lightning/tests/tests_pytorch/models/test_hooks.py

959 lines
38 KiB
Python

# Copyright The PyTorch Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from functools import partial, update_wrapper
from inspect import getmembers, isfunction
from unittest import mock
from unittest.mock import ANY, PropertyMock
import pytest
import torch
from torch.utils.data import DataLoader
from pytorch_lightning import __version__, Callback, LightningDataModule, LightningModule, Trainer
from pytorch_lightning.demos.boring_classes import BoringDataModule, BoringModel, RandomDataset
from tests_pytorch.helpers.runif import RunIf
class HookedDataModule(BoringDataModule):
def __init__(self, called):
super().__init__()
def call(hook, fn, *args, **kwargs):
out = fn(*args, **kwargs)
d = {"name": hook}
if args:
d["args"] = args
if kwargs:
d["kwargs"] = kwargs
called.append(d)
return out
for h in get_members(LightningDataModule):
attr = getattr(self, h)
partial_h = partial(call, h, attr)
update_wrapper(partial_h, attr)
setattr(self, h, partial_h)
@pytest.mark.parametrize("max_steps", [1, 2, 3])
def test_on_before_zero_grad_called(tmpdir, max_steps):
class CurrentTestModel(BoringModel):
on_before_zero_grad_called = 0
def on_before_zero_grad(self, optimizer):
self.on_before_zero_grad_called += 1
model = CurrentTestModel()
trainer = Trainer(default_root_dir=tmpdir, max_steps=max_steps, max_epochs=2)
assert 0 == model.on_before_zero_grad_called
trainer.fit(model)
assert max_steps == model.on_before_zero_grad_called
model.on_before_zero_grad_called = 0
trainer.test(model)
assert 0 == model.on_before_zero_grad_called
def test_training_epoch_end_metrics_collection(tmpdir):
"""Test that progress bar metrics also get collected at the end of an epoch."""
num_epochs = 3
class CurrentModel(BoringModel):
def training_step(self, *args, **kwargs):
output = super().training_step(*args, **kwargs)
self.log_dict({"step_metric": torch.tensor(-1), "shared_metric": 100}, logger=False, prog_bar=True)
return output
def training_epoch_end(self, outputs):
epoch = self.current_epoch
# both scalar tensors and Python numbers are accepted
self.log_dict(
{f"epoch_metric_{epoch}": torch.tensor(epoch), "shared_metric": 111}, logger=False, prog_bar=True
)
model = CurrentModel()
trainer = Trainer(max_epochs=num_epochs, default_root_dir=tmpdir, overfit_batches=2)
trainer.fit(model)
assert trainer.state.finished, f"Training failed with {trainer.state}"
metrics = trainer.progress_bar_callback.get_metrics(trainer, model)
# metrics added in training step should be unchanged by epoch end method
assert metrics["step_metric"] == -1
# a metric shared in both methods gets overwritten by epoch_end
assert metrics["shared_metric"] == 111
# metrics are kept after each epoch
for i in range(num_epochs):
assert metrics[f"epoch_metric_{i}"] == i
def test_training_epoch_end_metrics_collection_on_override(tmpdir):
"""Test that batch end metrics are collected when training_epoch_end is overridden at the end of an epoch."""
