class BaseTrainer:
def __init__(self,
model,
loss,
resume,
config,
train_loader,
val_loader=None,
train_logger=None):
self.model = model
self.loss = loss
self.config = config
self.train_loader = train_loader
self.val_loader = val_loader
self.train_logger = train_logger
self.logger = logging.getLogger(self.__class__.__name__)
self.do_validation = self.config['val']['is_val']
self.start_epoch = 1
self.improved = False
self.MEAN = config['dataset']['mean']
self.STD = config['dataset']['std']
# SETTING THE DEVICE
self.device, available_gpus = self._get_available_devices(
config['train']['n_gpu'])
if len(available_gpus) > 1:
self.model = torch.nn.DataParallel(self.model,
device_ids=available_gpus)
self.loss = torch.nn.DataParallel(self.loss,
device_ids=available_gpus)
self.model.to(self.device)
self.loss.to(self.device)
# CONFIG
self.epochs = self.config['train']['epochs']
self.save_period = self.config['train']['save_period']
writer_base_dir = config['train']['writer_dir']
check_dir(writer_base_dir)
writer_dir = os.path.join(
writer_base_dir, config['name'] + "_" + config['dataset']['name'])
check_dir(writer_dir)
self.writer = tensorboard.SummaryWriter(str(writer_dir))
# OPTIMIZER
if self.config['optimizer']['differential_lr']:
if isinstance(self.model, torch.nn.DataParallel):
trainable_params = [{
'params':
filter(lambda p: p.requires_grad,
self.model.module.get_decoder_params())
}, {
'params':
filter(lambda p: p.requires_grad,
self.model.module.get_backbone_params()),
'lr':
config['optimizer']['args']['lr'] / 10
}]
else:
trainable_params = [{
'params':
filter(lambda p: p.requires_grad,
self.model.get_decoder_params())
}, {
'params':
filter(lambda p: p.requires_grad,
self.model.get_backbone_params()),
'lr':
config['optimizer']['args']['lr'] / 10
}]
else:
trainable_params = filter(lambda p: p.requires_grad,
self.model.parameters())
self.optimizer = get_instance(torch.optim, 'optimizer', config,
trainable_params)
self.lr_scheduler = getattr(
lr_scheduler, config['lr_scheduler']['name'])(self.optimizer,
self.epochs,
len(train_loader))
# monitor
self.monitor = self.config['train']['monitor']['is_monitor']
if not self.monitor:
self.mnt_mode = None
self.mnt_best = 0
else:
self.mnt_mode = self.config['train']['monitor']['type']
self.mnt_metric = self.config['train']['monitor']['metric']
self.mnt_best = -math.inf if self.mnt_mode == 'max' else math.inf
self.early_stoping = self.config['train']['monitor']['early_stop']
self.not_improved_count = 0
# CHECKPOINTS
start_time = datetime.datetime.now().strftime('%m-%d_%H-%M')
checkpoint_base_dir = config['train']['save_dir']
check_dir(checkpoint_base_dir)
self.checkpoint_dir = os.path.join(
checkpoint_base_dir, config['name'] + "_" +
config['dataset']['name'] + "_" + start_time)
check_dir(self.checkpoint_dir)
def _get_available_devices(self, n_gpu):
sys_gpu = torch.cuda.device_count()
if sys_gpu == 0:
self.logger.warning('No GPUs detected, using the CPU')
n_gpu = 0
elif n_gpu > sys_gpu:
self.logger.warning(
'Nbr of GPU requested is {} but only {} are available'.format(
n_gpu, sys_gpu))
n_gpu = sys_gpu
device = torch.device('cuda:0' if n_gpu > 0 else 'cpu')
self.logger.info('Detected GPUs: {} Requested: {}'.format(
sys_gpu, n_gpu))
available_gpus = list(range(n_gpu))
return device, available_gpus
def _save_checkpoint(self, epoch, save_best=False):
state = {
'arch': type(self.model).__name__,
'epoch': epoch,
'logger': self.train_logger,
'state_dict': self.model.state_dict(),
'optimizer': self.optimizer.state_dict(),
'monitor_best': self.mnt_best,
'config': self.config
}
filename = str(
self.checkpoint_dir) + "/checkpoint-epoch" + str(epoch) + ".pth"
self.logger.info('\nSaving a checkpoint: {} ***'.format(filename))
torch.save(state, filename)
if save_best:
filename = os.path.join(self.checkpoint_dir, "best_epoch.pth")
self.logger.info("Saveing a best epoch")
torch.save(state, str(filename))
def _resume_checkpoint(self, resume_path):
self.logger.info('Loading checkpoint : {}'.format(resume_path))
checkpoint = torch.load(resume_path)
