def export(model,
filename_or_stream='export.pkl',
pickle_module=pickle,
pickle_protocol=2):
from fastai.torch_core import rank_distrib
import torch
"Export the content of `self` without the items and the optimizer state for inference"
if rank_distrib(): return # don't export if child proc
model._end_cleanup()
old_dbunch = model.dls
model.dls = model.dls.new_empty()
state = model.opt.state_dict() if model.opt is not None else None
model.opt = None
target = open(
model.path / filename_or_stream,
'wb') if is_pathlike(filename_or_stream) else filename_or_stream
with warnings.catch_warnings():
# To avoid the warning that come from PyTorch about model not being checked
warnings.simplefilter("ignore")
torch.save(model,
target,
pickle_module=pickle_module,
pickle_protocol=pickle_protocol)
model.create_opt()
if state is not None:
model.opt.load_state_dict(state)
model.dls = old_dbunch
def before_fit(self):
# Make sure this is a training run
self.run = not hasattr(self.learn, 'lr_finder') and \
not hasattr(self, "gather_preds") and rank_distrib() == 0
if not self.run:
return
# Log Hyper Parameters
params = self.learn.gather_args()
self.logger.log_hyperparams(params)
def before_fit(self):
self.run = (rank_distrib() == 0) and not (hasattr(
self.learn, 'lr_finder') or hasattr(self, "gather_preds"))
if self.run:
self.writer = SummaryWriter(log_dir=self.log_dir)
self.train_metrics = listify(
ifnone(self.train_metrics, self.learn.loss_func))
self.train_metric_names = listify(
ifnone(self.train_metric_names,
[self.get_name(s) for s in self.train_metrics]))
self.train_metric_names = [
'train_' + name for name in self.train_metric_names
]
self.smooth_dict = {
name: (0, 0)
for name in self.train_metric_names
}
def on_batch_end(self, **kwargs) -> None:
if dist.is_initialized() and rank_distrib() != 0: return
if kwargs['train']:
# print('loss', kwargs['last_loss'].item())
self.tb_writer.add_scalar('Train/loss', kwargs['last_loss'].item(),
kwargs['iteration'])
def before_fit(self):
self.run = not hasattr(self.learn, 'lr_finder') and hasattr(self, "gather_preds") and rank_distrib()==0
if self.run:
self._setup()