch.optim.Optimizer,
iterator: DataIterator,
train_dataset: Iterable[Instance],
validation_dataset: Optional[Iterable[Instance]] = None,
patience: Optional[int] = None,
validation_metric: str = "-loss",
validation_iterator: DataIterator = None,
shuffle: bool = True,
num_epochs: int = 20,
serialization_dir: Optional[str] = None,
num_serialized_models_to_keep: int = 20,
keep_serialized_model_every_num_seconds: int = None,
checkpointer: Checkpointer = None,
model_save_interval: float = None,
cuda_device: int = -1,
grad_norm: Optional[float] = None,
grad_clipping: Optional[float] = None,
learning_rate_scheduler: Optional[LearningRateScheduler] = None,
momentum_scheduler: Optional[MomentumScheduler] = None,
summary_interval: int = 100,
histogram_interval: int = None,
should_log_parameter_statistics: bool = True,
should_log_learning_rate: bool = False,
log_batch_size_period: Optional[int] = None,
moving_average: Optional[MovingAverage] = None,
distributed: bool = False,
local_rank: int = 0,
world_size: int = 1,
num_gradient_accumulation_steps: int = 1,
) -> None:
"""
A trainer for doing supervised learning. It just takes a labeled dataset
and a `DataIterator`, and uses the supplied `Optimizer` to learn the weights
for your model over some fixed number of epochs. You can also pass in a validation
dataset and enable early stopping. There are many other bells and whistles as well.
# Parameters
model : `Model`, required.
An AllenNLP model to be optimized. Pytorch Modules can also be optimized if
their `forward` method returns a dictionary with a "loss" key, containing a
scalar tensor representing the loss function to be optimized.
If you are training your model using GPUs, your model should already be
on the correct device. (If you use `Trainer.from_params` this will be
handled for you.)
optimizer : `torch.nn.Optimizer`, required.
An instance of a Pytorch Optimizer, instantiated with the parameters of the
model to be optimized.
iterator : `DataIterator`, required.
A method for iterating over a `Dataset`, yielding padded indexed batches.
train_dataset : `Dataset`, required.
A `Dataset` to train on. The dataset should have already been indexed.
validation_dataset : `Dataset`, optional, (default = None).
A `Dataset` to evaluate on. The dataset should have already been indexed.
patience : Optional[int] > 0, optional (default=None)
Number of epochs to be patient before early stopping: the training is stopped
after `patience` epochs with no improvement. If given, it must be `> 0`.
If None, early stopping is disabled.
validation_metric : str, optional (default="loss")
Validation metric to measure for whether to stop training using patience
and whether to serialize an `is_best` model each epoch. The metric name
must be prepended with either "+" or "-", which specifies whether the metric
is an increasing or decreasing function.
validation_iterator : `DataIterator`, optional (default=None)
An iterator to use for the validation set. If `None`, then
use the training `iterator`.
shuffle : `bool`, optional (default=True)
Whether to shuffle the instances in the iterator or not.
num_epochs : int, optional (default = 20)
Number of training epochs.
serialization_dir : str, optional (default=None)
Path to directory for saving and loading model files. Models will not be saved if
this parameter is not passed.
num_serialized_models_to_keep : `int`, optional (default=20)
Number of previous model checkpoints to retain. Default is to keep 20 checkpoints.
A value of None or -1 means all checkpoints will be kept.
keep_serialized_model_every_num_seconds : `int`, optional (default=None)
If num_serialized_models_to_keep is not None, then occasionally it's useful to
save models at a given interval in addition to the last num_serialized_models_to_keep.
To do so, specify keep_serialized_model_every_num_seconds as the number of seconds
between permanently saved checkpoints. Note that this option is only used if
num_serialized_models_to_keep is not None, otherwise all checkpoints are kept.
checkpointer : `Checkpointer`, optional (default=None)
An instance of class Checkpointer to use instead of the default. If a checkpointer is specified,
the arguments num_serialized_models_to_keep and keep_serialized_model_every_num_seconds should
not be specified. The caller is responsible for initializing the checkpointer so that it is
consistent with serialization_dir.
model_save_interval : `float`, optional (default=None)
If provided, then serialize models every `model_save_interval`
seconds within single epochs. In all cases, models are also saved
at the end of every epoch if `serialization_dir` is provided.
cuda_device : `int`, optional (default = -1)
An integer specifying the CUDA device(s) to use for this process. If -1, the CPU is used.
Data parallelism is controlled at the allennlp train level, so each trainer will have a single
GPU.
grad_norm : `float`, optional, (default = None).
If provided, gradient norms will be rescaled to have a maximum of this value.
grad_clipping : `float`, optional (default = `None`).
If provided, gradients will be clipped `during the backward pass` to have an (absolute)
maximum of this value. If you are getting `NaNs` in your gradients during training
that are not solved by using `grad_norm`, you may need this.
learning_rate_scheduler : `LearningRateScheduler`, optional (default = None)
If specified, the learning rate will be decayed with respect to
this schedule at the end of each epoch (or batch, if the scheduler implements
the `step_batch` method). If you use :class:`torch.optim.lr_scheduler.ReduceLROnPlateau`,
this will use the `validation_metric` provided to determine if learning has plateaued.
To support updating the learning rate on every batch, this can optionally implement
`step_batch(batch_num_total)` which updates the learning rate given the batch number.
momentum_scheduler : `MomentumScheduler`, optional (default = None)
If specified, the momentum will be updated at the end of each batch or epoch
according to the schedule.
summary_interval : `int`, optional, (default = 100)
Number of batches between logging scalars to tensorboard
histogram_interval : `int`, optional, (default = `None`)
If not None, then log histograms to tensorboard every `histogram_interval` batches.
When this parameter is specified, the following additional logging is enabled:
* Histograms of model parameters
* The ratio of parameter update norm to parameter norm
* Histogram of layer activations
We log histograms of the parameters returned by
`model.get_parameters_for_histogram_tensorboard_logging`.
The layer activations are logged for any modules in the `Model` that have
the attribute `should_log_activations` set to `True`. Logging
histograms requires a number of GPU-CPU copies during training and is typically
slow, so we recommend logging histograms relatively infrequently.
Note: only Modules that return tensors, tuples of tensors or dicts
with tensors as values currently support activation logging.
should_log_parameter_statistics : `bool`, optional, (default = True)
Whether to send parameter statistics (mean and standard deviation
of parameters and gradients) to tensorboard.
should_log_learning_rate : `bool`, optional, (default = False)
Whether to send parameter specific learning rate to tensorboard.
log_batch_size_period : `int`, optional, (default = `None`)
If defined, how often to log the average batch size.
moving_average : `MovingAverage`, optional, (default = None)
If provided, we will maintain moving averages for all parameters. During training, we
employ a shadow variable for each parameter, which maintains the moving average. During
evaluation, we backup the original parameters and assign the moving averages to corresponding
parameters. Be careful that when saving the checkpoint, we will save the moving averages of
parameters. This is necessary because we want the saved model to perform as well as the validated
model if we load it later. But this may cause problems if you restart the training from checkpoint.
distributed : `bool`, optional, (default = False)
If set, PyTorch's `DistributedDataParallel` is used to train the model in multiple GPUs. This also
requires `world_size` to be greater than 1.
local_rank : `int`, optional, (default = 0)
This is the unique identifier of the `Trainer` in a distributed process group. The GPU device id is
used as the rank.
world_size : `int`, (default = 1)
The number of `Trainer` workers participating in the distributed training.
num_gradient_accumulation_steps : `int`, optional, (default = 1)
Gradients are accumulated for the given number of steps before doing an optimizer step. This can
be useful to accommodate batches that are larger than the RAM size. Refer Thomas Wolf's
[post](https://tinyurl.com/y5mv44fw) for details on Gradient Accumulation.
"""
super().__init__(serialization_dir, cuda_device, distributed,
local_rank, world_size)
# I am not calling move_to_gpu here, because if the model is
# not already on the GPU then the optimizer is going to be wrong.
self.model = model
self.iterator = iterator
self._validation_iterator = validation_iterator
self.shuffle = shuffle
self.optimizer = optimizer
self.train_data = train_dataset
self._validation_data = validation_dataset
if patience is None: # no early stopping
if validation_dataset:
logger.warning(
"You provided a validation dataset but patience was set to None, "
"meaning that early stopping is disabled")
elif (not isinstance(patience, int)) or patience <= 0:
raise ConfigurationError(
'{} is an invalid value for "patience": it must be a positive integer '
"or None (if you want to disable early stopping)".format(
patience))
# For tracking is_best_so_far and should_stop_early
self._metric_tracker = MetricTracker(patience, validation_metric)
# Get rid of + or -
self._validation_metric = validation_metric[1:]
self._num_epochs = num_epochs
if checkpointer is not None:
# We can't easily check if these parameters were passed in, so check against their default values.
# We don't check against serialization_dir since it is also used by the parent class.
if (num_serialized_models_to_keep != 20
or keep_serialized_model_every_num_seconds is not None):
raise ConfigurationError(
"When passing a custom Checkpointer, you may not also pass in separate checkpointer "
"args 'num_serialized_models_to_keep' or 'keep_serialized_model_every_num_seconds'."
)
self._checkpointer = checkpointer
else:
self._checkpointer = Checkpointer(
serialization_dir,
keep_serialized_model_every_num_seconds,
num_serialized_models_to_keep,
)
self._model_save_interval = model_save_interval
self._grad_norm = grad_norm
self._grad_clipping = grad_clipping
self._learning_rate_scheduler = learning_rate_scheduler
self._momentum_scheduler = momentum_scheduler
self._moving_average = moving_average
# We keep the total batch number as an instance variable because it
# is used inside a closure for the hook which logs activations in
# `_enable_activation_logging`.
self._batch_num_total = 0
self._tensorboard = TensorboardWriter(
get_batch_num_total=lambda: self._batch_num_total,
serialization_dir=serialization_dir,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate,
)
self._log_batch_size_period = log_batch_size_period
self._last_log = 0.0 # time of last logging
self._num_gradient_accumulation_steps = num_gradient_accumulation_steps
# Enable activation logging.
if histogram_interval is not None:
self._tensorboard.enable_activation_logging(self.model)
# Using `DistributedDataParallel`(ddp) brings in a quirk wrt AllenNLP's `Model` interface and its
# usage. A `Model` object is wrapped by `ddp`, but assigning the wrapped model to `self.model`
# will break the usages such as `Model.get_regularization_penalty`, `Model.get_metrics`, etc.
#
# Hence a reference to Pytorch's object is maintained in the case of distributed training and in the
# normal case, reference to `Model` is retained. This reference is only used in
# these places: `model.__call__`, `model.train` and `model.eval`.
if self._distributed:
self._pytorch_model = DistributedDataParallel(
self.model,
device_ids=[self.cuda_device],
find_unused_parameters=True)
else:
self._pytorch_model = self.model
def rescale_gradients(self) -> Optional[float]:
return training_util.rescale_gradients(self.model, self._grad_norm)
def batch_loss(self, batch: TensorDict,
for_training: bool) -> torch.Tensor:
"""
Does a forward pass on the given batches and returns the `loss` value in the result.
If `for_training` is `True` also applies regularization penalty.
"""
batch = nn_util.move_to_device(batch, self.cuda_device)
output_dict = self._pytorch_model(**batch)
try:
loss = output_dict["loss"]
if for_training:
loss += self.model.get_regularization_penalty()
except KeyError:
if for_training:
raise RuntimeError(
"The model you are trying to optimize does not contain a"
" 'loss' key in the output of model.forward(inputs).")
loss = None
return loss
def _train_epoch(self, epoch: int) -> Dict[str, float]:
"""
Trains one epoch and returns metrics.
"""
logger.info("Epoch %d/%d", epoch, self._num_epochs - 1)
peak_cpu_usage = peak_memory_mb()
logger.info(f"Peak CPU memory usage MB: {peak_cpu_usage}")
gpu_usage = []
for gpu, memory in gpu_memory_mb().items():
gpu_usage.append((gpu, memory))
logger.info(f"GPU {gpu} memory usage MB: {memory}")
train_loss = 0.0
# Set the model to "train" mode.
self._pytorch_model.train()
# Get tqdm for the training batches
batch_generator = self.iterator(self.train_data,
num_epochs=1,
shuffle=self.shuffle)
batch_group_generator = lazy_groups_of(
batch_generator, self._num_gradient_accumulation_steps)
num_training_batches = math.ceil(
self.iterator.get_num_batches(self.train_data) /
self._num_gradient_accumulation_steps)
# Having multiple tqdm bars in case of distributed training will be a mess. Hence only the master's
# progress is shown
if self._master:
batch_group_generator_tqdm = Tqdm.tqdm(batch_group_generator,
total=num_training_batches)
else:
batch_group_generator_tqdm = batch_group_generator
self._last_log = time.time()
last_save_time = time.time()
batches_this_epoch = 0
if self._batch_num_total is None:
self._batch_num_total = 0
histogram_parameters = set(
self.model.get_parameters_for_histogram_tensorboard_logging())
logger.info("Training")
cumulative_batch_group_size = 0
for batch_group in batch_group_generator_tqdm:
batches_this_epoch += 1
self._batch_num_total += 1
batch_num_total = self._batch_num_total
self.optimizer.zero_grad()
for batch in batch_group:
loss = self.batch_loss(batch, for_training=True)
if torch.isnan(loss):
raise ValueError("nan loss encountered")
loss = loss / len(batch_group)
loss.backward()
train_loss += loss.item()
batch_grad_norm = self.rescale_gradients()
# This does nothing if batch_num_total is None or you are using a
# scheduler which doesn't update per batch.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step_batch(batch_num_total)
if self._momentum_scheduler:
self._momentum_scheduler.step_batch(batch_num_total)
if self._tensorboard.should_log_histograms_this_batch(
) and self._master:
# get the magnitude of parameter updates for logging
# We need a copy of current parameters to compute magnitude of updates,
# and copy them to CPU so large models won't go OOM on the GPU.
param_updates = {
name: param.detach().cpu().clone()
for name, param in self.model.named_parameters()
}
self.optimizer.step()
for name, param in self.model.named_parameters():
param_updates[name].sub_(param.detach().cpu())
update_norm = torch.norm(param_updates[name].view(-1))
param_norm = torch.norm(param.view(-1)).cpu()
self._tensorboard.add_train_scalar(
"gradient_update/" + name,
update_norm / (param_norm + 1e-7))
else:
self.optimizer.step()
# Update moving averages
if self._moving_average is not None:
self._moving_average.apply(batch_num_total)
# Update the description with the latest metrics
metrics = training_util.get_metrics(
self.model,
train_loss,
batches_this_epoch,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
# Updating tqdm only for the master as the trainers wouldn't have one
if self._master:
description = training_util.description_from_metrics(metrics)
batch_group_generator_tqdm.set_description(description,
refresh=False)
# Log parameter values to Tensorboard (only from the master)
if self._tensorboard.should_log_this_batch() and self._master:
self._tensorboard.log_parameter_and_gradient_statistics(
self.model, batch_grad_norm)
self._tensorboard.log_learning_rates(self.model,
self.optimizer)
self._tensorboard.add_train_scalar("loss/loss_train",
metrics["loss"])
self._tensorboard.log_metrics(
{"epoch_metrics/" + k: v
for k, v in metrics.items()})
if self._tensorboard.should_log_histograms_this_batch(
) and self._master:
self._tensorboard.log_histograms(self.model,
histogram_parameters)
if self._log_batch_size_period:
batch_group_size = sum(
training_util.get_batch_size(batch)
for batch in batch_group)
cumulative_batch_group_size += batch_group_size
if (batches_this_epoch - 1) % self._log_batch_size_period == 0:
average = cumulative_batch_group_size / batches_this_epoch
logger.info(
f"current batch size: {batch_group_size} mean batch size: {average}"
)
self._tensorboard.add_train_scalar("current_batch_size",
batch_group_size)
self._tensorboard.add_train_scalar("mean_batch_size",
average)
# Save model if needed.
if (self._model_save_interval is not None and
(time.time() - last_save_time > self._model_save_interval)
and self._master):
last_save_time = time.time()
self._save_checkpoint("{0}.{1}".format(
epoch, training_util.time_to_str(int(last_save_time))))
# Let all workers finish their epoch before computing
# the final statistics for the epoch.
if self._distributed:
dist.barrier()
metrics = training_util.get_metrics(
self.model,
train_loss,
batches_this_epoch,
reset=True,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
metrics["cpu_memory_MB"] = peak_cpu_usage
for (gpu_num, memory) in gpu_usage:
metrics["gpu_" + str(gpu_num) + "_memory_MB"] = memory
return metrics
def _validation_loss(self) -> Tuple[float, int]:
"""
Computes the validation loss. Returns it and the number of batches.