class OverriddenModel(BoringModel):
def __init__(self):
super().__init__()
self.len_outputs = 0
def on_train_epoch_start(self):
self.num_train_batches = 0
def training_epoch_end(self, outputs):
self.len_outputs = len(outputs)
def on_train_batch_end(self, outputs, batch, batch_idx):
self.num_train_batches += 1
class NotOverriddenModel(BoringModel):
def on_train_epoch_start(self):
self.num_train_batches = 0
def on_train_batch_end(self, outputs, batch, batch_idx):
self.num_train_batches += 1
overridden_model = OverriddenModel()
not_overridden_model = NotOverriddenModel()
not_overridden_model.training_epoch_end = None
trainer = Trainer(max_epochs=1, default_root_dir=tmpdir, overfit_batches=2)
trainer.fit(overridden_model)
assert overridden_model.len_outputs == overridden_model.num_train_batches
@pytest.mark.parametrize(
"accelerator,expected_device_str",
[
pytest.param("gpu", "cuda:0", marks=RunIf(min_cuda_gpus=1)),
pytest.param("mps", "mps:0", marks=RunIf(mps=True)),
],
)
@mock.patch(
"pytorch_lightning.strategies.Strategy.lightning_module",
new_callable=PropertyMock,
)
def test_apply_batch_transfer_handler(model_getter_mock, accelerator, expected_device_str):
expected_device = torch.device(expected_device_str)
class CustomBatch:
def __init__(self, data):
self.samples = data[0]
self.targets = data[1]
class CurrentTestModel(BoringModel):
rank = 0
transfer_batch_to_device_hook_rank = None
on_after_batch_transfer_hook_rank = None
def on_after_batch_transfer(self, batch, dataloader_idx):
assert dataloader_idx == 0
assert batch.samples.device == batch.targets.device == expected_device
self.on_after_batch_transfer_hook_rank = self.rank
self.rank += 1
batch.targets *= 2
return batch
def transfer_batch_to_device(self, batch, device, dataloader_idx):
assert dataloader_idx == 0
self.transfer_batch_to_device_hook_rank = self.rank
self.rank += 1
batch.samples = batch.samples.to(device)
batch.targets = batch.targets.to(device)
return batch
model = CurrentTestModel()
batch = CustomBatch((torch.zeros(5, 32), torch.ones(5, 1, dtype=torch.long)))
trainer = Trainer(accelerator=accelerator, devices=1)
# running .fit() would require us to implement custom data loaders, we mock the model reference instead
model_getter_mock.return_value = model
batch_gpu = trainer.strategy.batch_to_device(batch, expected_device)
assert model.transfer_batch_to_device_hook_rank == 0
assert model.on_after_batch_transfer_hook_rank == 1
assert batch_gpu.samples.device == batch_gpu.targets.device == expected_device
assert torch.allclose(batch_gpu.samples.cpu(), torch.zeros(5, 32))
assert torch.allclose(batch_gpu.targets.cpu(), torch.ones(5, 1, dtype=torch.long) * 2)
@RunIf(min_cuda_gpus=2, standalone=True)
def test_transfer_batch_hook_ddp(tmpdir):
"""Test custom data are properly moved to the right device using ddp."""
class CustomBatch:
def __init__(self, data):
self.samples = data[0]
def to(self, device, **kwargs):
self.samples = self.samples.to(device, **kwargs)
return self
def collate_fn(batch):
return CustomBatch(batch)
class TestModel(BoringModel):
def training_step(self, batch, batch_idx):
assert batch.samples.device == self.device
assert isinstance(batch_idx, int)
def train_dataloader(self):
return torch.utils.data.DataLoader(RandomDataset(32, 64), collate_fn=collate_fn)
model = TestModel()
model.validation_step = None
model.training_epoch_end = None
trainer = Trainer(
default_root_dir=tmpdir,
limit_train_batches=2,
limit_val_batches=0,
max_epochs=1,
strategy="ddp",
accelerator="gpu",
devices=2,
enable_progress_bar=False,
enable_model_summary=False,
)
trainer.fit(model)
def get_members(cls):
return {h for h, _ in getmembers(cls, predicate=isfunction) if not h.startswith("_")}
class HookedCallback(Callback):
def __init__(self, called):
def call(hook, fn, *args, **kwargs):
out = fn(*args, **kwargs)
d = {"name": f"Callback.{hook}"}
if args:
d["args"] = args
if kwargs:
d["kwargs"] = kwargs
called.append(d)
return out
for h in get_members(Callback):
attr = getattr(self, h)
partial_h = partial(call, h, attr)
update_wrapper(partial_h, attr)
setattr(self, h, partial_h)
def state_dict(*args, **kwargs):
return {"foo": True}
class HookedModel(BoringModel):
def __init__(self, called):
super().__init__()
pl_module_hooks = get_members(LightningModule)
# remove non-hooks
pl_module_hooks.difference_update({"optimizers", "log", "log_dict"})
# remove most `nn.Module` hooks
module_hooks = get_members(torch.nn.Module)
module_hooks.difference_update({"forward", "zero_grad", "train"})
pl_module_hooks.difference_update(module_hooks)
def call(hook, fn, *args, **kwargs):
out = fn(*args, **kwargs)
d = {"name": hook}
if args:
d["args"] = args
elif hook == "train":
# DeepSpeed calls `train(mode)` but we do not. Standardize
# https://github.com/microsoft/DeepSpeed/pull/571
d["args"] = (True,)
if kwargs:
d["kwargs"] = kwargs
called.append(d)
return out
for h in pl_module_hooks:
attr = getattr(self, h)
partial_h = partial(call, h, attr)
update_wrapper(partial_h, attr)
setattr(self, h, partial_h)
def validation_epoch_end(self, *args, **kwargs):
# `BoringModel` does not have a return for `validation_step_end` so this would fail
pass
def test_epoch_end(self, *args, **kwargs):
# `BoringModel` does not have a return for `test_step_end` so this would fail
pass
def _train_batch(self, *args, **kwargs):
if self.automatic_optimization:
return self._auto_train_batch(*args, **kwargs)
return self._manual_train_batch(*args, **kwargs)
@staticmethod
def _auto_train_batch(
trainer, model, batches, device=torch.device("cpu"), current_epoch=0, current_batch=0, **kwargs
):
using_native_amp = kwargs.get("amp_backend") == "native"
using_deepspeed = kwargs.get("strategy") == "deepspeed"
out = []
for i in range(current_batch, batches):
out.extend(
[
dict(name="on_before_batch_transfer", args=(ANY, 0)),
dict(name="transfer_batch_to_device", args=(ANY, device, 0)),
dict(name="on_after_batch_transfer", args=(ANY, 0)),
dict(name="Callback.on_train_batch_start", args=(trainer, model, ANY, i)),
dict(name="on_train_batch_start", args=(ANY, i)),
dict(name="forward", args=(ANY,)),
dict(name="training_step", args=(ANY, i)),
dict(name="training_step_end", args=(dict(loss=ANY),)),
dict(name="Callback.on_before_zero_grad", args=(trainer, model, ANY)),
dict(name="on_before_zero_grad", args=(ANY,)),
dict(name="optimizer_zero_grad", args=(current_epoch, i, ANY, 0)),
dict(name="Callback.on_before_backward", args=(trainer, model, ANY)),
dict(name="on_before_backward", args=(ANY,)),
# DeepSpeed handles backward internally
*([dict(name="backward", args=(ANY, ANY, 0))] if not using_deepspeed else []),
dict(name="Callback.on_after_backward", args=(trainer, model)),
dict(name="on_after_backward"),
# note: unscaling happens here in the case of AMP
dict(name="Callback.on_before_optimizer_step", args=(trainer, model, ANY, 0)),
dict(name="on_before_optimizer_step", args=(ANY, 0)),
*([dict(name="log_grad_norm", args=ANY)] if not using_deepspeed else []),
dict(
name="clip_gradients",
args=(ANY,),
kwargs=dict(gradient_clip_val=None, gradient_clip_algorithm=None),
),
dict(
name="configure_gradient_clipping",
args=(ANY, 0),
kwargs=dict(gradient_clip_val=None, gradient_clip_algorithm=None),
),
# this is after because it refers to the `LightningModule.optimizer_step` hook which encapsulates
# the actual call to `PrecisionPlugin.optimizer_step`
dict(
name="optimizer_step",
args=(current_epoch, i, ANY, 0, ANY),
kwargs=dict(on_tpu=False, using_lbfgs=False, using_native_amp=using_native_amp),
),
*(
[dict(name="lr_scheduler_step", args=(ANY, 0, None))]
if i == (trainer.num_training_batches - 1)
else []
),
dict(name="Callback.on_train_batch_end", args=(trainer, model, dict(loss=ANY), ANY, i)),
dict(name="on_train_batch_end", args=(dict(loss=ANY), ANY, i)),
]
)
return out
@staticmethod