self.start_epoch = checkpoint['epoch'] + 1
self.mnt_best = checkpoint['monitor_best']
self.not_improved_count = 0
if checkpoint['config']['model'] != self.config['model']:
self.logger.warning({
'Warning! Current model is not the same as the one in the checkpoint'
})
self.model.load_state_dict(checkpoint['state_dict'])
if checkpoint['config']['optimizer']['name'] != self.config[
'optimizer']['name']:
self.logger.warning({
'Warning! Current model is not the same as the one in the checkpoint'
})
self.model.load_state_dict(checkpoint['optimizer'])
self.train_logger = checkpoint['logger']
self.logger.info('Checkpoint <{}> (epoch {}) was loaded'.format(
resume_path, self.start_epoch))
def train(self):
for epoch in range(self.start_epoch, self.epochs + 1):
results = self._train_epoch(epoch)
if self.do_validation and epoch % self.config['val'][
'val_per_epochs'] == 0:
results = self._valid_epoch(epoch)
self.logger.info('\n*** val {} epoch ***'.format(epoch))
for k, v in results.items():
self.logger.info(' {}: {}'.format(k, v))
if self.train_logger is not None:
log = {'epoch': epoch, **results}
self.train_logger.add_entry(log)
else:
self.train_logger = Logger()
log = {'epoch': epoch, **results}
self.train_logger.add_entry(log)
if self.mnt_mode is not None and epoch % self.config['val'][
'val_per_epochs'] == 0:
try:
if self.mnt_mode == 'min':
self.improved = (log[self.mnt_metric] < self.mnt_best)
else:
self.improved = (log[self.mnt_metric] > self.mnt_best)
except KeyError:
self.logger.warning(
'The metrics being tracked ({}) has not been calculated. Training stops.'
.format(self.mnt_metric))
break
if self.improved:
self.mnt_best = log[self.mnt_metric]
self.not_improved_count = 0
else:
self.not_improved_count += 1
if self.not_improved_count > self.early_stoping:
self.logger.info(
'\nPerformance didn\'t improve for {} epochs'.format(
self.early_stoping))
self.logger.warning('Training Stop')
break
if epoch % self.save_period == 0:
self._save_checkpoint(epoch, save_best=self.improved)
def _train_epoch(self, epoch):
raise NotImplementedError
def _valid_epoch(self, epoch):
raise NotImplementedError
def _eval_metrics(self, output, target):
raise NotImplementedError
class BaseTrainer(object):
def __init__(self, model, losses, metrics, optimizer, resume, config, data_loader,
valid_data_loader=None, lr_scheduler=None, grad_clip=None):
self.config = config
self.data_loader = data_loader
self.valid_data_loader = valid_data_loader
self.do_validation = self.valid_data_loader is not None
self.lr_scheduler = lr_scheduler
self.lr_schedule_by_epoch = config['lr_scheduler']['by_epoch']
self.grad_clip = grad_clip
# Setup directory for checkpoint saving
start_time = datetime.datetime.now().strftime('%m%d_%H%M%S')
self.checkpoint_dir = os.path.join(config['trainer']['save_dir'], config['name'], start_time)
os.makedirs(self.checkpoint_dir, exist_ok=True)
# Build logger
logname = self.config['name']
logfile = os.path.join(self.checkpoint_dir, "logging.log")
self.logger = Logger(logname, logfile)
# Setup GPU device if available, move model into configured device
self.device, device_ids = self._prepare_device(config['n_gpu'])
self.model = model.to(self.device)
if len(device_ids) > 1:
self.model = torch.nn.DataParallel(model, device_ids=device_ids)
self.losses = losses
self.metrics = metrics
self.optimizer = optimizer
self.epochs = config['trainer']['epochs']
self.save_freq = config['trainer']['save_freq']
self.verbosity = config['trainer']['verbosity']
self.logger.info("Total epochs: {}".format(self.epochs))
# configuration to monitor model performance and save best
self.monitor = config['trainer']['monitor']
self.monitor_mode = config['trainer']['monitor_mode']
assert self.monitor_mode in ['min', 'max', 'off']
self.monitor_best = math.inf if self.monitor_mode == 'min' else -math.inf
self.start_epoch = 1
# setup visualization writer instance
writer_train_dir = os.path.join(config['visualization']['log_dir'], config['name'], start_time, "train")