"""
logger.info("Validating")
self._pytorch_model.eval()
# Replace parameter values with the shadow values from the moving averages.
if self._moving_average is not None:
self._moving_average.assign_average_value()
if self._validation_iterator is not None:
val_iterator = self._validation_iterator
else:
val_iterator = self.iterator
val_generator = val_iterator(self._validation_data,
num_epochs=1,
shuffle=False)
num_validation_batches = val_iterator.get_num_batches(
self._validation_data)
val_generator_tqdm = Tqdm.tqdm(val_generator,
total=num_validation_batches)
batches_this_epoch = 0
val_loss = 0
for batch in val_generator_tqdm:
loss = self.batch_loss(batch, for_training=False)
if loss is not None:
# You shouldn't necessarily have to compute a loss for validation, so we allow for
# `loss` to be None. We need to be careful, though - `batches_this_epoch` is
# currently only used as the divisor for the loss function, so we can safely only
# count those batches for which we actually have a loss. If this variable ever
# gets used for something else, we might need to change things around a bit.
batches_this_epoch += 1
val_loss += loss.detach().cpu().numpy()
# Update the description with the latest metrics
val_metrics = training_util.get_metrics(
self.model,
val_loss,
batches_this_epoch,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
description = training_util.description_from_metrics(val_metrics)
val_generator_tqdm.set_description(description, refresh=False)
# Now restore the original parameter values.
if self._moving_average is not None:
self._moving_average.restore()
return val_loss, batches_this_epoch
def train(self) -> Dict[str, Any]:
"""
Trains the supplied model with the supplied parameters.
"""
try:
epoch_counter = self._restore_checkpoint()
except RuntimeError:
traceback.print_exc()
raise ConfigurationError(
"Could not recover training from the checkpoint. Did you mean to output to "
"a different serialization directory or delete the existing serialization "
"directory?")
training_util.enable_gradient_clipping(self.model, self._grad_clipping)
logger.info("Beginning training.")
train_metrics: Dict[str, float] = {}
val_metrics: Dict[str, float] = {}
this_epoch_val_metric: float = None
metrics: Dict[str, Any] = {}
epochs_trained = 0
training_start_time = time.time()
metrics["best_epoch"] = self._metric_tracker.best_epoch
for key, value in self._metric_tracker.best_epoch_metrics.items():
metrics["best_validation_" + key] = value
for epoch in range(epoch_counter, self._num_epochs):
epoch_start_time = time.time()
train_metrics = self._train_epoch(epoch)
# get peak of memory usage
if "cpu_memory_MB" in train_metrics:
metrics["peak_cpu_memory_MB"] = max(
metrics.get("peak_cpu_memory_MB", 0),
train_metrics["cpu_memory_MB"])
for key, value in train_metrics.items():
if key.startswith("gpu_"):
metrics["peak_" + key] = max(metrics.get("peak_" + key, 0),
value)
if self._validation_data is not None:
with torch.no_grad():
# We have a validation set, so compute all the metrics on it.
val_loss, num_batches = self._validation_loss()
# It is safe again to wait till the validation is done. This is
# important to get the metrics right.
if self._distributed:
dist.barrier()
val_metrics = training_util.get_metrics(
self.model,
val_loss,
num_batches,
reset=True,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
# Check validation metric for early stopping
this_epoch_val_metric = val_metrics[
self._validation_metric]
self._metric_tracker.add_metric(this_epoch_val_metric)
if self._metric_tracker.should_stop_early():
logger.info("Ran out of patience. Stopping training.")
break
if self._master:
self._tensorboard.log_metrics(
train_metrics,
val_metrics=val_metrics,
log_to_console=True,
epoch=epoch + 1) # +1 because tensorboard doesn't like 0
# Create overall metrics dict
training_elapsed_time = time.time() - training_start_time
metrics["training_duration"] = str(
datetime.timedelta(seconds=training_elapsed_time))
metrics["training_start_epoch"] = epoch_counter
metrics["training_epochs"] = epochs_trained
metrics["epoch"] = epoch
for key, value in train_metrics.items():
metrics["training_" + key] = value
for key, value in val_metrics.items():
metrics["validation_" + key] = value
if self._metric_tracker.is_best_so_far():
# Update all the best_ metrics.
# (Otherwise they just stay the same as they were.)
metrics["best_epoch"] = epoch
for key, value in val_metrics.items():
metrics["best_validation_" + key] = value
self._metric_tracker.best_epoch_metrics = val_metrics
if self._serialization_dir and self._master:
dump_metrics(
os.path.join(self._serialization_dir,
f"metrics_epoch_{epoch}.json"), metrics)
# The Scheduler API is agnostic to whether your schedule requires a validation metric -
# if it doesn't, the validation metric passed here is ignored.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step(this_epoch_val_metric,
epoch)
if self._momentum_scheduler:
self._momentum_scheduler.step(this_epoch_val_metric, epoch)
if self._master:
self._save_checkpoint(epoch)
# Wait for the master to finish saving the checkpoint
if self._distributed:
dist.barrier()
epoch_elapsed_time = time.time() - epoch_start_time
logger.info("Epoch duration: %s",
datetime.timedelta(seconds=epoch_elapsed_time))
if epoch < self._num_epochs - 1:
training_elapsed_time = time.time() - training_start_time
estimated_time_remaining = training_elapsed_time * (
(self._num_epochs - epoch_counter) /
float(epoch - epoch_counter + 1) - 1)
formatted_time = str(
datetime.timedelta(seconds=int(estimated_time_remaining)))
logger.info("Estimated training time remaining: %s",
formatted_time)
epochs_trained += 1
# make sure pending events are flushed to disk and files are closed properly
self._tensorboard.close()
# Load the best model state before returning
best_model_state = self._checkpointer.best_model_state()
if best_model_state:
self.model.load_state_dict(best_model_state)
return metrics
def _save_checkpoint(self, epoch: Union[int, str]) -> None:
"""
Saves a checkpoint of the model to self._serialization_dir.
Is a no-op if self._serialization_dir is None.
# Parameters
epoch : Union[int, str], required.
The epoch of training. If the checkpoint is saved in the middle
of an epoch, the parameter is a string with the epoch and timestamp.
"""
# If moving averages are used for parameters, we save
# the moving average values into checkpoint, instead of the current values.
if self._moving_average is not None:
self._moving_average.assign_average_value()
# These are the training states we need to persist.
training_states = {
"metric_tracker": self._metric_tracker.state_dict(),
"optimizer": self.optimizer.state_dict(),
"batch_num_total": self._batch_num_total,
}
# If we have a learning rate or momentum scheduler, we should persist them too.
if self._learning_rate_scheduler is not None:
training_states[
"learning_rate_scheduler"] = self._learning_rate_scheduler.state_dict(
)
if self._momentum_scheduler is not None:
training_states[
"momentum_scheduler"] = self._momentum_scheduler.state_dict()
self._checkpointer.save_checkpoint(
model_state=self.model.state_dict(),
epoch=epoch,
training_states=training_states,
is_best_so_far=self._metric_tracker.is_best_so_far(),
)
# Restore the original values for parameters so that training will not be affected.
if self._moving_average is not None:
self._moving_average.restore()
def _restore_checkpoint(self) -> int:
"""
Restores the model and training state from the last saved checkpoint.
This includes an epoch count and optimizer state, which is serialized separately
from model parameters. This function should only be used to continue training -
if you wish to load a model for inference/load parts of a model into a new
computation graph, you should use the native Pytorch functions:
` model.load_state_dict(torch.load("/path/to/model/weights.th"))`
If `self._serialization_dir` does not exist or does not contain any checkpointed weights,
this function will do nothing and return 0.
# Returns
epoch: int
The epoch at which to resume training, which should be one after the epoch
in the saved training state.
"""
model_state, training_state = self._checkpointer.restore_checkpoint()
if not training_state:
# No checkpoint to restore, start at 0
return 0
self.model.load_state_dict(model_state)
self.optimizer.load_state_dict(training_state["optimizer"])
if (self._learning_rate_scheduler is not None
and "learning_rate_scheduler" in training_state):
self._learning_rate_scheduler.load_state_dict(
training_state["learning_rate_scheduler"])
if self._momentum_scheduler is not None and "momentum_scheduler" in training_state:
self._momentum_scheduler.load_state_dict(
training_state["momentum_scheduler"])
training_util.move_optimizer_to_cuda(self.optimizer)
# Currently the `training_state` contains a serialized `MetricTracker`.
if "metric_tracker" in training_state:
self._metric_tracker.load_state_dict(
training_state["metric_tracker"])
# It used to be the case that we tracked `val_metric_per_epoch`.
elif "val_metric_per_epoch" in training_state:
self._metric_tracker.clear()
self._metric_tracker.add_metrics(
training_state["val_metric_per_epoch"])
# And before that we didn't track anything.
else:
self._metric_tracker.clear()
if isinstance(training_state["epoch"], int):
epoch_to_return = training_state["epoch"] + 1
else:
epoch_to_return = int(training_state["epoch"].split(".")[0]) + 1
# For older checkpoints with batch_num_total missing, default to old behavior where
# it is unchanged.
batch_num_total = training_state.get("batch_num_total")
if batch_num_total is not None:
self._batch_num_total = batch_num_total
return epoch_to_return
# Requires custom from_params.
@classmethod
def from_params( # type: ignore
cls,
model: Model,
serialization_dir: str,
iterator: DataIterator,
train_data: Iterable[Instance],
validation_data: Optional[Iterable[Instance]],
params: Params,
validation_iterator: DataIterator = None,
local_rank: int = 0,
) -> "Trainer":
patience = params.pop_int("patience", None)
validation_metric = params.pop("validation_metric", "-loss")
shuffle = params.pop_bool("shuffle", True)
num_epochs = params.pop_int("num_epochs", 20)
cuda_device = parse_cuda_device(params.pop("cuda_device", -1))
grad_norm = params.pop_float("grad_norm", None)
grad_clipping = params.pop_float("grad_clipping", None)
lr_scheduler_params = params.pop("learning_rate_scheduler", None)
momentum_scheduler_params = params.pop("momentum_scheduler", None)
check_for_gpu(cuda_device)
if cuda_device >= 0:
# Moving model to GPU here so that the optimizer state gets constructed on
# the right device.
model = model.cuda(cuda_device)
parameters = [[n, p] for n, p in model.named_parameters()
if p.requires_grad]
optimizer = Optimizer.from_params(parameters, params.pop("optimizer"))
if "moving_average" in params:
moving_average = MovingAverage.from_params(
params.pop("moving_average"), parameters=parameters)
else:
moving_average = None
if lr_scheduler_params:
lr_scheduler = LearningRateScheduler.from_params(
optimizer, lr_scheduler_params)
else:
lr_scheduler = None
if momentum_scheduler_params:
momentum_scheduler = MomentumScheduler.from_params(
optimizer, momentum_scheduler_params)
else:
momentum_scheduler = None
if "checkpointer" in params:
if ("keep_serialized_model_every_num_seconds" in params
or "num_serialized_models_to_keep" in params):
raise ConfigurationError(
"Checkpointer may be initialized either from the 'checkpointer' key or from the "
"keys 'num_serialized_models_to_keep' and 'keep_serialized_model_every_num_seconds'"
" but the passed config uses both methods.")
checkpointer = Checkpointer.from_params(params.pop("checkpointer"))
else:
num_serialized_models_to_keep = params.pop_int(
"num_serialized_models_to_keep", 20)
keep_serialized_model_every_num_seconds = params.pop_int(
"keep_serialized_model_every_num_seconds", None)
checkpointer = Checkpointer(
serialization_dir=serialization_dir,
num_serialized_models_to_keep=num_serialized_models_to_keep,
keep_serialized_model_every_num_seconds=
keep_serialized_model_every_num_seconds,
)
model_save_interval = params.pop_float("model_save_interval", None)
summary_interval = params.pop_int("summary_interval", 100)
histogram_interval = params.pop_int("histogram_interval", None)
should_log_parameter_statistics = params.pop_bool(
"should_log_parameter_statistics", True)
should_log_learning_rate = params.pop_bool("should_log_learning_rate",
False)
log_batch_size_period = params.pop_int("log_batch_size_period", None)
distributed = params.pop_bool("distributed", False)
world_size = params.pop_int("world_size", 1)
num_gradient_accumulation_steps = params.pop(
"num_gradient_accumulation_steps", 1)
params.assert_empty(cls.__name__)
return cls(
model,
optimizer,
iterator,
train_data,
validation_data,
patience=patience,
validation_metric=validation_metric,
validation_iterator=validation_iterator,
shuffle=shuffle,
num_epochs=num_epochs,
serialization_dir=serialization_dir,
cuda_device=cuda_device,
grad_norm=grad_norm,
grad_clipping=grad_clipping,
learning_rate_scheduler=lr_scheduler,
momentum_scheduler=momentum_scheduler,
checkpointer=checkpointer,
model_save_interval=model_save_interval,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate,
log_batch_size_period=log_batch_size_period,
moving_average=moving_average,
distributed=distributed,
local_rank=local_rank,
world_size=world_size,
num_gradient_accumulation_steps=num_gradient_accumulation_steps,
)
class AdvTrainer(TrainerBase):
def __init__(
self,
model: Model,
optimizer: torch.optim.Optimizer,
data_loader: torch.utils.data.DataLoader,
adv_policy: adv_utils.AdvTrainingPolicy,
patience: Optional[int] = None,
validation_metric: str = "-loss",
validation_data_loader: torch.utils.data.DataLoader = None,
num_epochs: int = 20,
serialization_dir: Optional[str] = None,
num_serialized_models_to_keep: int = 20,
keep_serialized_model_every_num_seconds: int = None,
checkpointer: Checkpointer = None,
model_save_interval: float = None,
cuda_device: int = -1,
grad_norm: Optional[float] = None,
grad_clipping: Optional[float] = None,
learning_rate_scheduler: Optional[LearningRateScheduler] = None,
momentum_scheduler: Optional[MomentumScheduler] = None,
summary_interval: int = 100,
histogram_interval: int = None,
should_log_parameter_statistics: bool = True,
should_log_learning_rate: bool = False,
log_batch_size_period: Optional[int] = None,
moving_average: Optional[MovingAverage] = None,
batch_callbacks: List[BatchCallback] = None,
epoch_callbacks: List[EpochCallback] = None,
distributed: bool = False,
local_rank: int = 0,
world_size: int = 1,
opt_level: Optional[str] = None,
) -> None:
"""
A trainer for doing supervised learning. It just takes a labeled dataset
and a `DataLoader`, and uses the supplied `Optimizer` to learn the weights
for your model over some fixed number of epochs. You can also pass in a validation
dataloader and enable early stopping. There are many other bells and whistles as well.
# Parameters
model : `Model`, required.
An AllenNLP model to be optimized. Pytorch Modules can also be optimized if
their `forward` method returns a dictionary with a "loss" key, containing a
scalar tensor representing the loss function to be optimized.
If you are training your model using GPUs, your model should already be
on the correct device. (If you use `Trainer.from_params` this will be
handled for you.)
optimizer : `torch.nn.Optimizer`, required.
An instance of a Pytorch Optimizer, instantiated with the parameters of the
model to be optimized.
data_loader : `DataLoader`, required.
A pytorch `DataLoader` containing your `Dataset`, yielding padded indexed batches.
patience : Optional[int] > 0, optional (default=None)
Number of epochs to be patient before early stopping: the training is stopped
after `patience` epochs with no improvement. If given, it must be `> 0`.
If None, early stopping is disabled.
validation_metric : str, optional (default="loss")
Validation metric to measure for whether to stop training using patience
and whether to serialize an `is_best` model each epoch. The metric name
must be prepended with either "+" or "-", which specifies whether the metric
is an increasing or decreasing function.
validation_dataloader : `DataLoader`, optional (default=None)
A `DataLoader` to use for the validation set. If `None`, then
use the training `DataLoader` with the validation data.
num_epochs : int, optional (default = 20)
Number of training epochs.
serialization_dir : str, optional (default=None)
Path to directory for saving and loading model files. Models will not be saved if
this parameter is not passed.
num_serialized_models_to_keep : `int`, optional (default=20)
Number of previous model checkpoints to retain. Default is to keep 20 checkpoints.
A value of None or -1 means all checkpoints will be kept.
keep_serialized_model_every_num_seconds : `int`, optional (default=None)
If num_serialized_models_to_keep is not None, then occasionally it's useful to
save models at a given interval in addition to the last num_serialized_models_to_keep.
To do so, specify keep_serialized_model_every_num_seconds as the number of seconds
between permanently saved checkpoints. Note that this option is only used if
num_serialized_models_to_keep is not None, otherwise all checkpoints are kept.
checkpointer : `Checkpointer`, optional (default=None)
An instance of class Checkpointer to use instead of the default. If a checkpointer is specified,
the arguments num_serialized_models_to_keep and keep_serialized_model_every_num_seconds should
not be specified. The caller is responsible for initializing the checkpointer so that it is
consistent with serialization_dir.
model_save_interval : `float`, optional (default=None)
If provided, then serialize models every `model_save_interval`
seconds within single epochs. In all cases, models are also saved
at the end of every epoch if `serialization_dir` is provided.
cuda_device : `int`, optional (default = -1)
An integer specifying the CUDA device(s) to use for this process. If -1, the CPU is used.
Data parallelism is controlled at the allennlp train level, so each trainer will have a single
GPU.
grad_norm : `float`, optional, (default = None).
If provided, gradient norms will be rescaled to have a maximum of this value.
grad_clipping : `float`, optional (default = `None`).