def _manual_train_batch(trainer, model, batches, device=torch.device("cpu"), **kwargs):
using_deepspeed = kwargs.get("strategy") == "deepspeed"
out = []
for i in range(batches):
out.extend(
[
dict(name="on_before_batch_transfer", args=(ANY, 0)),
dict(name="transfer_batch_to_device", args=(ANY, device, 0)),
dict(name="on_after_batch_transfer", args=(ANY, 0)),
dict(name="Callback.on_train_batch_start", args=(trainer, model, ANY, i)),
dict(name="on_train_batch_start", args=(ANY, i)),
dict(name="forward", args=(ANY,)),
dict(name="Callback.on_before_backward", args=(trainer, model, ANY)),
dict(name="on_before_backward", args=(ANY,)),
# DeepSpeed handles backward internally
*([dict(name="backward", args=(ANY, None, None))] if not using_deepspeed else []),
dict(name="Callback.on_after_backward", args=(trainer, model)),
dict(name="on_after_backward"),
# `manual_backward` calls the previous 3
dict(name="manual_backward", args=(ANY,)),
dict(name="closure"),
dict(name="Callback.on_before_optimizer_step", args=(trainer, model, ANY, 0)),
dict(name="on_before_optimizer_step", args=(ANY, 0)),
*([dict(name="log_grad_norm", args=ANY)] if not using_deepspeed else []),
dict(name="training_step", args=(ANY, i)),
dict(name="training_step_end", args=(dict(loss=ANY),)),
dict(name="Callback.on_train_batch_end", args=(trainer, model, dict(loss=ANY), ANY, i)),
dict(name="on_train_batch_end", args=(dict(loss=ANY), ANY, i)),
]
)
return out
@staticmethod
def _eval_epoch(fn, trainer, model, batches, key, device=torch.device("cpu")):
outputs = {key: ANY}
return [
dict(name=f"Callback.on_{fn}_epoch_start", args=(trainer, model)),
dict(name=f"on_{fn}_epoch_start"),
*HookedModel._eval_batch(fn, trainer, model, batches, key, device=device),
dict(name=f"{fn}_epoch_end", args=([outputs] * batches,)),
dict(name=f"Callback.on_{fn}_epoch_end", args=(trainer, model)),
dict(name=f"on_{fn}_epoch_end"),
]
@staticmethod
def _eval_batch(fn, trainer, model, batches, key, device=torch.device("cpu")):
out = []
outputs = {key: ANY}
for i in range(batches):
out.extend(
[
dict(name="on_before_batch_transfer", args=(ANY, 0)),
dict(name="transfer_batch_to_device", args=(ANY, device, 0)),
dict(name="on_after_batch_transfer", args=(ANY, 0)),
dict(name=f"Callback.on_{fn}_batch_start", args=(trainer, model, ANY, i, 0)),
dict(name=f"on_{fn}_batch_start", args=(ANY, i, 0)),
dict(name="forward", args=(ANY,)),
dict(name=f"{fn}_step", args=(ANY, i)),
dict(name=f"{fn}_step_end", args=(outputs,)),
dict(name=f"Callback.on_{fn}_batch_end", args=(trainer, model, outputs, ANY, i, 0)),
dict(name=f"on_{fn}_batch_end", args=(outputs, ANY, i, 0)),
]
)
return out
@staticmethod
def _predict_batch(trainer, model, batches):
out = []
for i in range(batches):
out.extend(
[
dict(name="on_before_batch_transfer", args=(ANY, 0)),
dict(name="transfer_batch_to_device", args=(ANY, torch.device("cpu"), 0)),
dict(name="on_after_batch_transfer", args=(ANY, 0)),
dict(name="Callback.on_predict_batch_start", args=(trainer, model, ANY, i, 0)),
dict(name="on_predict_batch_start", args=(ANY, i, 0)),
dict(name="forward", args=(ANY,)),
dict(name="predict_step", args=(ANY, i)),
# TODO: `predict_step_end`
dict(name="Callback.on_predict_batch_end", args=(trainer, model, ANY, ANY, i, 0)),
dict(name="on_predict_batch_end", args=(ANY, ANY, i, 0)),
]
)
return out
@pytest.mark.parametrize(
"kwargs",
[
{},
# these precision plugins modify the optimization flow, so testing them explicitly
pytest.param(
dict(accelerator="gpu", devices=1, precision=16, amp_backend="native"), marks=RunIf(min_cuda_gpus=1)
),
pytest.param(
dict(accelerator="gpu", devices=1, precision=16, amp_backend="apex"),
marks=RunIf(min_cuda_gpus=1, amp_apex=True),
),
pytest.param(
dict(accelerator="gpu", devices=1, precision=16, strategy="deepspeed"),
marks=RunIf(min_cuda_gpus=1, standalone=True, deepspeed=True),
),
],
)
@pytest.mark.parametrize("automatic_optimization", (True, False))
