Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

fix incompatibilites lightning1.9 #328

Merged
merged 5 commits into from
Mar 13, 2023
Merged

fix incompatibilites lightning1.9 #328

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4e22396
adapt lightning 1.9
anhtu293 Mar 12, 2023
dc179f9
fix log condition
anhtu293 Mar 12, 2023
b2dd2b3
wip: better handle should_accumulate()
anhtu293 Mar 12, 2023
f637c0f
update trainer init in lightning 1.9
anhtu293 Mar 12, 2023
489d3b2
remove handle for lightning1.4
anhtu293 Mar 13, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions alonet/callbacks/metrics_callback.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,7 +121,6 @@ def _process_train_metrics(self, outputs):
def _log_train_metrics(self, pl_module, trainer):
# Log the results
for key in self.metrics:

if "histogram" in key and len(self.metrics[key]) > 0:
hist = torch.cat(self.metrics[key]).to("cpu")
log_hist(trainer, f"train/{key}", hist)
Expand All @@ -138,7 +137,7 @@ def _log_train_metrics(self, pl_module, trainer):
)

@rank_zero_only
def on_train_batch_end(self, trainer, pl_module, outputs, batch, batch_idx, dataloader_idx):
def on_train_batch_end(self, trainer, pl_module, outputs, batch, batch_idx):
"""Method called after each training batch. This class is a pytorch lightning callback, therefore
this method will by automatically called by pytorch lightning.

Expand Down Expand Up @@ -176,7 +175,8 @@ def on_train_batch_end(self, trainer, pl_module, outputs, batch, batch_idx, data
)

self._process_train_metrics(outputs)
if trainer.fit_loop.should_accumulate() or (trainer.global_step + 1) % trainer.log_every_n_steps != 0:
should_accumulate = trainer.fit_loop._should_accumulate()
if should_accumulate or (trainer.global_step + 1) % trainer.log_every_n_steps != 0:
return

self._log_train_metrics(pl_module, trainer)
Expand Down
6 changes: 2 additions & 4 deletions alonet/common/pl_helpers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -169,7 +169,6 @@ def load_training(
weights_path = getattr(args, "weights", None) if args is not None else None
no_run_id = args.no_run_id if no_run_id is None and "no_run_id" in args else no_run_id


if "weights" in kwargs and kwargs["weights"] is not None: # Highest priority
weights_path = kwargs["weights"]

Expand Down Expand Up @@ -273,13 +272,12 @@ def run_pl_training(
# Init trainer and run training
trainer = pl.Trainer.from_argparse_args(
args,
# default_root_dir=expe_dir,
gpus=-1 if not args.cpu else 0,
accelerator="gpu" if not args.cpu else "cpu",
auto_select_gpus=not args.cpu,
logger=logger,
callbacks=callbacks,
resume_from_checkpoint=resume_from_checkpoint,
accelerator=None if torch.cuda.device_count() <= 1 else "ddp",
strategy="ddp" if torch.cuda.device_count() >= 2 else None,
**pl_trainer,
)

Expand Down