If provided, gradients will be clipped `during the backward pass` to have an (absolute)
maximum of this value. If you are getting `NaNs` in your gradients during training
that are not solved by using `grad_norm`, you may need this.
learning_rate_scheduler : `LearningRateScheduler`, optional (default = None)
If specified, the learning rate will be decayed with respect to
this schedule at the end of each epoch (or batch, if the scheduler implements
the `step_batch` method). If you use `torch.optim.lr_scheduler.ReduceLROnPlateau`,
this will use the `validation_metric` provided to determine if learning has plateaued.
To support updating the learning rate on every batch, this can optionally implement
`step_batch(batch_num_total)` which updates the learning rate given the batch number.
momentum_scheduler : `MomentumScheduler`, optional (default = None)
If specified, the momentum will be updated at the end of each batch or epoch
according to the schedule.
summary_interval : `int`, optional, (default = 100)
Number of batches between logging scalars to tensorboard
histogram_interval : `int`, optional, (default = `None`)
If not None, then log histograms to tensorboard every `histogram_interval` batches.
When this parameter is specified, the following additional logging is enabled:
* Histograms of model parameters
* The ratio of parameter update norm to parameter norm
* Histogram of layer activations
We log histograms of the parameters returned by
`model.get_parameters_for_histogram_tensorboard_logging`.
The layer activations are logged for any modules in the `Model` that have
the attribute `should_log_activations` set to `True`. Logging
histograms requires a number of GPU-CPU copies during training and is typically
slow, so we recommend logging histograms relatively infrequently.
Note: only Modules that return tensors, tuples of tensors or dicts
with tensors as values currently support activation logging.
should_log_parameter_statistics : `bool`, optional, (default = True)
Whether to send parameter statistics (mean and standard deviation
of parameters and gradients) to tensorboard.
should_log_learning_rate : `bool`, optional, (default = False)
Whether to send parameter specific learning rate to tensorboard.
log_batch_size_period : `int`, optional, (default = `None`)
If defined, how often to log the average batch size.
moving_average : `MovingAverage`, optional, (default = None)
If provided, we will maintain moving averages for all parameters. During training, we
employ a shadow variable for each parameter, which maintains the moving average. During
evaluation, we backup the original parameters and assign the moving averages to corresponding
parameters. Be careful that when saving the checkpoint, we will save the moving averages of
parameters. This is necessary because we want the saved model to perform as well as the validated
model if we load it later. But this may cause problems if you restart the training from checkpoint.
distributed : `bool`, optional, (default = False)
If set, PyTorch's `DistributedDataParallel` is used to train the model in multiple GPUs. This also
requires `world_size` to be greater than 1.
local_rank : `int`, optional, (default = 0)
This is the unique identifier of the `Trainer` in a distributed process group. The GPU device id is
used as the rank.
world_size : `int`, (default = 1)
The number of `Trainer` workers participating in the distributed training.
opt_level : `str`, optional, (default = `None`)
Each opt_level establishes a set of properties that govern Amp’s implementation of pure or mixed
precision training. Must be a choice of `"O0"`, `"O1"`, `"O2"`, or `"O3"`.
See the Apex [documentation](https://nvidia.github.io/apex/amp.html#opt-levels-and-properties) for
more details. If `None`, Amp is not used. Defaults to `None`.
"""
super().__init__(serialization_dir, cuda_device, distributed,
local_rank, world_size)
# I am not calling move_to_gpu here, because if the model is
# not already on the GPU then the optimizer is going to be wrong.
self.model = model
self.data_loader = data_loader
self._validation_data_loader = validation_data_loader
self.optimizer = optimizer
self.adv_policy = adv_policy
if patience is None: # no early stopping
if validation_data_loader:
logger.warning(
"You provided a validation dataset but patience was set to None, "
"meaning that early stopping is disabled")
elif (not isinstance(patience, int)) or patience <= 0:
raise ConfigurationError(
'{} is an invalid value for "patience": it must be a positive integer '
"or None (if you want to disable early stopping)".format(
patience))
# For tracking is_best_so_far and should_stop_early
self._metric_tracker = MetricTracker(patience, validation_metric)
# Get rid of + or -
self._validation_metric = validation_metric[1:]
self._num_epochs = num_epochs
if checkpointer is not None:
# We can't easily check if these parameters were passed in, so check against their default values.
# We don't check against serialization_dir since it is also used by the parent class.
if (num_serialized_models_to_keep != 20
or keep_serialized_model_every_num_seconds is not None):
raise ConfigurationError(
"When passing a custom Checkpointer, you may not also pass in separate checkpointer "
"args 'num_serialized_models_to_keep' or 'keep_serialized_model_every_num_seconds'."
)
self._checkpointer = checkpointer
else:
self._checkpointer = Checkpointer(
serialization_dir,
keep_serialized_model_every_num_seconds,
num_serialized_models_to_keep,
)
self._model_save_interval = model_save_interval
self._grad_norm = grad_norm
self._grad_clipping = grad_clipping
self._learning_rate_scheduler = learning_rate_scheduler
self._momentum_scheduler = momentum_scheduler
self._moving_average = moving_average
self._batch_callbacks = batch_callbacks or []
self._epoch_callbacks = epoch_callbacks or []
# We keep the total batch number as an instance variable because it
# is used inside a closure for the hook which logs activations in
# `_enable_activation_logging`.
self._batch_num_total = 0
self._tensorboard = TensorboardWriter(
get_batch_num_total=lambda: self._batch_num_total,
serialization_dir=serialization_dir,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate,
)
self._log_batch_size_period = log_batch_size_period
self._last_log = 0.0 # time of last logging
# Enable activation logging.
if histogram_interval is not None:
self._tensorboard.enable_activation_logging(self.model)
# Enable automatic mixed precision training with NVIDIA Apex.
self._opt_level = opt_level
if self._opt_level is not None:
if amp is None:
raise ConfigurationError((
"Apex not installed but opt_level was provided. Please install NVIDIA's Apex to enable"
" automatic mixed precision (AMP) training. See: https://github.com/NVIDIA/apex."
))
self.model, self.optimizer = amp.initialize(
self.model, self.optimizer, opt_level=self._opt_level)
# Using `DistributedDataParallel`(ddp) brings in a quirk wrt AllenNLP's `Model` interface and its
# usage. A `Model` object is wrapped by `ddp`, but assigning the wrapped model to `self.model`
# will break the usages such as `Model.get_regularization_penalty`, `Model.get_metrics`, etc.
#
# Hence a reference to Pytorch's object is maintained in the case of distributed training and in the
# normal case, reference to `Model` is retained. This reference is only used in
# these places: `model.__call__`, `model.train` and `model.eval`.
if self._distributed:
self._pytorch_model = DistributedDataParallel(
self.model,
device_ids=[self.cuda_device],
find_unused_parameters=True)
else:
self._pytorch_model = self.model
def rescale_gradients(self) -> Optional[float]:
"""
Performs gradient rescaling. Is a no-op if gradient rescaling is not enabled.
"""
if self._grad_norm:
if self._opt_level is not None:
# See: https://nvidia.github.io/apex/advanced.html#gradient-clipping
parameters_to_clip = [
p for p in amp.master_params(self.optimizer)
if p.grad is not None
]
else:
parameters_to_clip = [
p for p in self.model.parameters() if p.grad is not None
]
return training_util.sparse_clip_norm(parameters_to_clip,
self._grad_norm)
else:
return None
def batch_loss(self, batch: TensorDict,
for_training: bool) -> torch.Tensor:
"""
Does a forward pass on the given batches and returns the `loss` value in the result.
If `for_training` is `True` also applies regularization penalty.
"""
adv_utils.reset_embedding_hook()
batch = nn_util.move_to_device(batch, self.cuda_device)
output_dict = self._pytorch_model(**batch)
try:
loss = output_dict["loss"]
if for_training and self.model.get_regularization_penalty():
loss += self.model.get_regularization_penalty()
except KeyError:
if for_training:
raise RuntimeError(
"The model you are trying to optimize does not contain a"
" 'loss' key in the output of model.forward(inputs).")
loss = None
return loss
def _train_epoch(self, epoch: int) -> Dict[str, float]:
"""
Trains one epoch and returns metrics.
"""
logger.info("Epoch %d/%d", epoch, self._num_epochs - 1)
# peak_cpu_usage = common_util.peak_memory_mb()
# logger.info(f"Peak CPU memory usage MB: {peak_cpu_usage}")
gpu_usage = []
for gpu, memory in common_util.peak_gpu_memory().items():
gpu_usage.append((gpu, memory))
logger.info(f"GPU {gpu} memory usage MB: {memory}")
train_loss = 0.0
# Set the model to "train" mode.
self._pytorch_model.train()
if isinstance(
self.adv_policy,
(adv_utils.HotFlipPolicy, adv_utils.RandomNeighbourPolicy)):
hooks = adv_utils.register_embedding_hook(self.model)
embedding_matrix = util.find_embedding_layer(self.model).weight
# Get tqdm for the training batches
batch_generator = iter(self.data_loader)
logger.info("Training")
num_training_batches = len(self.data_loader)
# Having multiple tqdm bars in case of distributed training will be a mess. Hence only the master's
# progress is shown
if self._master:
batch_generator_tqdm = Tqdm.tqdm(batch_generator,
total=num_training_batches)
else:
batch_generator_tqdm = batch_generator
self._last_log = time.time()
last_save_time = time.time()
batches_this_epoch = 0
if self._batch_num_total is None:
self._batch_num_total = 0
histogram_parameters = set(
self.model.get_parameters_for_histogram_tensorboard_logging())
cumulative_batch_group_size = 0
done_early = False
for batch in batch_generator_tqdm:
if self._distributed:
# Check whether the other workers have stopped already (due to differing amounts of
# data in each). If so, we can't proceed because we would hang when we hit the
# barrier implicit in Model.forward. We use a IntTensor instead a BoolTensor
# here because NCCL process groups apparently don't support BoolTensor.
done = torch.tensor(0, device=self.cuda_device)
torch.distributed.all_reduce(done,
torch.distributed.ReduceOp.SUM)
if done.item() > 0:
done_early = True
logger.warning(
f"Worker {torch.distributed.get_rank()} finishing training early! "
"This implies that there is an imbalance in your training "
"data across the workers and that some amount of it will be "
"ignored. A small amount of this is fine, but a major imbalance "
"should be avoided. Note: This warning will appear unless your "
"data is perfectly balanced.")
break
batches_this_epoch += 1
self._batch_num_total += 1
batch_num_total = self._batch_num_total
self.optimizer.zero_grad()
# normal samples
loss = self.batch_loss(batch, for_training=True)
if torch.isnan(loss):
raise ValueError("nan loss encountered")
if self._opt_level is not None:
with amp.scale_loss(loss, self.optimizer) as scaled_loss:
scaled_loss.backward()
else:
loss.backward()
train_loss += loss.item()
# adversarial samples
if self.adv_policy.adv_iteration > 0:
adv_utils.set_adv_mode(True)
adv_fields = batch[self.adv_policy.adv_field]
token_key = adv_utils.guess_token_key_from_field(adv_fields)
raw_tokens = adv_fields['tokens'][token_key].cuda()
adv_tokens = raw_tokens.clone()
for adv_idx in range(self.adv_policy.adv_iteration):
if isinstance(self.adv_policy, adv_utils.HotFlipPolicy):
fw, bw = adv_utils.read_embedding_hook(
self.adv_policy.forward_order)
adv_tokens = adv_utils.hotflip(
raw_tokens=raw_tokens,
adv_tokens=adv_tokens,
embeds=fw,
grads=bw,
embedding_matrix=embedding_matrix,
searcher=self.adv_policy.searcher,
replace_num=self.adv_policy.replace_num,
)
elif isinstance(self.adv_policy,
adv_utils.RandomNeighbourPolicy):
adv_tokens = adv_utils.random_swap(
raw_tokens=raw_tokens,
adv_tokens=adv_tokens,
searcher=self.adv_policy.searcher,
replace_num=self.adv_policy.replace_num,
)
elif isinstance(self.adv_policy,
adv_utils.DoItYourselfPolicy):
adv_utils.send("step", self.adv_policy.step)
else:
raise Exception
adv_fields['tokens'][token_key] = adv_tokens
loss = self.batch_loss(
batch,
for_training=True) / self.adv_policy.adv_iteration
if torch.isnan(loss):
raise ValueError("nan loss encountered")
if self._opt_level is not None:
with amp.scale_loss(loss,
self.optimizer) as scaled_loss:
scaled_loss.backward()
else:
loss.backward()
train_loss += loss.item()
adv_utils.set_adv_mode(False)
batch_grad_norm = self.rescale_gradients()
torch.cuda.empty_cache()
# This does nothing if batch_num_total is None or you are using a
# scheduler which doesn't update per batch.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step_batch(batch_num_total)
if self._momentum_scheduler:
self._momentum_scheduler.step_batch(batch_num_total)
if self._tensorboard.should_log_histograms_this_batch(
) and self._master:
# get the magnitude of parameter updates for logging
# We need a copy of current parameters to compute magnitude of updates,
# and copy them to CPU so large models won't go OOM on the GPU.
param_updates = {
name: param.detach().cpu().clone()
for name, param in self.model.named_parameters()
}
self.optimizer.step()
for name, param in self.model.named_parameters():
param_updates[name].sub_(param.detach().cpu())
update_norm = torch.norm(param_updates[name].view(-1))
param_norm = torch.norm(param.view(-1)).cpu()
self._tensorboard.add_train_scalar(
"gradient_update/" + name,
update_norm / (param_norm + 1e-7))
else:
self.optimizer.step()
# Update moving averages
if self._moving_average is not None:
self._moving_average.apply(batch_num_total)
# Update the description with the latest metrics
metrics = training_util.get_metrics(
self.model,
train_loss,
batches_this_epoch,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
# Updating tqdm only for the master as the trainers wouldn't have one
if self._master:
description = training_util.description_from_metrics(metrics)
batch_generator_tqdm.set_description(description,
refresh=False)
if self._master:
for callback in self._batch_callbacks:
callback(
self,
epoch,
batches_this_epoch,
is_training=True,
)
# Log parameter values to Tensorboard (only from the master)
if self._tensorboard.should_log_this_batch() and self._master:
self._tensorboard.log_parameter_and_gradient_statistics(
self.model, batch_grad_norm)
self._tensorboard.log_learning_rates(self.model,
self.optimizer)
self._tensorboard.add_train_scalar("loss/loss_train",
metrics["loss"])
self._tensorboard.log_metrics(
{"epoch_metrics/" + k: v
for k, v in metrics.items()})
if self._tensorboard.should_log_histograms_this_batch(
) and self._master:
self._tensorboard.log_histograms(self.model,
histogram_parameters)
if self._log_batch_size_period:
batch_group_size = sum(
training_util.get_batch_size(batch)
for batch in batch_group)
cumulative_batch_group_size += batch_group_size
if (batches_this_epoch - 1) % self._log_batch_size_period == 0:
average = cumulative_batch_group_size / batches_this_epoch
logger.info(
f"current batch size: {batch_group_size} mean batch size: {average}"
)
self._tensorboard.add_train_scalar("current_batch_size",
batch_group_size)
self._tensorboard.add_train_scalar("mean_batch_size",
average)
# Save model if needed.
if (self._model_save_interval is not None and
(time.time() - last_save_time > self._model_save_interval)
and self._master):
last_save_time = time.time()
self._save_checkpoint("{0}.{1}".format(
epoch, training_util.time_to_str(int(last_save_time))))
if self._distributed and not done_early:
logger.warning(
f"Worker {torch.distributed.get_rank()} completed its entire epoch (training)."
)
# Indicate that we're done so that any workers that have remaining data stop the epoch early.
done = torch.tensor(1, device=self.cuda_device)
torch.distributed.all_reduce(done, torch.distributed.ReduceOp.SUM)
assert done.item()
# Let all workers finish their epoch before computing
# the final statistics for the epoch.
if self._distributed:
dist.barrier()
if isinstance(
self.adv_policy,
(adv_utils.HotFlipPolicy, adv_utils.RandomNeighbourPolicy)):
for hook in hooks:
hook.remove()
metrics = training_util.get_metrics(
self.model,
train_loss,
batches_this_epoch,
reset=True,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
# metrics["cpu_memory_MB"] = peak_cpu_usage
for (gpu_num, memory) in gpu_usage:
metrics["gpu_" + str(gpu_num) + "_memory_MB"] = memory
return metrics
def _validation_loss(self) -> Tuple[float, int]:
"""
Computes the validation loss. Returns it and the number of batches.