def test_trainer_model_hook_system_fit(tmpdir, kwargs, automatic_optimization):
called = []
class TestModel(HookedModel):
def __init__(self, *args):
super().__init__(*args)
self.automatic_optimization = automatic_optimization
def training_step(self, batch, batch_idx):
if self.automatic_optimization:
return super().training_step(batch, batch_idx)
loss = self.step(batch[0])
opt = self.optimizers()
opt.zero_grad()
self.manual_backward(loss)
opt.step(lambda: called.append({"name": "closure"}))
return {"loss": loss}
model = TestModel(called)
callback = HookedCallback(called)
train_batches = 2
val_batches = 2
trainer = Trainer(
default_root_dir=tmpdir,
max_epochs=1,
limit_train_batches=train_batches,
limit_val_batches=val_batches,
enable_progress_bar=False,
enable_model_summary=False,
callbacks=[callback],
track_grad_norm=1,
**kwargs,
)
trainer.fit(model)
saved_ckpt = {
"callbacks": ANY,
"epoch": 0,
"global_step": train_batches,
"lr_schedulers": ANY,
"optimizer_states": ANY,
"pytorch-lightning_version": __version__,
"state_dict": ANY,
"loops": ANY,
}
if kwargs.get("amp_backend") == "native" or kwargs.get("amp_backend") == "apex":
saved_ckpt[trainer.precision_plugin.__class__.__qualname__] = ANY
device = torch.device("cuda:0" if "accelerator" in kwargs and kwargs["accelerator"] == "gpu" else "cpu")
expected = [
dict(name="configure_callbacks"),
dict(name="prepare_data"),
# DeepSpeed needs the batch size to figure out throughput logging
*([dict(name="train_dataloader")] if kwargs.get("strategy") == "deepspeed" else []),
dict(name="Callback.setup", args=(trainer, model), kwargs=dict(stage="fit")),
dict(name="setup", kwargs=dict(stage="fit")),
dict(name="configure_sharded_model"),
dict(name="configure_optimizers"),
dict(name="Callback.on_fit_start", args=(trainer, model)),
dict(name="on_fit_start"),
dict(name="Callback.on_sanity_check_start", args=(trainer, model)),
dict(name="val_dataloader"),
dict(name="train", args=(False,)),
dict(name="on_validation_model_eval"),
dict(name="zero_grad"),
dict(name="Callback.on_validation_start", args=(trainer, model)),
dict(name="on_validation_start"),
*model._eval_epoch("validation", trainer, model, val_batches, "x", device=device),
dict(name="Callback.on_validation_end", args=(trainer, model)),
dict(name="on_validation_end"),
dict(name="train", args=(True,)),
dict(name="on_validation_model_train"),
dict(name="Callback.on_sanity_check_end", args=(trainer, model)),
# duplicate `train` because `_run_train` calls it again in case validation wasn't run
dict(name="train", args=(True,)),
dict(name="train_dataloader"),
dict(name="Callback.on_train_start", args=(trainer, model)),
dict(name="on_train_start"),
dict(name="Callback.on_train_epoch_start", args=(trainer, model)),
dict(name="on_train_epoch_start"),
*model._train_batch(trainer, model, train_batches, device=device, **kwargs),
dict(name="train", args=(False,)),
dict(name="on_validation_model_eval"),
dict(name="zero_grad"),
dict(name="Callback.on_validation_start", args=(trainer, model)),
dict(name="on_validation_start"),
*model._eval_epoch("validation", trainer, model, val_batches, "x", device=device),
dict(name="Callback.on_validation_end", args=(trainer, model)),
dict(name="on_validation_end"),
dict(name="train", args=(True,)),
dict(name="on_validation_model_train"),
dict(name="training_epoch_end", args=([dict(loss=ANY)] * train_batches,)),
dict(name="Callback.on_train_epoch_end", args=(trainer, model)),
# `ModelCheckpoint.save_checkpoint` is called here from `Callback.on_train_epoch_end`
dict(name="Callback.state_dict"),
dict(name="Callback.on_save_checkpoint", args=(trainer, model, saved_ckpt)),
dict(name="on_save_checkpoint", args=(saved_ckpt,)),
dict(name="on_train_epoch_end"),
dict(name="Callback.on_train_end", args=(trainer, model)),
dict(name="on_train_end"),
dict(name="Callback.on_fit_end", args=(trainer, model)),
dict(name="on_fit_end"),
