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.

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

Add "monitor" to saved ModelCheckpoints #4383

Merged
merged 6 commits into from Oct 28, 2020
Merged
Show file tree
Hide file tree
Changes from 3 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
7 changes: 5 additions & 2 deletions CHANGELOG.md
Expand Up @@ -12,7 +12,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
- Added `dirpath` and `filename` parameter in `ModelCheckpoint` ([#4213](https://github.com/PyTorchLightning/pytorch-lightning/pull/4213))


- Added plugins docs and DDPPlugin to customize ddp across all accelerators([#4258](https://github.com/PyTorchLightning/pytorch-lightning/pull/4285))
- Added plugins docs and DDPPlugin to customize ddp across all accelerators ([#4258](https://github.com/PyTorchLightning/pytorch-lightning/pull/4285))


- Added `strict` option to the scheduler dictionary ([#3586](https://github.com/PyTorchLightning/pytorch-lightning/pull/3586))
Expand All @@ -21,7 +21,10 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
- Added `fsspec` support for profilers ([#4162](https://github.com/PyTorchLightning/pytorch-lightning/pull/4162))


- Added autogenerated helptext to `Trainer.add_argparse_args`. ([#4344](https://github.com/PyTorchLightning/pytorch-lightning/pull/4344))
- Added autogenerated helptext to `Trainer.add_argparse_args` ([#4344](https://github.com/PyTorchLightning/pytorch-lightning/pull/4344))


- Added "best_model_monitor" key to saved `ModelCheckpoints` ([#4383](https://github.com/PyTorchLightning/pytorch-lightning/pull/4383))


### Changed
Expand Down
3 changes: 1 addition & 2 deletions pytorch_lightning/callbacks/model_checkpoint.py
Expand Up @@ -131,8 +131,6 @@ class ModelCheckpoint(Callback):

CHECKPOINT_JOIN_CHAR = "-"
CHECKPOINT_NAME_LAST = "last"
CHECKPOINT_STATE_BEST_SCORE = "checkpoint_callback_best_model_score"
CHECKPOINT_STATE_BEST_PATH = "checkpoint_callback_best_model_path"
awaelchli marked this conversation as resolved.
Show resolved Hide resolved

def __init__(
self,
Expand Down Expand Up @@ -187,6 +185,7 @@ def on_validation_end(self, trainer, pl_module):

def on_save_checkpoint(self, trainer, pl_module) -> Dict[str, Any]:
return {
"best_model_monitor": self.monitor,
carmocca marked this conversation as resolved.
Show resolved Hide resolved
"best_model_score": self.best_model_score,
"best_model_path": self.best_model_path,
}
Expand Down
5 changes: 1 addition & 4 deletions tests/checkpointing/test_model_checkpoint.py
Expand Up @@ -507,10 +507,7 @@ def test_model_checkpoint_save_last_checkpoint_contents(tmpdir):
assert all(ckpt_last_epoch[k] == ckpt_last[k] for k in ("epoch", "global_step"))

ch_type = type(model_checkpoint)
assert all(list(
ckpt_last["callbacks"][ch_type][k] == ckpt_last_epoch["callbacks"][ch_type][k]
for k in ("best_model_score", "best_model_path")
))
assert ckpt_last["callbacks"][ch_type] == ckpt_last_epoch["callbacks"][ch_type]

# it is easier to load the model objects than to iterate over the raw dict of tensors
model_last_epoch = EvalModelTemplate.load_from_checkpoint(path_last_epoch)
Expand Down