"""
logger.info("Validating")
self._pytorch_model.eval()
# Replace parameter values with the shadow values from the moving averages.
if self._moving_average is not None:
self._moving_average.assign_average_value()
if self._validation_data_loader is not None:
validation_data_loader = self._validation_data_loader
else:
raise ConfigurationError(
"Validation results cannot be calculated without a validation_data_loader"
)
val_generator_tqdm = Tqdm.tqdm(validation_data_loader)
batches_this_epoch = 0
val_loss = 0
done_early = False
for batch in val_generator_tqdm:
if self._distributed:
# Check whether the other workers have stopped already (due to differing amounts of
# data in each). If so, we can't proceed because we would hang when we hit the
# barrier implicit in Model.forward. We use a IntTensor instead a BoolTensor
# here because NCCL process groups apparently don't support BoolTensor.
done = torch.tensor(0, device=self.cuda_device)
torch.distributed.all_reduce(done,
torch.distributed.ReduceOp.SUM)
if done.item() > 0:
done_early = True
logger.warning(
f"Worker {torch.distributed.get_rank()} finishing validation early! "
"This implies that there is an imbalance in your validation "
"data across the workers and that some amount of it will be "
"ignored. A small amount of this is fine, but a major imbalance "
"should be avoided. Note: This warning will appear unless your "
"data is perfectly balanced.")
break
loss = self.batch_loss(batch, for_training=False)
if loss is not None:
# You shouldn't necessarily have to compute a loss for validation, so we allow for
# `loss` to be None. We need to be careful, though - `batches_this_epoch` is
# currently only used as the divisor for the loss function, so we can safely only
# count those batches for which we actually have a loss. If this variable ever
# gets used for something else, we might need to change things around a bit.
batches_this_epoch += 1
val_loss += loss.detach().cpu().numpy()
# Update the description with the latest metrics
val_metrics = training_util.get_metrics(
self.model,
val_loss,
batches_this_epoch,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
description = training_util.description_from_metrics(val_metrics)
val_generator_tqdm.set_description(description, refresh=False)
if self._distributed and not done_early:
logger.warning(
f"Worker {torch.distributed.get_rank()} completed its entire epoch (validation)."
)
# Indicate that we're done so that any workers that have remaining data stop validation early.
done = torch.tensor(1, device=self.cuda_device)
torch.distributed.all_reduce(done, torch.distributed.ReduceOp.SUM)
assert done.item()
# Now restore the original parameter values.
if self._moving_average is not None:
self._moving_average.restore()
return val_loss, batches_this_epoch
def train(self) -> Dict[str, Any]:
"""
Trains the supplied model with the supplied parameters.
"""
try:
epoch_counter = self._restore_checkpoint()
logger.warning(
f"Restore from epoch {epoch_counter} in checkpoint...")
except RuntimeError:
traceback.print_exc()
raise ConfigurationError(
"Could not recover training from the checkpoint. Did you mean to output to "
"a different serialization directory or delete the existing serialization "
"directory?")
training_util.enable_gradient_clipping(self.model, self._grad_clipping)
logger.info("Beginning training.")
val_metrics: Dict[str, float] = {}
this_epoch_val_metric: float = None
metrics: Dict[str, Any] = {}
epochs_trained = 0
training_start_time = time.time()
metrics["best_epoch"] = self._metric_tracker.best_epoch
for key, value in self._metric_tracker.best_epoch_metrics.items():
metrics["best_validation_" + key] = value
for callback in self._epoch_callbacks:
callback(self,
metrics={},
epoch=-1 if epoch_counter == 0 else epoch_counter)
for epoch in range(epoch_counter, self._num_epochs):
epoch_start_time = time.time()
train_metrics = self._train_epoch(epoch)
# get peak of memory usage
if "cpu_memory_MB" in train_metrics:
metrics["peak_cpu_memory_MB"] = max(
metrics.get("peak_cpu_memory_MB", 0),
train_metrics["cpu_memory_MB"])
for key, value in train_metrics.items():
if key.startswith("gpu_"):
metrics["peak_" + key] = max(metrics.get("peak_" + key, 0),
value)
if self._validation_data_loader is not None:
with torch.no_grad():
# We have a validation set, so compute all the metrics on it.
val_loss, num_batches = self._validation_loss()
# It is safe again to wait till the validation is done. This is
# important to get the metrics right.
if self._distributed:
dist.barrier()
val_metrics = training_util.get_metrics(
self.model,
val_loss,
num_batches,
reset=True,
world_size=self._world_size,
cuda_device=[self.cuda_device],
)
# Check validation metric for early stopping
this_epoch_val_metric = val_metrics[
self._validation_metric]
self._metric_tracker.add_metric(this_epoch_val_metric)
if self._metric_tracker.should_stop_early():
logger.info("Ran out of patience. Stopping training.")
break
if self._master:
self._tensorboard.log_metrics(
train_metrics,
val_metrics=val_metrics,
log_to_console=True,
epoch=epoch + 1) # +1 because tensorboard doesn't like 0
# Create overall metrics dict
training_elapsed_time = time.time() - training_start_time
metrics["training_duration"] = str(
datetime.timedelta(seconds=training_elapsed_time))
metrics["training_start_epoch"] = epoch_counter
metrics["training_epochs"] = epochs_trained
metrics["epoch"] = epoch
for key, value in train_metrics.items():
metrics["training_" + key] = value
for key, value in val_metrics.items():
metrics["validation_" + key] = value
if self._metric_tracker.is_best_so_far():
# Update all the best_ metrics.
# (Otherwise they just stay the same as they were.)
metrics["best_epoch"] = epoch
for key, value in val_metrics.items():
metrics["best_validation_" + key] = value
self._metric_tracker.best_epoch_metrics = val_metrics
if self._serialization_dir and self._master:
common_util.dump_metrics(
os.path.join(self._serialization_dir,
f"metrics_epoch_{epoch}.json"), metrics)
# The Scheduler API is agnostic to whether your schedule requires a validation metric -
# if it doesn't, the validation metric passed here is ignored.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step(this_epoch_val_metric)
if self._momentum_scheduler:
self._momentum_scheduler.step(this_epoch_val_metric)
if self._master:
self._save_checkpoint(epoch)
# Wait for the master to finish saving the checkpoint
if self._distributed:
dist.barrier()
for callback in self._epoch_callbacks:
callback(self, metrics=metrics, epoch=epoch)
epoch_elapsed_time = time.time() - epoch_start_time
logger.info("Epoch duration: %s",
datetime.timedelta(seconds=epoch_elapsed_time))
if epoch < self._num_epochs - 1:
training_elapsed_time = time.time() - training_start_time
estimated_time_remaining = training_elapsed_time * (
(self._num_epochs - epoch_counter) /
float(epoch - epoch_counter + 1) - 1)
formatted_time = str(
datetime.timedelta(seconds=int(estimated_time_remaining)))
logger.info("Estimated training time remaining: %s",
formatted_time)
epochs_trained += 1
# make sure pending events are flushed to disk and files are closed properly
self._tensorboard.close()
# Load the best model state before returning
best_model_state = self._checkpointer.best_model_state()
if best_model_state:
self.model.load_state_dict(best_model_state)
return metrics
def _save_checkpoint(self, epoch: Union[int, str]) -> None:
"""
Saves a checkpoint of the model to self._serialization_dir.
Is a no-op if self._serialization_dir is None.
# Parameters
epoch : Union[int, str], required.
The epoch of training. If the checkpoint is saved in the middle
of an epoch, the parameter is a string with the epoch and timestamp.
"""
# If moving averages are used for parameters, we save
# the moving average values into checkpoint, instead of the current values.
if self._moving_average is not None:
self._moving_average.assign_average_value()
# These are the training states we need to persist.
training_states = {
"metric_tracker": self._metric_tracker.state_dict(),
"optimizer": self.optimizer.state_dict(),
"batch_num_total": self._batch_num_total,
}
# If we have a learning rate or momentum scheduler, we should persist them too.
if self._learning_rate_scheduler is not None:
training_states[
"learning_rate_scheduler"] = self._learning_rate_scheduler.state_dict(
)
if self._momentum_scheduler is not None:
training_states[
"momentum_scheduler"] = self._momentum_scheduler.state_dict()
self._checkpointer.save_checkpoint(
model_state=self.model.state_dict(),
epoch=epoch,
training_states=training_states,
is_best_so_far=self._metric_tracker.is_best_so_far(),
)
# Restore the original values for parameters so that training will not be affected.
if self._moving_average is not None:
self._moving_average.restore()
def _restore_checkpoint(self) -> int:
"""
Restores the model and training state from the last saved checkpoint.
This includes an epoch count and optimizer state, which is serialized separately
from model parameters. This function should only be used to continue training -
if you wish to load a model for inference/load parts of a model into a new
computation graph, you should use the native Pytorch functions:
` model.load_state_dict(torch.load("/path/to/model/weights.th"))`
If `self._serialization_dir` does not exist or does not contain any checkpointed weights,
this function will do nothing and return 0.
# Returns
epoch: int
The epoch at which to resume training, which should be one after the epoch
in the saved training state.
"""
model_state, training_state = self._checkpointer.restore_checkpoint()
if not training_state:
# No checkpoint to restore, start at 0
return 0
self.model.load_state_dict(model_state)
self.optimizer.load_state_dict(training_state["optimizer"])
if (self._learning_rate_scheduler is not None
and "learning_rate_scheduler" in training_state):
self._learning_rate_scheduler.load_state_dict(
training_state["learning_rate_scheduler"])
if self._momentum_scheduler is not None and "momentum_scheduler" in training_state:
self._momentum_scheduler.load_state_dict(
training_state["momentum_scheduler"])
training_util.move_optimizer_to_cuda(self.optimizer)
# Currently the `training_state` contains a serialized `MetricTracker`.
if "metric_tracker" in training_state:
self._metric_tracker.load_state_dict(
training_state["metric_tracker"])
# It used to be the case that we tracked `val_metric_per_epoch`.
elif "val_metric_per_epoch" in training_state:
self._metric_tracker.clear()
self._metric_tracker.add_metrics(
training_state["val_metric_per_epoch"])
# And before that we didn't track anything.
else:
self._metric_tracker.clear()
if isinstance(training_state["epoch"], int):
epoch_to_return = training_state["epoch"] + 1
else:
epoch_to_return = int(training_state["epoch"].split(".")[0]) + 1
# For older checkpoints with batch_num_total missing, default to old behavior where
# it is unchanged.
batch_num_total = training_state.get("batch_num_total")
if batch_num_total is not None:
self._batch_num_total = batch_num_total
return epoch_to_return
@classmethod
def from_partial_objects(
cls,
model: Model,
serialization_dir: str,
data_loader: DataLoader,
validation_data_loader: DataLoader = None,
local_rank: int = 0,
patience: int = None,
validation_metric: str = "-loss",
num_epochs: int = 20,
cuda_device: int = -1,
grad_norm: float = None,
grad_clipping: float = None,
model_save_interval: float = None,
summary_interval: int = 100,
histogram_interval: int = None,
should_log_parameter_statistics: bool = True,
should_log_learning_rate: bool = False,
log_batch_size_period: int = None,
distributed: bool = None,
world_size: int = 1,
opt_level: Optional[str] = None,
no_grad: List[str] = None,
optimizer: Lazy[Optimizer] = None,
learning_rate_scheduler: Lazy[LearningRateScheduler] = None,
momentum_scheduler: Lazy[MomentumScheduler] = None,
moving_average: Lazy[MovingAverage] = None,
checkpointer: Lazy[Checkpointer] = None,
) -> "Trainer":
"""
This method exists so that we can have a documented method to construct this class using
`FromParams`. If you are not using `FromParams` or config files, you can safely ignore this
method.
The reason we can't just use `__init__` with `FromParams` here is because there are
sequential dependencies to this class's arguments. Anything that has a `Lazy[]` type
annotation needs something from one of the non-`Lazy` arguments. The `Optimizer` needs to
have the parameters from the `Model` before it's constructed, and the `Schedulers` need to
have the `Optimizer`. Because of this, the typical way we construct things `FromParams`
doesn't work, so we use `Lazy` to allow for constructing the objects sequentially.
If you're not using `FromParams`, you can just construct these arguments in the right order
yourself in your code and call the constructor directly.
"""
check_for_gpu(cuda_device)
if cuda_device >= 0:
# Moving model to GPU here so that the optimizer state gets constructed on
# the right device.
model = model.cuda(cuda_device)
if no_grad:
for name, parameter in model.named_parameters():
if any(re.search(regex, name) for regex in no_grad):
parameter.requires_grad_(False)
common_util.log_frozen_and_tunable_parameter_names(model)
parameters = [[n, p] for n, p in model.named_parameters()
if p.requires_grad]
optimizer_ = optimizer.construct(model_parameters=parameters)
if not optimizer_:
optimizer_ = Optimizer.default(parameters)
try:
batches_per_epoch = len(data_loader)
except TypeError:
# If the dataset is lazy, it won't have a length.
batches_per_epoch = None
moving_average_ = moving_average.construct(parameters=parameters)
learning_rate_scheduler_ = learning_rate_scheduler.construct(
optimizer=optimizer_,
num_epochs=num_epochs,
num_steps_per_epoch=batches_per_epoch)
momentum_scheduler_ = momentum_scheduler.construct(
optimizer=optimizer_)
checkpointer_ = checkpointer.construct() or Checkpointer(
serialization_dir)
return cls(
model,
optimizer_,
data_loader,
patience=patience,
validation_metric=validation_metric,
validation_data_loader=validation_data_loader,
num_epochs=num_epochs,
serialization_dir=serialization_dir,
cuda_device=cuda_device,
grad_norm=grad_norm,
grad_clipping=grad_clipping,
learning_rate_scheduler=learning_rate_scheduler_,
momentum_scheduler=momentum_scheduler_,
checkpointer=checkpointer_,
model_save_interval=model_save_interval,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate,
log_batch_size_period=log_batch_size_period,
moving_average=moving_average_,
distributed=distributed,
local_rank=local_rank,
world_size=world_size,
opt_level=opt_level,
)
class PtTrainer(TrainerBase):
def __init__(
self,
model: Model,
optimizer: torch.optim.Optimizer,
iterator: DataIterator,
train_dataset: Iterable[Instance],
validation_dataset: Optional[Iterable[Instance]] = None,
max_src_len: int = None,
patience: Optional[int] = None,
validation_metric: str = "-loss",
validation_iterator: DataIterator = None,
batch_size: int = 1,
shuffle: bool = True,
num_epochs: int = 20,
serialization_dir: Optional[str] = None,
num_serialized_models_to_keep: int = 20,
keep_serialized_model_every_num_seconds: int = None,
checkpointer: Checkpointer = None,
model_save_interval: float = None,
cuda_device: Union[int, List] = -1,
grad_norm: Optional[float] = None,
grad_clipping: Optional[float] = None,
learning_rate_scheduler: Optional[LearningRateScheduler] = None,
momentum_scheduler: Optional[MomentumScheduler] = None,
summary_interval: int = 100,
histogram_interval: int = None,
should_log_parameter_statistics: bool = True,
should_log_learning_rate: bool = False,
log_batch_size_period: Optional[int] = None,
moving_average: Optional[MovingAverage] = None) -> None:
super().__init__(serialization_dir, cuda_device)
# I am not calling move_to_gpu here, because if the model is
# not already on the GPU then the optimizer is going to be wrong.
self.model = model
self.iterator = iterator
self._validation_iterator = validation_iterator
self.shuffle = shuffle
self.optimizer = optimizer
self.train_data = train_dataset
self._validation_data = validation_dataset
self.max_src_len = max_src_len
self.batch_size = batch_size
# For tracking is_best_so_far and should_stop_early
self._metric_tracker = MetricTracker(patience, validation_metric)
# Get rid of + or -
self._validation_metric = validation_metric[1:]
self._num_epochs = num_epochs
if checkpointer is not None:
# We can't easily check if these parameters were passed in, so check against their default values.
# We don't check against serialization_dir since it is also used by the parent class.
if num_serialized_models_to_keep != 20 or \
keep_serialized_model_every_num_seconds is not None:
raise ConfigurationError(
"When passing a custom Checkpointer, you may not also pass in separate checkpointer "
"args 'num_serialized_models_to_keep' or 'keep_serialized_model_every_num_seconds'."
)
self._checkpointer = checkpointer
else:
self._checkpointer = Checkpointer(
serialization_dir, keep_serialized_model_every_num_seconds,
num_serialized_models_to_keep)
self._model_save_interval = model_save_interval
self._grad_norm = grad_norm
self._grad_clipping = grad_clipping
self._learning_rate_scheduler = learning_rate_scheduler
self._momentum_scheduler = momentum_scheduler
self._moving_average = moving_average
# We keep the total batch number as an instance variable because it
# is used inside a closure for the hook which logs activations in
# ``_enable_activation_logging``.
self._batch_num_total = 0
self._tensorboard = TensorboardWriter(
get_batch_num_total=lambda: self._batch_num_total,
serialization_dir=serialization_dir,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate)
self._log_batch_size_period = log_batch_size_period
self._last_log = 0.0 # time of last logging
# Enable activation logging.
if histogram_interval is not None:
self._tensorboard.enable_activation_logging(self.model)
def rescale_gradients(self) -> Optional[float]:
return training_util.rescale_gradients(self.model, self._grad_norm)
def batch_loss(self, batch_group: List[TensorDict],
for_training: bool) -> torch.Tensor:
"""
Does a forward pass on the given batches and returns the ``loss`` value in the result.
If ``for_training`` is `True` also applies regularization penalty.