writer_valid_dir = os.path.join(config['visualization']['log_dir'], config['name'], start_time, "valid")
self.writer_train = WriterTensorboardX(writer_train_dir, self.logger, config['visualization']['tensorboardX'])
self.writer_valid = WriterTensorboardX(writer_valid_dir, self.logger, config['visualization']['tensorboardX'])
# Save configuration file into checkpoint directory
config_save_path = os.path.join(self.checkpoint_dir, 'config.json')
with open(config_save_path, 'w') as handle:
json.dump(config, handle, indent=4, sort_keys=False)
# Resume
if resume:
self._resume_checkpoint(resume)
def train(self):
for epoch in range(self.start_epoch, self.epochs + 1):
print("----------------------------------------------------------------")
self.logger.info("[EPOCH %d/%d]" % (epoch, self.epochs))
start_time = time()
result = self._train_epoch(epoch)
finish_time = time()
self.logger.info("Finish at {}, Runtime: {:.3f} [s]".format(datetime.datetime.now(), finish_time-start_time))
# save logged informations into log dict
log = {}
for key, value in result.items():
if key == 'train_metrics':
log.update({'train_' + mtr.func.__name__ : value[i] for i, mtr in enumerate(self.metrics)})
elif key == 'valid_metrics':
log.update({'valid_' + mtr.func.__name__ : value[i] for i, mtr in enumerate(self.metrics)})
else:
log[key] = value
# print logged informations to the screen
if self.logger is not None:
self.logger.add_entry(log)
if self.verbosity >= 1:
for key, value in sorted(list(log.items())):
self.logger.info('{:15s}: {}'.format(str(key), value))
# evaluate model performance according to configured metric, save best checkpoint as model_best
best = False
if self.monitor_mode != 'off':
try:
if (self.monitor_mode == 'min' and log[self.monitor] < self.monitor_best) or\
(self.monitor_mode == 'max' and log[self.monitor] > self.monitor_best):
self.logger.info("Monitor improved from %f to %f" % (self.monitor_best, log[self.monitor]))
self.monitor_best = log[self.monitor]
best = True
except KeyError:
if epoch == 1:
msg = "Warning: Can\'t recognize metric named '{}' ".format(self.monitor)\
+ "for performance monitoring. model_best checkpoint won\'t be updated."
self.logger.warning(msg)
# Save checkpoint
self._save_checkpoint(epoch, save_best=best)
def _train_epoch(self, epoch):
self.logger.info("Train on epoch...")
self.model.train()
self.writer_train.set_step(epoch)
# Perform training
total_loss = 0
total_metrics = np.zeros(len(self.metrics))
n_iter = len(self.data_loader)
train_pbar = tqdm(enumerate(self.data_loader), total=n_iter)
for batch_idx, data in train_pbar:
# Send data to device
for key, value in data.items():
data[key] = value.to(self.device)
# Forward and Backward
output, losses, loss = self._forward(data)
self._backward(loss)
# Learning rate scheduler by iteration
if self.lr_scheduler is not None and not self.lr_schedule_by_epoch:
self.lr_scheduler.step()
# Accumulate loss and metrics
loss_iter = loss.item()
total_loss += loss_iter
metrics_iter = self._eval_metrics(output, data)
total_metrics += metrics_iter
# Visualize results
if (batch_idx==n_iter-2) and (self.verbosity>=2):
self._visualize_results(output, data)
# tqdm progress bar
if self.verbosity>=1:
pbar_dict = self._get_progress_bar_dict(losses, loss_iter, metrics_iter)
train_pbar.set_postfix(**pbar_dict)
# Learning rate scheduler by epoch
if self.lr_scheduler is not None and self.lr_schedule_by_epoch:
self.lr_scheduler.step()
# Record log
total_loss /= len(self.data_loader)
total_metrics /= len(self.data_loader)
log = {'train_loss': total_loss, 'train_metrics': total_metrics.tolist()}
# Write training result to TensorboardX
self.writer_train.add_scalar('loss', total_loss)
for i, metric in enumerate(self.metrics):
self.writer_train.add_scalar('metrics/%s'%(metric.func.__name__), total_metrics[i])
if self.verbosity>=2:
for i in range(len(self.optimizer.param_groups)):
self.writer_train.add_scalar('lr/group%d'%(i), self.optimizer.param_groups[i]['lr'])
# Perform validating
if self.do_validation:
self.logger.info("Validate on epoch...")