"""
if self._multiple_gpu:
output_dict = training_util.data_parallel(batch_group, self.model,
self._cuda_devices)
else:
assert len(batch_group) == 1
batch = batch_group[0]
batch = nn_util.move_to_device(batch, self._cuda_devices[0])
output_dict = self.model(**batch)
try:
loss = output_dict["loss"]
if for_training:
loss += self.model.get_regularization_penalty()
except KeyError:
if for_training:
raise RuntimeError(
"The model you are trying to optimize does not contain a"
" 'loss' key in the output of model.forward(inputs).")
loss = None
return loss
def _train_epoch(self, epoch: int) -> Dict[str, float]:
train_loss = 0.0
self.model.train()
num_gpus = len(self._cuda_devices)
if getattr(self, "train_dataset", None) is None:
self.train_dataset = DMDataSet(data=self.train_data[0],
batch_size=self.batch_size,
num_gpus=num_gpus,
shuffle=True)
self.train_dataset.set_epoch(epoch)
num_training_batches = math.ceil(
len(self.train_dataset) / self.batch_size / num_gpus)
self._last_log = time.time()
batches_this_epoch = 0
if self._batch_num_total is None:
self._batch_num_total = 0
logger.info("Training")
train_generator_tqdm = Tqdm.tqdm(self.train_dataset,
total=num_training_batches)
for batch_group in train_generator_tqdm:
# print('gpu num: ', len(batch_group))
# print('batch_size: ', len(batch_group[0]["source_tokens"]["tokens"]))
# gpu_data = batch_group[0]
# src_data = gpu_data["source_tokens"]["tokens"]
# tgt_data = gpu_data["target_tokens"]["tokens"]
# for sdata, tdata in zip(src_data, tgt_data):
# s = ''.join([self.model.vocab.get_token_from_index(x, "source_tokens") if x != 0 else '' for x in sdata.numpy()])
# t = ''.join([self.model.vocab.get_token_from_index(x, "target_tokens") if x != 0 else '' for x in tdata.numpy()])
# print(s)
# print(t)
batches_this_epoch += 1
self._batch_num_total += 1
batch_num_total = self._batch_num_total
self.optimizer.zero_grad()
loss = self.batch_loss(batch_group, for_training=True)
if torch.isnan(loss):
raise ValueError("nan loss encountered")
loss.backward()
train_loss += loss.item()
batch_grad_norm = self.rescale_gradients()
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step_batch(batch_num_total)
if self._momentum_scheduler:
self._momentum_scheduler.step_batch(batch_num_total)
self.optimizer.step()
# Update moving averages
if self._moving_average is not None:
self._moving_average.apply(batch_num_total)
# Update the description with the latest metrics
metrics = training_util.get_metrics(self.model, train_loss,
batches_this_epoch)
description = training_util.description_from_metrics(metrics)
train_generator_tqdm.set_description(description, refresh=False)
# Log parameter values to Tensorboard
if self._tensorboard.should_log_this_batch():
self._tensorboard.log_learning_rates(self.model,
self.optimizer)
self._tensorboard.add_train_scalar("loss/loss_train",
metrics["loss"])
self._tensorboard.log_metrics(
{"epoch_metrics/" + k: v
for k, v in metrics.items()})
metrics = training_util.get_metrics(self.model,
train_loss,
batches_this_epoch,
reset=True)
return metrics
def _validation_loss(self) -> Tuple[float, int]:
logger.info("Validating")
self.model.eval()
# Replace parameter values with the shadow values from the moving averages.
if self._moving_average is not None:
self._moving_average.assign_average_value()
if self._validation_iterator is not None:
val_iterator = self._validation_iterator
else:
val_iterator = self.iterator
num_gpus = len(self._cuda_devices)
if getattr(self, "val_dataset", None) is None:
self.val_dataset = DMDataSet(data=self._validation_data[0],
batch_size=self.batch_size,
num_gpus=num_gpus,
shuffle=False)
num_validation_batches = math.ceil(
len(self.val_dataset) / self.batch_size / num_gpus)
val_generator_tqdm = Tqdm.tqdm(self.val_dataset,
total=num_validation_batches)
batches_this_epoch = 0
val_loss = 0
for batch_group in val_generator_tqdm:
loss = self.batch_loss(batch_group, for_training=False)
if loss is not None:
batches_this_epoch += 1
val_loss += loss.detach().cpu().numpy()
# Update the description with the latest metrics
val_metrics = training_util.get_metrics(self.model, val_loss,
batches_this_epoch)
description = training_util.description_from_metrics(val_metrics)
val_generator_tqdm.set_description(description, refresh=False)
# Now restore the original parameter values.
if self._moving_average is not None:
self._moving_average.restore()
return val_loss, batches_this_epoch
def train(self) -> Dict[str, Any]:
"""
Trains the supplied model with the supplied parameters.
"""
try:
epoch_counter = self._restore_checkpoint()
except RuntimeError:
traceback.print_exc()
raise ConfigurationError(
"Could not recover training from the checkpoint. Did you mean to output to "
"a different serialization directory or delete the existing serialization "
"directory?")
training_util.enable_gradient_clipping(self.model, self._grad_clipping)
logger.info("Beginning training.")
train_metrics: Dict[str, float] = {}
val_metrics: Dict[str, float] = {}
this_epoch_val_metric: float = None
metrics: Dict[str, Any] = {}
epochs_trained = 0
training_start_time = time.time()
metrics['best_epoch'] = self._metric_tracker.best_epoch
for key, value in self._metric_tracker.best_epoch_metrics.items():
metrics["best_validation_" + key] = value
for epoch in range(epoch_counter, self._num_epochs):
epoch_start_time = time.time()
train_metrics = self._train_epoch(epoch)
if self._validation_data is not None:
with torch.no_grad():
# We have a validation set, so compute all the metrics on it.
val_loss, num_batches = self._validation_loss()
val_metrics = training_util.get_metrics(self.model,
val_loss,
num_batches,
reset=True)
# Check validation metric for early stopping
this_epoch_val_metric = val_metrics[
self._validation_metric]
self._metric_tracker.add_metric(this_epoch_val_metric)
if self._metric_tracker.should_stop_early():
logger.info("Ran out of patience. Stopping training.")
break
self._tensorboard.log_metrics(
train_metrics,
val_metrics=val_metrics,
log_to_console=True,
epoch=epoch + 1) # +1 because tensorboard doesn't like 0
# Create overall metrics dict
training_elapsed_time = time.time() - training_start_time
metrics["training_duration"] = str(
datetime.timedelta(seconds=training_elapsed_time))
metrics["training_start_epoch"] = epoch_counter
metrics["training_epochs"] = epochs_trained
metrics["epoch"] = epoch
for key, value in train_metrics.items():
metrics["training_" + key] = value
for key, value in val_metrics.items():
metrics["validation_" + key] = value
if self._metric_tracker.is_best_so_far():
# Update all the best_ metrics.
# (Otherwise they just stay the same as they were.)
metrics['best_epoch'] = epoch
for key, value in val_metrics.items():
metrics["best_validation_" + key] = value
self._metric_tracker.best_epoch_metrics = val_metrics
if self._serialization_dir:
dump_metrics(
os.path.join(self._serialization_dir,
f'metrics_epoch_{epoch}.json'), metrics)
# The Scheduler API is agnostic to whether your schedule requires a validation metric -
# if it doesn't, the validation metric passed here is ignored.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step(this_epoch_val_metric,
epoch)
if self._momentum_scheduler:
self._momentum_scheduler.step(this_epoch_val_metric, epoch)
self._save_checkpoint(epoch)
epoch_elapsed_time = time.time() - epoch_start_time
logger.info("Epoch duration: %s",
datetime.timedelta(seconds=epoch_elapsed_time))
if epoch < self._num_epochs - 1:
training_elapsed_time = time.time() - training_start_time
estimated_time_remaining = training_elapsed_time * \
((self._num_epochs - epoch_counter) / float(epoch - epoch_counter + 1) - 1)
formatted_time = str(
datetime.timedelta(seconds=int(estimated_time_remaining)))
logger.info("Estimated training time remaining: %s",
formatted_time)
epochs_trained += 1
# make sure pending events are flushed to disk and files are closed properly
self._tensorboard.close()
# Load the best model state before returning
best_model_state = self._checkpointer.best_model_state()
if best_model_state:
self.model.load_state_dict(best_model_state)
return metrics
def _save_checkpoint(self, epoch: Union[int, str]) -> None:
"""
Saves a checkpoint of the model to self._serialization_dir.
Is a no-op if self._serialization_dir is None.
Parameters
----------
epoch : Union[int, str], required.
The epoch of training. If the checkpoint is saved in the middle
of an epoch, the parameter is a string with the epoch and timestamp.
"""
# If moving averages are used for parameters, we save
# the moving average values into checkpoint, instead of the current values.
if self._moving_average is not None:
self._moving_average.assign_average_value()
# These are the training states we need to persist.
training_states = {
"metric_tracker": self._metric_tracker.state_dict(),
"optimizer": self.optimizer.state_dict(),
"batch_num_total": self._batch_num_total
}
# If we have a learning rate or momentum scheduler, we should persist them too.
if self._learning_rate_scheduler is not None:
training_states[
"learning_rate_scheduler"] = self._learning_rate_scheduler.state_dict(
)
if self._momentum_scheduler is not None:
training_states[
"momentum_scheduler"] = self._momentum_scheduler.state_dict()
self._checkpointer.save_checkpoint(
model_state=self.model.state_dict(),
epoch=epoch,
training_states=training_states,
is_best_so_far=self._metric_tracker.is_best_so_far())
# Restore the original values for parameters so that training will not be affected.
if self._moving_average is not None:
self._moving_average.restore()
def _restore_checkpoint(self) -> int:
"""
Restores the model and training state from the last saved checkpoint.
This includes an epoch count and optimizer state, which is serialized separately
from model parameters. This function should only be used to continue training -
if you wish to load a model for inference/load parts of a model into a new
computation graph, you should use the native Pytorch functions:
`` model.load_state_dict(torch.load("/path/to/model/weights.th"))``
If ``self._serialization_dir`` does not exist or does not contain any checkpointed weights,
this function will do nothing and return 0.
Returns
-------
epoch: int
The epoch at which to resume training, which should be one after the epoch
in the saved training state.
"""
model_state, training_state = self._checkpointer.restore_checkpoint()
if not training_state:
# No checkpoint to restore, start at 0
return 0
self.model.load_state_dict(model_state)
self.optimizer.load_state_dict(training_state["optimizer"])
if self._learning_rate_scheduler is not None and "learning_rate_scheduler" in training_state:
self._learning_rate_scheduler.load_state_dict(
training_state["learning_rate_scheduler"])
if self._momentum_scheduler is not None and "momentum_scheduler" in training_state:
self._momentum_scheduler.load_state_dict(
training_state["momentum_scheduler"])
training_util.move_optimizer_to_cuda(self.optimizer)
# Currently the ``training_state`` contains a serialized ``MetricTracker``.
if "metric_tracker" in training_state:
self._metric_tracker.load_state_dict(
training_state["metric_tracker"])
# It used to be the case that we tracked ``val_metric_per_epoch``.
elif "val_metric_per_epoch" in training_state:
self._metric_tracker.clear()
self._metric_tracker.add_metrics(
training_state["val_metric_per_epoch"])
# And before that we didn't track anything.
else:
self._metric_tracker.clear()
if isinstance(training_state["epoch"], int):
epoch_to_return = training_state["epoch"] + 1
else:
epoch_to_return = int(training_state["epoch"].split('.')[0]) + 1
# For older checkpoints with batch_num_total missing, default to old behavior where
# it is unchanged.
batch_num_total = training_state.get('batch_num_total')
if batch_num_total is not None:
self._batch_num_total = batch_num_total
return epoch_to_return
# Requires custom from_params.
@classmethod
def from_params(cls,
params: Params,
serialization_dir: str,
recover: bool = False,
cache_directory: str = None,
cache_prefix: str = None) -> 'PtTrainer':
max_src_len = params.dataset_reader.get('max_src_len', None)
all_datasets = training_util.datasets_from_params(
params, cache_directory, cache_prefix)
datasets_for_vocab_creation = set(
params.pop("datasets_for_vocab_creation", all_datasets))
for dataset in datasets_for_vocab_creation:
if dataset not in all_datasets:
raise ConfigurationError(
f"invalid 'dataset_for_vocab_creation' {dataset}")
logger.info(
"From dataset instances, %s will be considered for vocabulary creation.",
", ".join(datasets_for_vocab_creation))
if recover and os.path.exists(
os.path.join(serialization_dir, "vocabulary")):
vocab = Vocabulary.from_files(
os.path.join(serialization_dir, "vocabulary"))
params.pop("vocabulary", {})
else:
vocab = Vocabulary.from_params(params.pop(
"vocabulary", {}), (instance
for key, dataset in all_datasets.items()
if key in datasets_for_vocab_creation
for instance in dataset))
model = Model.from_params(vocab=vocab, params=params.pop('model'))
# If vocab extension is ON for training, embedding extension should also be
# done. If vocab and embeddings are already in sync, it would be a no-op.
model.extend_embedder_vocab()
# Initializing the model can have side effect of expanding the vocabulary
vocab.save_to_files(os.path.join(serialization_dir, "vocabulary"))
iterator = DataIterator.from_params(params.pop("iterator"))
iterator.index_with(model.vocab)
validation_iterator_params = params.pop("validation_iterator", None)
if validation_iterator_params:
validation_iterator = DataIterator.from_params(
validation_iterator_params)
validation_iterator.index_with(model.vocab)
else:
validation_iterator = None
train_data = all_datasets['train']
validation_data = all_datasets.get('validation')
test_data = all_datasets.get('test')
trainer_params = params.pop("trainer")
no_grad_regexes = trainer_params.pop("no_grad", ())
for name, parameter in model.named_parameters():
if any(re.search(regex, name) for regex in no_grad_regexes):
parameter.requires_grad_(False)
frozen_parameter_names, tunable_parameter_names = \
get_frozen_and_tunable_parameter_names(model)
logger.info("Following parameters are Frozen (without gradient):")
for name in frozen_parameter_names:
logger.info(name)
logger.info("Following parameters are Tunable (with gradient):")
for name in tunable_parameter_names:
logger.info(name)
params = trainer_params
patience = params.pop_int("patience", None)
validation_metric = params.pop("validation_metric", "-loss")
shuffle = params.pop_bool("shuffle", True)
num_epochs = params.pop_int("num_epochs", 20)
cuda_device = parse_cuda_device(params.pop("cuda_device", -1))
grad_norm = params.pop_float("grad_norm", None)
grad_clipping = params.pop_float("grad_clipping", None)
lr_scheduler_params = params.pop("learning_rate_scheduler", None)
momentum_scheduler_params = params.pop("momentum_scheduler", None)
if isinstance(cuda_device, list):
model_device = cuda_device[0]
else:
model_device = cuda_device
if model_device >= 0:
# Moving model to GPU here so that the optimizer state gets constructed on
# the right device.
model = model.cuda(model_device)
parameters = [[n, p] for n, p in model.named_parameters()
if p.requires_grad]
optimizer = Optimizer.from_params(parameters, params.pop("optimizer"))
if "moving_average" in params:
moving_average = MovingAverage.from_params(
params.pop("moving_average"), parameters=parameters)
else:
moving_average = None
if lr_scheduler_params:
lr_scheduler = LearningRateScheduler.from_params(
optimizer, lr_scheduler_params)
else:
lr_scheduler = None
if momentum_scheduler_params:
momentum_scheduler = MomentumScheduler.from_params(
optimizer, momentum_scheduler_params)
else:
momentum_scheduler = None
if 'checkpointer' in params:
if 'keep_serialized_model_every_num_seconds' in params or \
'num_serialized_models_to_keep' in params:
raise ConfigurationError(
"Checkpointer may be initialized either from the 'checkpointer' key or from the "
"keys 'num_serialized_models_to_keep' and 'keep_serialized_model_every_num_seconds'"
" but the passed config uses both methods.")
checkpointer = Checkpointer.from_params(params.pop("checkpointer"))
else:
num_serialized_models_to_keep = params.pop_int(
"num_serialized_models_to_keep", 20)
keep_serialized_model_every_num_seconds = params.pop_int(
"keep_serialized_model_every_num_seconds", None)
checkpointer = Checkpointer(
serialization_dir=serialization_dir,
num_serialized_models_to_keep=num_serialized_models_to_keep,
keep_serialized_model_every_num_seconds=
keep_serialized_model_every_num_seconds)
model_save_interval = params.pop_float("model_save_interval", None)
summary_interval = params.pop_int("summary_interval", 100)
histogram_interval = params.pop_int("histogram_interval", None)
should_log_parameter_statistics = params.pop_bool(
"should_log_parameter_statistics", True)
should_log_learning_rate = params.pop_bool("should_log_learning_rate",
False)
log_batch_size_period = params.pop_int("log_batch_size_period", None)
return cls(
model,
optimizer,
iterator,
train_data,
validation_data,
patience=patience,
validation_metric=validation_metric,
validation_iterator=validation_iterator,
max_src_len=max_src_len,
shuffle=shuffle,
num_epochs=num_epochs,
serialization_dir=serialization_dir,
cuda_device=cuda_device,
grad_norm=grad_norm,
grad_clipping=grad_clipping,
learning_rate_scheduler=lr_scheduler,
momentum_scheduler=momentum_scheduler,
checkpointer=checkpointer,
model_save_interval=model_save_interval,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate,
log_batch_size_period=log_batch_size_period,
moving_average=moving_average,
batch_size=iterator._batch_size)
class MatchingTrainer(TrainerBase):
def __init__(
self,
model: Model,
optimizer: torch.optim.Optimizer,
iterator: DataIterator,
train_dataset: Iterable[Instance],
validation_dataset: Optional[Iterable[Instance]] = None,
patience: Optional[int] = None,
validation_metric: str = "-loss",
validation_iterator: DataIterator = None,
retrieve_text: bool = True,
shuffle: bool = True,
num_epochs: int = 20,
serialization_dir: Optional[str] = None,
num_serialized_models_to_keep: int = 20,
keep_serialized_model_every_num_seconds: int = None,
checkpointer: Checkpointer = None,
model_save_interval: float = None,
cuda_device: Union[int, List] = -1,
grad_norm: Optional[float] = None,
grad_clipping: Optional[float] = None,
learning_rate_scheduler: Optional[LearningRateScheduler] = None,
momentum_scheduler: Optional[MomentumScheduler] = None,
summary_interval: int = 100,
histogram_interval: int = None,
should_log_parameter_statistics: bool = True,
should_log_learning_rate: bool = False,
log_batch_size_period: Optional[int] = None,
moving_average: Optional[MovingAverage] = None) -> None:
"""
A trainer for doing supervised learning. It just takes a labeled dataset
and a ``DataIterator``, and uses the supplied ``Optimizer`` to learn the weights
for your model over some fixed number of epochs. You can also pass in a validation
dataset and enable early stopping. There are many other bells and whistles as well.
Parameters
----------
model : ``Model``, required.
An AllenNLP model to be optimized. Pytorch Modules can also be optimized if
their ``forward`` method returns a dictionary with a "loss" key, containing a
scalar tensor representing the loss function to be optimized.
If you are training your model using GPUs, your model should already be
on the correct device. (If you use `Trainer.from_params` this will be
handled for you.)
optimizer : ``torch.nn.Optimizer``, required.
An instance of a Pytorch Optimizer, instantiated with the parameters of the
model to be optimized.
iterator : ``DataIterator``, required.
A method for iterating over a ``Dataset``, yielding padded indexed batches.
train_dataset : ``Dataset``, required.
A ``Dataset`` to train on. The dataset should have already been indexed.
validation_dataset : ``Dataset``, optional, (default = None).
A ``Dataset`` to evaluate on. The dataset should have already been indexed.
patience : Optional[int] > 0, optional (default=None)
Number of epochs to be patient before early stopping: the training is stopped
after ``patience`` epochs with no improvement. If given, it must be ``> 0``.
If None, early stopping is disabled.
validation_metric : str, optional (default="loss")
Validation metric to measure for whether to stop training using patience
and whether to serialize an ``is_best`` model each epoch. The metric name
must be prepended with either "+" or "-", which specifies whether the metric
is an increasing or decreasing function.
validation_iterator : ``DataIterator``, optional (default=None)
An iterator to use for the validation set. If ``None``, then
use the training `iterator`.
shuffle: ``bool``, optional (default=True)
Whether to shuffle the instances in the iterator or not.
num_epochs : int, optional (default = 20)
Number of training epochs.
serialization_dir : str, optional (default=None)
Path to directory for saving and loading model files. Models will not be saved if
this parameter is not passed.
num_serialized_models_to_keep : ``int``, optional (default=20)
Number of previous model checkpoints to retain. Default is to keep 20 checkpoints.
A value of None or -1 means all checkpoints will be kept.
keep_serialized_model_every_num_seconds : ``int``, optional (default=None)
If num_serialized_models_to_keep is not None, then occasionally it's useful to
save models at a given interval in addition to the last num_serialized_models_to_keep.
To do so, specify keep_serialized_model_every_num_seconds as the number of seconds
between permanently saved checkpoints. Note that this option is only used if
num_serialized_models_to_keep is not None, otherwise all checkpoints are kept.
checkpointer : ``Checkpointer``, optional (default=None)
An instance of class Checkpointer to use instead of the default. If a checkpointer is specified,
the arguments num_serialized_models_to_keep and keep_serialized_model_every_num_seconds should
not be specified. The caller is responsible for initializing the checkpointer so that it is
consistent with serialization_dir.
model_save_interval : ``float``, optional (default=None)
If provided, then serialize models every ``model_save_interval``
seconds within single epochs. In all cases, models are also saved
at the end of every epoch if ``serialization_dir`` is provided.
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.
grad_norm : ``float``, optional, (default = None).
If provided, gradient norms will be rescaled to have a maximum of this value.
grad_clipping : ``float``, optional (default = ``None``).
If provided, gradients will be clipped `during the backward pass` to have an (absolute)
maximum of this value. If you are getting ``NaNs`` in your gradients during training
that are not solved by using ``grad_norm``, you may need this.
learning_rate_scheduler : ``LearningRateScheduler``, optional (default = None)
If specified, the learning rate will be decayed with respect to
this schedule at the end of each epoch (or batch, if the scheduler implements
the ``step_batch`` method). If you use :class:`torch.optim.lr_scheduler.ReduceLROnPlateau`,
this will use the ``validation_metric`` provided to determine if learning has plateaued.
To support updating the learning rate on every batch, this can optionally implement
``step_batch(batch_num_total)`` which updates the learning rate given the batch number.
momentum_scheduler : ``MomentumScheduler``, optional (default = None)
If specified, the momentum will be updated at the end of each batch or epoch
according to the schedule.
summary_interval: ``int``, optional, (default = 100)
Number of batches between logging scalars to tensorboard
histogram_interval : ``int``, optional, (default = ``None``)
If not None, then log histograms to tensorboard every ``histogram_interval`` batches.
When this parameter is specified, the following additional logging is enabled:
* Histograms of model parameters
* The ratio of parameter update norm to parameter norm
* Histogram of layer activations
We log histograms of the parameters returned by
``model.get_parameters_for_histogram_tensorboard_logging``.
The layer activations are logged for any modules in the ``Model`` that have
the attribute ``should_log_activations`` set to ``True``. Logging
histograms requires a number of GPU-CPU copies during training and is typically
slow, so we recommend logging histograms relatively infrequently.
Note: only Modules that return tensors, tuples of tensors or dicts
with tensors as values currently support activation logging.
should_log_parameter_statistics : ``bool``, optional, (default = True)
Whether to send parameter statistics (mean and standard deviation
of parameters and gradients) to tensorboard.
should_log_learning_rate : ``bool``, optional, (default = False)
Whether to send parameter specific learning rate to tensorboard.
log_batch_size_period : ``int``, optional, (default = ``None``)
If defined, how often to log the average batch size.
moving_average: ``MovingAverage``, optional, (default = None)
If provided, we will maintain moving averages for all parameters. During training, we
employ a shadow variable for each parameter, which maintains the moving average. During
evaluation, we backup the original parameters and assign the moving averages to corresponding
parameters. Be careful that when saving the checkpoint, we will save the moving averages of
parameters. This is necessary because we want the saved model to perform as well as the validated
model if we load it later. But this may cause problems if you restart the training from checkpoint.
"""
super().__init__(serialization_dir, cuda_device)
# I am not calling move_to_gpu here, because if the model is
# not already on the GPU then the optimizer is going to be wrong.
self.model = model
self.iterator = iterator
self._validation_iterator = validation_iterator
self.shuffle = shuffle
self.optimizer = optimizer
self.train_data = train_dataset
self._validation_data = validation_dataset
if patience is None: # no early stopping
if validation_dataset:
logger.warning(
'You provided a validation dataset but patience was set to None, '
'meaning that early stopping is disabled')
elif (not isinstance(patience, int)) or patience <= 0:
raise ConfigurationError(
'{} is an invalid value for "patience": it must be a positive integer '
'or None (if you want to disable early stopping)'.format(
patience))
# For tracking is_best_so_far and should_stop_early
self._metric_tracker = MetricTracker(patience, validation_metric)
# Get rid of + or -
self._validation_metric = validation_metric[1:]
self._num_epochs = num_epochs
if checkpointer is not None:
# We can't easily check if these parameters were passed in, so check against their default values.
# We don't check against serialization_dir since it is also used by the parent class.
if num_serialized_models_to_keep != 20 or \
keep_serialized_model_every_num_seconds is not None:
raise ConfigurationError(
"When passing a custom Checkpointer, you may not also pass in separate checkpointer "
"args 'num_serialized_models_to_keep' or 'keep_serialized_model_every_num_seconds'."
)
self._checkpointer = checkpointer
else:
self._checkpointer = Checkpointer(
serialization_dir, keep_serialized_model_every_num_seconds,
num_serialized_models_to_keep)
self._model_save_interval = model_save_interval
self._grad_norm = grad_norm
self._grad_clipping = grad_clipping
self._learning_rate_scheduler = learning_rate_scheduler
self._momentum_scheduler = momentum_scheduler
self._moving_average = moving_average
# We keep the total batch number as an instance variable because it
# is used inside a closure for the hook which logs activations in
# ``_enable_activation_logging``.
self._batch_num_total = 0
self._tensorboard = TensorboardWriter(
get_batch_num_total=lambda: self._batch_num_total,
serialization_dir=serialization_dir,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate)
self._log_batch_size_period = log_batch_size_period
self._last_log = 0.0 # time of last logging
# Enable activation logging.
if histogram_interval is not None:
self._tensorboard.enable_activation_logging(self.model)
self.retrieve_text = retrieve_text
if retrieve_text:
self.train_text_db = []
self.train_segment_ids_db = []
for instance in train_dataset:
self.train_text_db.append(instance.fields['token_ids'].tokens)
self.train_token_indexers = instance.fields[
'token_ids']._token_indexers
self.train_segment_ids_db.append(
instance.fields['segment_ids'].array)
self.train_indices = range(len(self.train_text_db))
if validation_dataset:
self.val_text_db = []
self.val_segment_ids_db = []
for instance in validation_dataset:
self.val_text_db.append(
instance.fields['token_ids'].tokens)
self.val_token_indexers = instance.fields[
'token_ids']._token_indexers
self.val_segment_ids_db.append(
instance.fields['segment_ids'].array)
self.val_indices = range(len(self.val_text_db))
else:
self.train_image_db = []
# self.category_weights = defaultdict(int)
for instance in train_dataset:
if 'category' in instance.fields:
self.category_weights[
instance.fields['category'].metadata] += 1
self.train_image_db.append(instance.fields['images'].array)
# self.category_weights = {category: round(np.log(float(len(train_dataset))/self.category_weights[category]), 5) for category in self.category_weights}"""
if validation_dataset:
self.val_image_db = []
for instance in validation_dataset:
self.val_image_db.append(instance.fields['images'].array)
self.num_negative_samples = 2
def rescale_gradients(self) -> Optional[float]:
return training_util.rescale_gradients(self.model, self._grad_norm)
def batch_loss(self, batch_group: List[TensorDict],
for_training: bool) -> torch.Tensor:
"""
Does a forward pass on the given batches and returns the ``loss`` value in the result.
If ``for_training`` is `True` also applies regularization penalty.
"""
if self._multiple_gpu:
output_dict = training_util.data_parallel(batch_group, self.model,
self._cuda_devices)
else:
assert len(batch_group) == 1
batch = batch_group[0]
batch = nn_util.move_to_device(batch, self._cuda_devices[0])
output_dict = self.model(**batch)
try:
loss = output_dict["loss"]
if for_training:
loss += self.model.get_regularization_penalty()
except KeyError:
if for_training:
raise RuntimeError(
"The model you are trying to optimize does not contain a"
" 'loss' key in the output of model.forward(inputs).")
loss = None
return loss
def _train_epoch(self, epoch: int) -> Dict[str, float]:
"""
Trains one epoch and returns metrics.
"""
logger.info("Epoch %d/%d", epoch, self._num_epochs - 1)
peak_cpu_usage = peak_memory_mb()
logger.info(f"Peak CPU memory usage MB: {peak_cpu_usage}")
gpu_usage = []
for gpu, memory in gpu_memory_mb().items():
gpu_usage.append((gpu, memory))
logger.info(f"GPU {gpu} memory usage MB: {memory}")
train_loss = 0.0
# Set the model to "train" mode.
self.model.train()
num_gpus = len(self._cuda_devices)
# Get tqdm for the training batches
raw_train_generator = self.iterator(self.train_data,
num_epochs=1,
shuffle=self.shuffle)
train_generator = lazy_groups_of(raw_train_generator, num_gpus)
num_training_batches = math.ceil(
self.iterator.get_num_batches(self.train_data) / num_gpus)
self._last_log = time.time()
last_save_time = time.time()
batches_this_epoch = 0
if self._batch_num_total is None:
self._batch_num_total = 0
histogram_parameters = set(
self.model.get_parameters_for_histogram_tensorboard_logging())
logger.info("Training")
train_generator_tqdm = Tqdm.tqdm(train_generator,
total=num_training_batches)
cumulative_batch_size = 0
for batch_group in train_generator_tqdm:
batches_this_epoch += 1
self._batch_num_total += 1
batch_num_total = self._batch_num_total
self.optimizer.zero_grad()
images = []
text = []
segment_ids = []
labels = []
for i in range(len(batch_group[0]['images'])):
positive_index = random.randint(0, self.num_negative_samples)
labels.append(positive_index)
if self.retrieve_text:
instance_text = []
instance_segment_ids = []
for j in range(self.num_negative_samples + 1):
if j == positive_index:
instance_text.append(batch_group[0]['token_ids']
['tokens'][i, :].tolist())
instance_segment_ids.append(
batch_group[0]['segment_ids'][i].tolist())
else:
negative_sample_index = random.choice(
self.train_indices)
text_field = TextField(
self.train_text_db[negative_sample_index],
self.train_token_indexers)
text_field.index(self.model.vocab)
padding_lengths = text_field.get_padding_lengths()
instance_text.append(
text_field.as_tensor(
padding_lengths=padding_lengths)
['tokens'].tolist())
instance_segment_ids.append(
self.train_segment_ids_db[
negative_sample_index].tolist())
text += instance_text
segment_ids += instance_segment_ids
else:
instance_images = [
None for _ in range(self.num_negative_samples + 1)
]
for j in range(self.num_negative_samples + 1):
if j == positive_index:
instance_images[j] = np.expand_dims(
batch_group[0]['images'][i].numpy(), 0)
else:
instance_images[j] = np.expand_dims(
random.choice(self.train_image_db), 0)
images += instance_images
matching_label_field_name = "labels"
if self.retrieve_text:
max_text_len = max([len(sequence) for sequence in text])
text = [
sequence +
[0 for _ in range(max_text_len - len(sequence))]
for sequence in text
]
batch_group[0]['token_ids'] = {
'tokens': torch.LongTensor(text)
}
segment_ids = [
sequence +
[0 for _ in range(max_text_len - len(sequence))]
for sequence in segment_ids
]
batch_group[0]['segment_ids'] = torch.from_numpy(
np.array(segment_ids, dtype=np.int64))
else:
batch_group[0]['images'] = torch.from_numpy(np.vstack(images))
batch_group[0][matching_label_field_name] = torch.from_numpy(
np.array(labels, dtype=np.int64))
loss = self.batch_loss(batch_group, for_training=True)
if torch.isnan(loss):
raise ValueError("nan loss encountered")
loss.backward()
train_loss += loss.item()
batch_grad_norm = self.rescale_gradients()
# This does nothing if batch_num_total is None or you are using a
# scheduler which doesn't update per batch.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step_batch(batch_num_total)
if self._momentum_scheduler:
self._momentum_scheduler.step_batch(batch_num_total)
if self._tensorboard.should_log_histograms_this_batch():
# get the magnitude of parameter updates for logging
# We need a copy of current parameters to compute magnitude of updates,
# and copy them to CPU so large models won't go OOM on the GPU.
param_updates = {
name: param.detach().cpu().clone()
for name, param in self.model.named_parameters()
}
self.optimizer.step()
for name, param in self.model.named_parameters():
param_updates[name].sub_(param.detach().cpu())
update_norm = torch.norm(param_updates[name].view(-1, ))
param_norm = torch.norm(param.view(-1, )).cpu()
self._tensorboard.add_train_scalar(
"gradient_update/" + name,
update_norm / (param_norm + 1e-7))
else:
self.optimizer.step()
# Update moving averages
if self._moving_average is not None:
self._moving_average.apply(batch_num_total)
# Update the description with the latest metrics
metrics = training_util.get_metrics(self.model, train_loss,
batches_this_epoch)
description = training_util.description_from_metrics(metrics)
train_generator_tqdm.set_description(description, refresh=False)
# Log parameter values to Tensorboard
if self._tensorboard.should_log_this_batch():
self._tensorboard.log_parameter_and_gradient_statistics(
self.model, batch_grad_norm)
self._tensorboard.log_learning_rates(self.model,
self.optimizer)
self._tensorboard.add_train_scalar("loss/loss_train",
metrics["loss"])
self._tensorboard.log_metrics(
{"epoch_metrics/" + k: v
for k, v in metrics.items()})
if self._tensorboard.should_log_histograms_this_batch():
self._tensorboard.log_histograms(self.model,
histogram_parameters)
if self._log_batch_size_period:
cur_batch = sum([
training_util.get_batch_size(batch)
for batch in batch_group
])
cumulative_batch_size += cur_batch
if (batches_this_epoch - 1) % self._log_batch_size_period == 0:
average = cumulative_batch_size / batches_this_epoch
logger.info(
f"current batch size: {cur_batch} mean batch size: {average}"
)
self._tensorboard.add_train_scalar("current_batch_size",
cur_batch)
self._tensorboard.add_train_scalar("mean_batch_size",
average)
# Save model if needed.
if self._model_save_interval is not None and (
time.time() - last_save_time > self._model_save_interval):
last_save_time = time.time()
self._save_checkpoint('{0}.{1}'.format(
epoch, training_util.time_to_str(int(last_save_time))))
metrics = training_util.get_metrics(self.model,
train_loss,
batches_this_epoch,
reset=True)
metrics['cpu_memory_MB'] = peak_cpu_usage
for (gpu_num, memory) in gpu_usage:
metrics['gpu_' + str(gpu_num) + '_memory_MB'] = memory
return metrics
def _validation_loss(self) -> Tuple[float, int]:
"""
Computes the validation loss. Returns it and the number of batches.