val_log = self._valid_epoch(epoch)
log = {**log, **val_log}
return log
def _valid_epoch(self, epoch):
self.model.eval()
total_val_loss = 0
total_val_metrics = np.zeros(len(self.metrics))
n_iter = len(self.valid_data_loader)
self.writer_valid.set_step(epoch)
with torch.no_grad():
# Validate
for batch_idx, data in tqdm(enumerate(self.valid_data_loader), total=n_iter):
# Send data to device
for key, value in data.items():
data[key] = value.to(self.device)
# Forward
output, _, loss = self._forward(data)
# Accumulate loss and metrics
total_val_loss += loss.item()
total_val_metrics += self._eval_metrics(output, data)
# Visualize results
if (batch_idx==n_iter-2) and (self.verbosity>=2):
self._visualize_results(output, data)
# Record log
total_val_loss /= len(self.valid_data_loader)
total_val_metrics /= len(self.valid_data_loader)
val_log = {
'valid_loss': total_val_loss,
'valid_metrics': total_val_metrics.tolist(),
}
# Write validating result to TensorboardX
self.writer_valid.add_scalar('loss', total_val_loss)
for i, metric in enumerate(self.metrics):
self.writer_valid.add_scalar('metrics/%s'%(metric.func.__name__), total_val_metrics[i])
return val_log
def _forward(self, data):
self.optimizer.zero_grad()
output = self.model(**data)
losses = self.losses(**output, **data)
loss = self._sum_losses(losses)
return output, losses, loss
def _backward(self, loss):
loss.backward()
if self.grad_clip is not None:
clip_grad.clip_grad_norm_(filter(lambda p: p.requires_grad, self.model.parameters()), **self.grad_clip)
self.optimizer.step()
def _eval_metrics(self, output, data):
acc_metrics = np.zeros(len(self.metrics))
for i, metric in enumerate(self.metrics):
acc_metrics[i] += metric(**output, **data)
return acc_metrics
def _sum_losses(self, losses):
loss = sum(loss_val for loss_val in list(losses.values()))
return loss
def _visualize_results(self, output, data):
pass
def _get_progress_bar_dict(self, losses, loss, metrics):
pbar_dict = dict()
pbar_dict['lr'] = self.optimizer.param_groups[0]['lr']
for key, val in losses.items():
pbar_dict['loss_%s'%(key)] = val.item()
pbar_dict['loss'] = loss
if self.verbosity>=3:
for i, metric in enumerate(self.metrics):
pbar_dict['%s'%(metric.func.__name__)] = metrics[i]
return pbar_dict
def _prepare_device(self, n_gpu_use):
"""
setup GPU device if available, move model into configured device
"""
n_gpu = torch.cuda.device_count()
if n_gpu_use > 0 and n_gpu == 0:
self.logger.warning("Warning: There\'s no GPU available on this machine, training will be performed on CPU.")
n_gpu_use = 0
if n_gpu_use > n_gpu:
msg = "Warning: The number of GPU\'s configured to use is {}, but only {} are available on this machine.".format(n_gpu_use, n_gpu)
self.logger.warning(msg)
n_gpu_use = n_gpu
device = torch.device('cuda:0' if n_gpu_use > 0 else 'cpu')
list_ids = list(range(n_gpu_use))
return device, list_ids
def _save_checkpoint(self, epoch, save_best=False):
"""
Saving checkpoints
:param epoch: current epoch number
:param log: logging information of the epoch
:param save_best: if True, rename the saved checkpoint to 'model_best.pth'
"""
# Construct savedict
arch = type(self.model).__name__
state = {
'arch': arch,
'epoch': epoch,
'state_dict': self.model.state_dict(),
'optimizer': self.optimizer.state_dict(),
'monitor_best': self.monitor_best,
'config': self.config
}
# Save checkpoint for each epoch
if self.save_freq is not None: # Use None mode to avoid over disk space with large models
if epoch % self.save_freq == 0:
filename = os.path.join(self.checkpoint_dir, 'epoch{}.pth'.format(epoch))
torch.save(state, filename)
self.logger.info("Saving checkpoint at {}".format(filename))
# Save the best checkpoint
if save_best:
best_path = os.path.join(self.checkpoint_dir, 'model_best.pth')
torch.save(state, best_path)
self.logger.info("Saving current best at {}".format(best_path))
else:
self.logger.info("Monitor is not improved from %f" % (self.monitor_best))
def _resume_checkpoint(self, resume_path):
"""
Resume from saved checkpoints
:param resume_path: Checkpoint path to be resumed
"""
self.logger.info("Loading checkpoint: {}".format(resume_path))
checkpoint = torch.load(resume_path)
self.start_epoch = checkpoint['epoch'] + 1
self.monitor_best = checkpoint['monitor_best']
# load architecture params from checkpoint.
if checkpoint['config']['arch'] != self.config['arch']:
self.logger.warning('Warning: Architecture configuration given in config file is different from that of checkpoint. ' + \
'This may yield an exception while state_dict is being loaded.')
self.model.load_state_dict(checkpoint['state_dict'], strict=True)
# load optimizer state from checkpoint only when optimizer type is not changed.
if checkpoint['config']['optimizer']['type'] != self.config['optimizer']['type']:
self.logger.warning('Warning: Optimizer type given in config file is different from that of checkpoint. ' + \
'Optimizer parameters not being resumed.')
else:
self.optimizer.load_state_dict(checkpoint['optimizer'])
self.logger.info("Checkpoint '{}' (epoch {}) loaded".format(resume_path, self.start_epoch-1))