"""
logger.info("Validating")
self.model.eval()
# Replace parameter values with the shadow values from the moving averages.
if self._moving_average is not None:
self._moving_average.assign_average_value()
if self._validation_iterator is not None:
val_iterator = self._validation_iterator
else:
val_iterator = self.iterator
num_gpus = len(self._cuda_devices)
raw_val_generator = val_iterator(self._validation_data,
num_epochs=1,
shuffle=False)
val_generator = lazy_groups_of(raw_val_generator, num_gpus)
num_validation_batches = math.ceil(
val_iterator.get_num_batches(self._validation_data) / num_gpus)
val_generator_tqdm = Tqdm.tqdm(val_generator,
total=num_validation_batches)
batches_this_epoch = 0
val_loss = 0
for batch_group in val_generator_tqdm:
images = []
text = []
segment_ids = []
labels = []
num_negative_samples = self.num_negative_samples * 10
for i in range(len(batch_group[0]['images'])):
positive_index = random.randint(0, num_negative_samples)
labels.append(positive_index)
if self.retrieve_text:
instance_text = []
instance_segment_ids = []
for j in range(num_negative_samples + 1):
if j == positive_index:
instance_text.append(batch_group[0]['token_ids']
['tokens'][i].tolist())
instance_segment_ids.append(
batch_group[0]['segment_ids'][i].tolist())
else:
negative_sample_index = random.choice(
self.val_indices)
text_field = TextField(
self.val_text_db[negative_sample_index],
self.val_token_indexers)
text_field.index(self.model.vocab)
padding_lengths = text_field.get_padding_lengths()
instance_text.append(
text_field.as_tensor(
padding_lengths=padding_lengths)
['tokens'].tolist())
instance_segment_ids.append(
self.val_segment_ids_db[negative_sample_index].
tolist())
text += instance_text
segment_ids += instance_segment_ids
else:
instance_images = [
None for _ in range(num_negative_samples + 1)
]
for j in range(num_negative_samples + 1):
if j == positive_index:
instance_images[j] = np.expand_dims(
batch_group[0]['images'][i].numpy(), 0)
else:
instance_images[j] = np.expand_dims(
random.choice(self.val_image_db), 0)
images += instance_images
matching_label_field_name = "labels"
if self.retrieve_text:
max_text_len = max([len(sequence) for sequence in text])
text = [
sequence +
[0 for _ in range(max_text_len - len(sequence))]
for sequence in text
]
batch_group[0]['token_ids'] = {
'tokens': torch.LongTensor(text)
}
segment_ids = [
sequence +
[0 for _ in range(max_text_len - len(sequence))]
for sequence in segment_ids
]
batch_group[0]['segment_ids'] = torch.from_numpy(
np.array(segment_ids, dtype=np.int64))
else:
batch_group[0]['images'] = torch.from_numpy(np.vstack(images))
batch_group[0][matching_label_field_name] = torch.from_numpy(
np.array(labels, dtype=np.int64))
loss = self.batch_loss(batch_group, for_training=False)
if loss is not None:
# You shouldn't necessarily have to compute a loss for validation, so we allow for
# `loss` to be None. We need to be careful, though - `batches_this_epoch` is
# currently only used as the divisor for the loss function, so we can safely only
# count those batches for which we actually have a loss. If this variable ever
# gets used for something else, we might need to change things around a bit.
batches_this_epoch += 1
val_loss += loss.detach().cpu().numpy()
# Update the description with the latest metrics
val_metrics = training_util.get_metrics(self.model, val_loss,
batches_this_epoch)
description = training_util.description_from_metrics(val_metrics)
val_generator_tqdm.set_description(description, refresh=False)
# Now restore the original parameter values.
if self._moving_average is not None:
self._moving_average.restore()
return val_loss, batches_this_epoch
def train(self) -> Dict[str, Any]:
"""
Trains the supplied model with the supplied parameters.
"""
try:
epoch_counter = self._restore_checkpoint()
except RuntimeError:
traceback.print_exc()
raise ConfigurationError(
"Could not recover training from the checkpoint. Did you mean to output to "
"a different serialization directory or delete the existing serialization "
"directory?")
training_util.enable_gradient_clipping(self.model, self._grad_clipping)
logger.info("Beginning training.")
train_metrics: Dict[str, float] = {}
val_metrics: Dict[str, float] = {}
this_epoch_val_metric: float = None
metrics: Dict[str, Any] = {}
epochs_trained = 0
training_start_time = time.time()
metrics['best_epoch'] = self._metric_tracker.best_epoch
for key, value in self._metric_tracker.best_epoch_metrics.items():
metrics["best_validation_" + key] = value
for epoch in range(epoch_counter, self._num_epochs):
epoch_start_time = time.time()
train_metrics = self._train_epoch(epoch)
# get peak of memory usage
if 'cpu_memory_MB' in train_metrics:
metrics['peak_cpu_memory_MB'] = max(
metrics.get('peak_cpu_memory_MB', 0),
train_metrics['cpu_memory_MB'])
for key, value in train_metrics.items():
if key.startswith('gpu_'):
metrics["peak_" + key] = max(metrics.get("peak_" + key, 0),
value)
if self._validation_data is not None:
with torch.no_grad():
# We have a validation set, so compute all the metrics on it.
val_loss, num_batches = self._validation_loss()
val_metrics = training_util.get_metrics(self.model,
val_loss,
num_batches,
reset=True)
# Check validation metric for early stopping
this_epoch_val_metric = val_metrics[
self._validation_metric]
self._metric_tracker.add_metric(this_epoch_val_metric)
if self._metric_tracker.should_stop_early():
logger.info("Ran out of patience. Stopping training.")
break
self._tensorboard.log_metrics(
train_metrics,
val_metrics=val_metrics,
log_to_console=True,
epoch=epoch + 1) # +1 because tensorboard doesn't like 0
# Create overall metrics dict
training_elapsed_time = time.time() - training_start_time
metrics["training_duration"] = str(
datetime.timedelta(seconds=training_elapsed_time))
metrics["training_start_epoch"] = epoch_counter
metrics["training_epochs"] = epochs_trained
metrics["epoch"] = epoch
for key, value in train_metrics.items():
metrics["training_" + key] = value
for key, value in val_metrics.items():
metrics["validation_" + key] = value
if self._metric_tracker.is_best_so_far():
# Update all the best_ metrics.
# (Otherwise they just stay the same as they were.)
metrics['best_epoch'] = epoch
for key, value in val_metrics.items():
metrics["best_validation_" + key] = value
self._metric_tracker.best_epoch_metrics = val_metrics
if self._serialization_dir:
dump_metrics(
os.path.join(self._serialization_dir,
f'metrics_epoch_{epoch}.json'), metrics)
# The Scheduler API is agnostic to whether your schedule requires a validation metric -
# if it doesn't, the validation metric passed here is ignored.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step(this_epoch_val_metric,
epoch)
if self._momentum_scheduler:
self._momentum_scheduler.step(this_epoch_val_metric, epoch)
self._save_checkpoint(epoch)
epoch_elapsed_time = time.time() - epoch_start_time
logger.info("Epoch duration: %s",
datetime.timedelta(seconds=epoch_elapsed_time))
if epoch < self._num_epochs - 1:
training_elapsed_time = time.time() - training_start_time
estimated_time_remaining = training_elapsed_time * \
((self._num_epochs - epoch_counter) / float(epoch - epoch_counter + 1) - 1)
formatted_time = str(
datetime.timedelta(seconds=int(estimated_time_remaining)))
logger.info("Estimated training time remaining: %s",
formatted_time)
epochs_trained += 1
# make sure pending events are flushed to disk and files are closed properly
self._tensorboard.close()
# Load the best model state before returning
best_model_state = self._checkpointer.best_model_state()
if best_model_state:
self.model.load_state_dict(best_model_state)
return metrics
def _save_checkpoint(self, epoch: Union[int, str]) -> None:
"""
Saves a checkpoint of the model to self._serialization_dir.
Is a no-op if self._serialization_dir is None.
Parameters
----------
epoch : Union[int, str], required.
The epoch of training. If the checkpoint is saved in the middle
of an epoch, the parameter is a string with the epoch and timestamp.
"""
# If moving averages are used for parameters, we save
# the moving average values into checkpoint, instead of the current values.
if self._moving_average is not None:
self._moving_average.assign_average_value()
# These are the training states we need to persist.
training_states = {
"metric_tracker": self._metric_tracker.state_dict(),
"optimizer": self.optimizer.state_dict(),
"batch_num_total": self._batch_num_total
}
# If we have a learning rate or momentum scheduler, we should persist them too.
if self._learning_rate_scheduler is not None:
training_states[
"learning_rate_scheduler"] = self._learning_rate_scheduler.state_dict(
)
if self._momentum_scheduler is not None:
training_states[
"momentum_scheduler"] = self._momentum_scheduler.state_dict()
self._checkpointer.save_checkpoint(
model_state=self.model.state_dict(),
epoch=epoch,
training_states=training_states,
is_best_so_far=self._metric_tracker.is_best_so_far())
# Restore the original values for parameters so that training will not be affected.
if self._moving_average is not None:
self._moving_average.restore()
def _restore_checkpoint(self) -> int:
"""
Restores the model and training state from the last saved checkpoint.
This includes an epoch count and optimizer state, which is serialized separately
from model parameters. This function should only be used to continue training -
if you wish to load a model for inference/load parts of a model into a new
computation graph, you should use the native Pytorch functions:
`` model.load_state_dict(torch.load("/path/to/model/weights.th"))``
If ``self._serialization_dir`` does not exist or does not contain any checkpointed weights,
this function will do nothing and return 0.
Returns
-------
epoch: int
The epoch at which to resume training, which should be one after the epoch
in the saved training state.
"""
model_state, training_state = self._checkpointer.restore_checkpoint()
if not training_state:
# No checkpoint to restore, start at 0
return 0
self.model.load_state_dict(model_state)
self.optimizer.load_state_dict(training_state["optimizer"])
if self._learning_rate_scheduler is not None and "learning_rate_scheduler" in training_state:
self._learning_rate_scheduler.load_state_dict(
training_state["learning_rate_scheduler"])
if self._momentum_scheduler is not None and "momentum_scheduler" in training_state:
self._momentum_scheduler.load_state_dict(
training_state["momentum_scheduler"])
training_util.move_optimizer_to_cuda(self.optimizer)
# Currently the ``training_state`` contains a serialized ``MetricTracker``.
if "metric_tracker" in training_state:
self._metric_tracker.load_state_dict(
training_state["metric_tracker"])
# It used to be the case that we tracked ``val_metric_per_epoch``.
elif "val_metric_per_epoch" in training_state:
self._metric_tracker.clear()
self._metric_tracker.add_metrics(
training_state["val_metric_per_epoch"])
# And before that we didn't track anything.
else:
self._metric_tracker.clear()
if isinstance(training_state["epoch"], int):
epoch_to_return = training_state["epoch"] + 1
else:
epoch_to_return = int(training_state["epoch"].split('.')[0]) + 1
# For older checkpoints with batch_num_total missing, default to old behavior where
# it is unchanged.
batch_num_total = training_state.get('batch_num_total')
if batch_num_total is not None:
self._batch_num_total = batch_num_total
return epoch_to_return
# Requires custom from_params.
@classmethod
def from_params(
cls, # type: ignore
params: Params,
serialization_dir: str,
recover: bool,
cache_directory: str,
cache_prefix: str) -> 'MatchingTrainer':
# pylint: disable=arguments-differ
pieces = TrainerPieces.from_params(
params, # pylint: disable=no-member
serialization_dir,
recover,
cache_directory,
cache_prefix)
model = pieces.model
iterator = pieces.iterator
train_data = pieces.train_dataset
validation_data = pieces.validation_dataset
params = pieces.params
validation_iterator = pieces.validation_iterator
patience = params.pop_int("patience", None)
validation_metric = params.pop("validation_metric", "-loss")
shuffle = params.pop_bool("shuffle", True)
num_epochs = params.pop_int("num_epochs", 20)
cuda_device = parse_cuda_device(params.pop("cuda_device", -1))
grad_norm = params.pop_float("grad_norm", None)
grad_clipping = params.pop_float("grad_clipping", None)
lr_scheduler_params = params.pop("learning_rate_scheduler", None)
momentum_scheduler_params = params.pop("momentum_scheduler", None)
retrieve_text = params.pop("retrieve_text", True)
if isinstance(cuda_device, list):
model_device = cuda_device[0]
else:
model_device = cuda_device
if model_device >= 0:
# Moving model to GPU here so that the optimizer state gets constructed on
# the right device.
model = model.cuda(model_device)
parameters = [[n, p] for n, p in model.named_parameters()
if p.requires_grad]
optimizer = Optimizer.from_params(parameters, params.pop("optimizer"))
if "moving_average" in params:
moving_average = MovingAverage.from_params(
params.pop("moving_average"), parameters=parameters)
else:
moving_average = None
if lr_scheduler_params:
lr_scheduler = LearningRateScheduler.from_params(
optimizer, lr_scheduler_params)
else:
lr_scheduler = None
if momentum_scheduler_params:
momentum_scheduler = MomentumScheduler.from_params(
optimizer, momentum_scheduler_params)
else:
momentum_scheduler = None
if 'checkpointer' in params:
if 'keep_serialized_model_every_num_seconds' in params or \
'num_serialized_models_to_keep' in params:
raise ConfigurationError(
"Checkpointer may be initialized either from the 'checkpointer' key or from the "
"keys 'num_serialized_models_to_keep' and 'keep_serialized_model_every_num_seconds'"
" but the passed config uses both methods.")
checkpointer = Checkpointer.from_params(params.pop("checkpointer"))
else:
num_serialized_models_to_keep = params.pop_int(
"num_serialized_models_to_keep", 20)
keep_serialized_model_every_num_seconds = params.pop_int(
"keep_serialized_model_every_num_seconds", None)
checkpointer = Checkpointer(
serialization_dir=serialization_dir,
num_serialized_models_to_keep=num_serialized_models_to_keep,
keep_serialized_model_every_num_seconds=
keep_serialized_model_every_num_seconds)
model_save_interval = params.pop_float("model_save_interval", None)
summary_interval = params.pop_int("summary_interval", 100)
histogram_interval = params.pop_int("histogram_interval", None)
should_log_parameter_statistics = params.pop_bool(
"should_log_parameter_statistics", True)
should_log_learning_rate = params.pop_bool("should_log_learning_rate",
False)
log_batch_size_period = params.pop_int("log_batch_size_period", None)
params.assert_empty(cls.__name__)
return cls(
model,
optimizer,
iterator,
train_data,
validation_data,
patience=patience,
retrieve_text=retrieve_text,
validation_metric=validation_metric,
validation_iterator=validation_iterator,
shuffle=shuffle,
num_epochs=num_epochs,
serialization_dir=serialization_dir,
cuda_device=cuda_device,
grad_norm=grad_norm,
grad_clipping=grad_clipping,
learning_rate_scheduler=lr_scheduler,
momentum_scheduler=momentum_scheduler,
checkpointer=checkpointer,
model_save_interval=model_save_interval,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate,
log_batch_size_period=log_batch_size_period,
moving_average=moving_average)
class DistributeTrainer(DistributedTrainerBase):
"""
NOTE: only work in nprocess_ngpus
"""
def __init__(
self,
rank: int,
worldsize: int,
ngpus_per_node: int,
cuda_device: Union[int, List],
model: Model,
optimizer: torch.optim.Optimizer,
iterator: DataIterator,
train_dataset: Iterable[Instance],
validation_dataset: Optional[Iterable[Instance]] = None,
patience: Optional[int] = None,
validation_metric: str = "-loss",
validation_iterator: DataIterator = None,
shuffle: bool = True,
num_epochs: int = 20,
serialization_dir: Optional[str] = None,
num_serialized_models_to_keep: int = 20,
keep_serialized_model_every_num_seconds: int = None,
checkpointer: Checkpointer = None,
model_save_interval: float = None,
grad_norm: Optional[float] = None,
grad_clipping: Optional[float] = None,
learning_rate_scheduler: Optional[LearningRateScheduler] = None,
momentum_scheduler: Optional[MomentumScheduler] = None,
summary_interval: int = 100,
histogram_interval: int = None,
should_log_parameter_statistics: bool = True,
should_log_learning_rate: bool = False,
log_batch_size_period: Optional[int] = None,
moving_average: Optional[MovingAverage] = None) -> None:
super().__init__(rank, worldsize, ngpus_per_node, cuda_device,
serialization_dir)
self.model = model
self.iterator = iterator
self._validation_iterator = validation_iterator
self.shuffle = shuffle
self.optimizer = optimizer
self.train_data = train_dataset
self._validation_data = validation_dataset
self._metric_tracker = MetricTracker(patience, validation_metric)
self._validation_metric = validation_metric[1:]
self._num_epochs = num_epochs
# NOTE: although We have ckpter for everyone, only rank 0 of each node should be able to ckpt
if checkpointer is not None:
self._checkpointer = checkpointer
else:
self._checkpointer = Checkpointer(
serialization_dir, keep_serialized_model_every_num_seconds,
num_serialized_models_to_keep)
self._model_save_interval = model_save_interval
self._grad_norm = grad_norm
self._grad_clipping = grad_clipping
self._learning_rate_scheduler = learning_rate_scheduler
self._momentum_scheduler = momentum_scheduler
self._moving_average = moving_average
# We keep the total batch number as an instance variable because it
# is used inside a closure for the hook which logs activations in
# ``_enable_activation_logging``.
self._batch_num_total = 0
# NOTE: log.
serialization_dir = os.path.join(serialization_dir, str(rank))
self._tensorboard = TensorboardWriter(
get_batch_num_total=lambda: self._batch_num_total,
serialization_dir=serialization_dir,
summary_interval=summary_interval,
histogram_interval=histogram_interval,
should_log_parameter_statistics=should_log_parameter_statistics,
should_log_learning_rate=should_log_learning_rate)
self._log_batch_size_period = log_batch_size_period
self._last_log = 0.0 # time of last logging
# Enable activation logging.
if histogram_interval is not None:
self._tensorboard.enable_activation_logging(self.model)
def rescale_gradients(self) -> Optional[float]:
return training_util.rescale_gradients(self.model, self._grad_norm)
def batch_loss(self, batch_group: List[TensorDict],
for_training: bool) -> torch.Tensor:
assert len(batch_group) == 1
batch = batch_group[0]
batch = nn_util.move_to_device(batch, self._cuda_device[0])
output_dict = self.model(**batch)
try:
loss = output_dict["loss"]
if for_training:
loss += self.model.get_regularization_penalty()
except KeyError:
if for_training:
raise RuntimeError(
"The model you are trying to optimize does not contain a"
" 'loss' key in the output of model.forward(inputs).")
loss = None
return loss
def _train_epoch(self, epoch: int) -> Dict[str, float]:
"""
Trains one epoch and returns metrics.
only report system utils when we are local rank 0 at each machine.
"""
logger.info("Rank %d: Epoch %d/%d", self._rank, epoch,
self._num_epochs - 1)
peak_cpu_usage = peak_memory_mb()
if self._is_chief:
logger.info(f"Peak CPU memory usage MB: {peak_cpu_usage}")
train_loss = 0.0
# Set the model to "train" mode.
self.model.train()
# should be 1 anyway, because we are only dealing with nprocess_with_ngpus
num_gpus = len(self._cuda_device)
# TODO: Implementation of whether the generator should take into account of worldsize.
# Get tqdm for the training batches
raw_train_generator = self.iterator(self.train_data,
num_epochs=1,
shuffle=self.shuffle)
train_generator = lazy_groups_of(raw_train_generator, num_gpus)
num_training_batches = math.ceil(
self.iterator.get_num_batches(self.train_data) / num_gpus)
self._last_log = time.time()
last_save_time = time.time()
batches_this_epoch = 0
if self._batch_num_total is None:
self._batch_num_total = 0
histogram_parameters = set(
self.model.get_parameters_for_histogram_tensorboard_logging())
logger.info("Training")
train_generator_tqdm = Tqdm.tqdm(train_generator,
total=num_training_batches)
cumulative_batch_size = 0
# NOTE: only work in nprocess_ngpus
device = torch.device("cuda:%d" % self._cuda_device[0])
for batch_group in train_generator_tqdm:
batches_this_epoch += 1
self._batch_num_total += 1
batch_num_total = self._batch_num_total
self.optimizer.zero_grad()
loss = self.batch_loss(batch_group, for_training=True)
if torch.isnan(loss):
raise ValueError("nan loss encountered")
loss.backward()
train_loss += loss.item()
batch_grad_norm = self.rescale_gradients()
# This does nothing if batch_num_total is None or you are using a
# scheduler which doesn't update per batch.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step_batch(batch_num_total)
if self._momentum_scheduler:
self._momentum_scheduler.step_batch(batch_num_total)
if self._is_chief:
# only chief do tensorboard
if self._tensorboard.should_log_histograms_this_batch():
# get the magnitude of parameter updates for logging
# We need a copy of current parameters to compute magnitude of updates,
# and copy them to CPU so large models won't go OOM on the GPU.
param_updates = {
name: param.detach().cpu().clone()
for name, param in self.model.named_parameters()
}
self.optimizer.step()
for name, param in self.model.named_parameters():
param_updates[name].sub_(param.detach().cpu())
update_norm = torch.norm(param_updates[name].view(
-1, ))
param_norm = torch.norm(param.view(-1, )).cpu()
self._tensorboard.add_train_scalar(
"gradient_update/" + name,
update_norm / (param_norm + 1e-7))
else:
self.optimizer.step()
else:
self.optimizer.step()
# Update moving averages
# NOTE: not sure whether this need to be average
if self._moving_average is not None:
self._moving_average.apply(batch_num_total)
metrics = get_metrics(self.model, device, self._worldsize,
train_loss, batches_this_epoch)
description = training_util.description_from_metrics(metrics)
train_generator_tqdm.set_description(
("Rank %d: " % self._rank) + description, refresh=False)
if self._is_chief:
# Log parameter values to Tensorboard
if self._tensorboard.should_log_this_batch():
self._tensorboard.log_parameter_and_gradient_statistics(
self.model, batch_grad_norm)
self._tensorboard.log_learning_rates(
self.model, self.optimizer)
self._tensorboard.add_train_scalar("loss/loss_train",
metrics["loss"])
self._tensorboard.log_metrics(
{"epoch_metrics/" + k: v
for k, v in metrics.items()})
if self._tensorboard.should_log_histograms_this_batch():
self._tensorboard.log_histograms(self.model,
histogram_parameters)
if self._log_batch_size_period:
cur_batch = sum([
training_util.get_batch_size(batch)
for batch in batch_group
])
cumulative_batch_size += cur_batch
if (batches_this_epoch - 1) % self._log_batch_size_period == 0:
average = cumulative_batch_size / batches_this_epoch
logger.info(
f"rank {self._rank}, current batch size: {cur_batch} mean batch size: {average}"
)
if self._is_chief:
self._tensorboard.add_train_scalar(
"current_batch_size", cur_batch)
self._tensorboard.add_train_scalar(
"mean_batch_size", average)
if self._is_chief:
# Save model if needed.
if self._model_save_interval is not None and (
time.time() - last_save_time >
self._model_save_interval):
last_save_time = time.time()
self._save_checkpoint('{0}.{1}'.format(
epoch, training_util.time_to_str(int(last_save_time))))
metrics = get_metrics(self.model, device, self._worldsize,
train_loss, batches_this_epoch)
metrics['cpu_memory_MB'] = peak_cpu_usage
return metrics
def _validation_loss(self) -> Tuple[float, int]:
"""
Computes the validation loss. Returns it and the number of batches.
"""
logger.info("Rank %d Validating", self._rank)
self.model.eval()
# Replace parameter values with the shadow values from the moving averages.
if self._moving_average is not None:
self._moving_average.assign_average_value()
if self._validation_iterator is not None:
val_iterator = self._validation_iterator
else:
val_iterator = self.iterator
num_gpus = len(self._cuda_device)
raw_val_generator = val_iterator(self._validation_data,
num_epochs=1,
shuffle=False)
val_generator = lazy_groups_of(raw_val_generator, num_gpus)
num_validation_batches = math.ceil(
val_iterator.get_num_batches(self._validation_data) / num_gpus)
val_generator_tqdm = Tqdm.tqdm(val_generator,
total=num_validation_batches)
batches_this_epoch = 0
val_loss = 0
for batch_group in val_generator_tqdm:
loss = self.batch_loss(batch_group, for_training=False)
if loss is not None:
# You shouldn't necessarily have to compute a loss for validation, so we allow for
# `loss` to be None. We need to be careful, though - `batches_this_epoch` is
# currently only used as the divisor for the loss function, so we can safely only
# count those batches for which we actually have a loss. If this variable ever
# gets used for something else, we might need to change things around a bit.
batches_this_epoch += 1
val_loss += loss.detach().cpu().numpy()
# Update the description with the latest metrics
val_metrics = training_util.get_metrics(self.model, val_loss,
batches_this_epoch)
description = training_util.description_from_metrics(val_metrics)
val_generator_tqdm.set_description(description, refresh=False)
# Now restore the original parameter values.
if self._moving_average is not None:
self._moving_average.restore()
return val_loss, batches_this_epoch
def train(self) -> Dict[str, Any]:
"""
Trains the supplied model with the supplied parameters.
"""
try:
epoch_counter = self._restore_checkpoint()
except RuntimeError:
traceback.print_exc()
raise ConfigurationError(
"Could not recover training from the checkpoint. Did you mean to output to "
"a different serialization directory or delete the existing serialization "
"directory?")
training_util.enable_gradient_clipping(self.model, self._grad_clipping)
logger.info("Rank %d Beginning training.", self._rank)
train_metrics: Dict[str, float] = {}
val_metrics: Dict[str, float] = {}
this_epoch_val_metric: float = None
metrics: Dict[str, Any] = {}
epochs_trained = 0
training_start_time = time.time()
metrics['best_epoch'] = self._metric_tracker.best_epoch
for key, value in self._metric_tracker.best_epoch_metrics.items():
metrics["best_validation_" + key] = value
for epoch in range(epoch_counter, self._num_epochs):
epoch_start_time = time.time()
train_metrics = self._train_epoch(epoch)
# get peak of memory usage
if self._is_chief:
if 'cpu_memory_MB' in train_metrics:
metrics['peak_cpu_memory_MB'] = max(
metrics.get('peak_cpu_memory_MB', 0),
train_metrics['cpu_memory_MB'])
if self._validation_data is not None:
with torch.no_grad():
# We have a validation set, so compute all the metrics on it.
val_loss, num_batches = self._validation_loss()
val_metrics = training_util.get_metrics(self.model,
val_loss,
num_batches,
reset=True)
# Check validation metric for early stopping
this_epoch_val_metric = val_metrics[
self._validation_metric]
self._metric_tracker.add_metric(this_epoch_val_metric)
if self._metric_tracker.should_stop_early():
logger.info(
"Ran out of patience. Stopping training.")
break
if self._is_chief:
self._tensorboard.log_metrics(
train_metrics,
val_metrics=val_metrics,
log_to_console=True,
epoch=epoch +
1) # +1 because tensorboard doesn't like 0
# Create overall metrics dict
training_elapsed_time = time.time() - training_start_time
metrics["training_duration"] = str(
datetime.timedelta(seconds=training_elapsed_time))
metrics["training_start_epoch"] = epoch_counter
metrics["training_epochs"] = epochs_trained
metrics["epoch"] = epoch
for key, value in train_metrics.items():
metrics["training_" + key] = value
for key, value in val_metrics.items():
metrics["validation_" + key] = value
if self._metric_tracker.is_best_so_far() and self._is_chief:
# Update all the best_ metrics.
# (Otherwise they just stay the same as they were.)
metrics['best_epoch'] = epoch
for key, value in val_metrics.items():
metrics["best_validation_" + key] = value
self._metric_tracker.best_epoch_metrics = val_metrics
if self._serialization_dir and self._is_chief:
dump_metrics(
os.path.join(self._serialization_dir,
f'metrics_epoch_{epoch}.json'), metrics)
# The Scheduler API is agnostic to whether your schedule requires a validation metric -
# if it doesn't, the validation metric passed here is ignored.
if self._learning_rate_scheduler:
self._learning_rate_scheduler.step(this_epoch_val_metric,
epoch)
if self._momentum_scheduler:
self._momentum_scheduler.step(this_epoch_val_metric, epoch)
if self._is_chief:
self._save_checkpoint(epoch)
epoch_elapsed_time = time.time() - epoch_start_time
logger.info("Rank %d Epoch duration: %s", self._rank,
datetime.timedelta(seconds=epoch_elapsed_time))
if epoch < self._num_epochs - 1:
training_elapsed_time = time.time() - training_start_time
estimated_time_remaining = training_elapsed_time * \
((self._num_epochs - epoch_counter) / float(epoch - epoch_counter + 1) - 1)
formatted_time = str(
datetime.timedelta(
seconds=int(estimated_time_remaining)))
logger.info(
"Rank %d, Estimated training time remaining: %s",
self._rank, formatted_time)
epochs_trained += 1
# make sure pending events are flushed to disk and files are closed properly
self._tensorboard.close()
# Load the best model state before returning
if self._is_chief:
best_model_state = self._checkpointer.best_model_state()
if best_model_state:
self.model.load_state_dict(best_model_state)
return metrics
def _save_checkpoint(self, epoch: Union[int, str]) -> None:
"""
Saves a checkpoint of the model to self._serialization_dir.
Is a no-op if self._serialization_dir is None.
Parameters
----------
epoch : Union[int, str], required.
The epoch of training. If the checkpoint is saved in the middle
of an epoch, the parameter is a string with the epoch and timestamp.
"""
# If moving averages are used for parameters, we save
# the moving average values into checkpoint, instead of the current values.
if self._moving_average is not None:
self._moving_average.assign_average_value()
# These are the training states we need to persist.
training_states = {
"metric_tracker": self._metric_tracker.state_dict(),
"optimizer": self.optimizer.state_dict(),
"batch_num_total": self._batch_num_total
}
# If we have a learning rate or momentum scheduler, we should persist them too.
if self._learning_rate_scheduler is not None:
training_states[
"learning_rate_scheduler"] = self._learning_rate_scheduler.state_dict(
)
if self._momentum_scheduler is not None:
training_states[
"momentum_scheduler"] = self._momentum_scheduler.state_dict()
self._checkpointer.save_checkpoint(
model_state=self.model.state_dict(),
epoch=epoch,
training_states=training_states,
is_best_so_far=self._metric_tracker.is_best_so_far())
# Restore the original values for parameters so that training will not be affected.
if self._moving_average is not None:
self._moving_average.restore()
def _restore_checkpoint(self) -> int:
"""
Restores the model and training state from the last saved checkpoint.
This includes an epoch count and optimizer state, which is serialized separately
from model parameters. This function should only be used to continue training -
if you wish to load a model for inference/load parts of a model into a new
computation graph, you should use the native Pytorch functions:
`` model.load_state_dict(torch.load("/path/to/model/weights.th"))``
If ``self._serialization_dir`` does not exist or does not contain any checkpointed weights,
this function will do nothing and return 0.
Returns
-------
epoch: int
The epoch at which to resume training, which should be one after the epoch
in the saved training state.
"""
model_state, training_state = self._checkpointer.restore_checkpoint()
if not training_state:
# No checkpoint to restore, start at 0
return 0
self.model.load_state_dict(model_state)
self.optimizer.load_state_dict(training_state["optimizer"])
if self._learning_rate_scheduler is not None and "learning_rate_scheduler" in training_state:
self._learning_rate_scheduler.load_state_dict(
training_state["learning_rate_scheduler"])
if self._momentum_scheduler is not None and "momentum_scheduler" in training_state:
self._momentum_scheduler.load_state_dict(
training_state["momentum_scheduler"])
training_util.move_optimizer_to_cuda(self.optimizer)
# Currently the ``training_state`` contains a serialized ``MetricTracker``.
if "metric_tracker" in training_state:
self._metric_tracker.load_state_dict(
training_state["metric_tracker"])
# It used to be the case that we tracked ``val_metric_per_epoch``.
elif "val_metric_per_epoch" in training_state:
self._metric_tracker.clear()
self._metric_tracker.add_metrics(
training_state["val_metric_per_epoch"])
# And before that we didn't track anything.
else:
self._metric_tracker.clear()
if isinstance(training_state["epoch"], int):
epoch_to_return = training_state["epoch"] + 1
else:
epoch_to_return = int(training_state["epoch"].split('.')[0]) + 1
# For older checkpoints with batch_num_total missing, default to old behavior where
# it is unchanged.
batch_num_total = training_state.get('batch_num_total')
if batch_num_total is not None:
self._batch_num_total = batch_num_total
return epoch_to_return