dict(name="Callback.teardown", args=(trainer, model), kwargs=dict(stage="fit")),
dict(name="teardown", kwargs=dict(stage="fit")),
]
assert called == expected
def test_trainer_model_hook_system_fit_no_val_and_resume_max_epochs(tmpdir):
# initial training to get a checkpoint
model = BoringModel()
trainer = Trainer(
default_root_dir=tmpdir,
max_epochs=1,
limit_train_batches=2,
limit_val_batches=0,
enable_progress_bar=False,
enable_model_summary=False,
callbacks=[HookedCallback([])],
)
trainer.fit(model)
best_model_path = trainer.checkpoint_callback.best_model_path
called = []
callback = HookedCallback(called)
# already performed 1 step, resume and do 2 more
trainer = Trainer(
default_root_dir=tmpdir,
max_epochs=2,
limit_train_batches=2,
limit_val_batches=0,
enable_progress_bar=False,
enable_model_summary=False,
callbacks=[callback],
track_grad_norm=1,
)
# resume from checkpoint with HookedModel
model = HookedModel(called)
trainer.fit(model, ckpt_path=best_model_path)
loaded_ckpt = {
"callbacks": ANY,
"epoch": 0,
"global_step": 2,
"lr_schedulers": ANY,
"optimizer_states": ANY,
"pytorch-lightning_version": __version__,
"state_dict": ANY,
"loops": ANY,
}
saved_ckpt = {**loaded_ckpt, "global_step": 4, "epoch": 1}
expected = [
dict(name="configure_callbacks"),
dict(name="prepare_data"),
dict(name="Callback.setup", args=(trainer, model), kwargs=dict(stage="fit")),
dict(name="setup", kwargs=dict(stage="fit")),
dict(name="on_load_checkpoint", args=(loaded_ckpt,)),
dict(name="Callback.on_load_checkpoint", args=(trainer, model, loaded_ckpt)),
dict(name="Callback.load_state_dict", args=({"foo": True},)),
dict(name="configure_sharded_model"),
dict(name="configure_optimizers"),
dict(name="Callback.on_fit_start", args=(trainer, model)),
dict(name="on_fit_start"),
dict(name="train", args=(True,)),
dict(name="train_dataloader"),
dict(name="Callback.on_train_start", args=(trainer, model)),
dict(name="on_train_start"),
dict(name="Callback.on_train_epoch_start", args=(trainer, model)),
dict(name="on_train_epoch_start"),
*model._train_batch(trainer, model, 2, current_epoch=1, current_batch=0),
dict(name="training_epoch_end", args=([dict(loss=ANY)] * 2,)),
dict(name="Callback.on_train_epoch_end", args=(trainer, model)),
dict(name="Callback.state_dict"),
dict(name="Callback.on_save_checkpoint", args=(trainer, model, saved_ckpt)),
dict(name="on_save_checkpoint", args=(saved_ckpt,)),
dict(name="on_train_epoch_end"),
dict(name="Callback.on_train_end", args=(trainer, model)),
dict(name="on_train_end"),
dict(name="Callback.on_fit_end", args=(trainer, model)),
dict(name="on_fit_end"),
dict(name="Callback.teardown", args=(trainer, model), kwargs=dict(stage="fit")),
dict(name="teardown", kwargs=dict(stage="fit")),
]
assert called == expected
def test_trainer_model_hook_system_fit_no_val_and_resume_max_steps(tmpdir):
# initial training to get a checkpoint
model = BoringModel()
trainer = Trainer(
default_root_dir=tmpdir,
max_steps=1,
limit_val_batches=0,
enable_progress_bar=False,
enable_model_summary=False,
callbacks=[HookedCallback([])],
)
trainer.fit(model)
best_model_path = trainer.checkpoint_callback.best_model_path
# resume from checkpoint with HookedModel
called = []
model = HookedModel(called)
callback = HookedCallback(called)
# already performed 1 step, resume and do 2 more
train_batches = 2
steps_after_reload = 1 + train_batches
trainer = Trainer(
default_root_dir=tmpdir,
max_steps=steps_after_reload,
limit_val_batches=0,
enable_progress_bar=False,
enable_model_summary=False,
callbacks=[callback],
track_grad_norm=1,
)
trainer.fit(model, ckpt_path=best_model_path)
loaded_ckpt = {
"callbacks": ANY,
"epoch": 0,
"global_step": 1,
"lr_schedulers": ANY,
"optimizer_states": ANY,
"pytorch-lightning_version": __version__,
"state_dict": ANY,
"loops": ANY,
}
saved_ckpt = {**loaded_ckpt, "global_step": steps_after_reload}
expected = [
dict(name="configure_callbacks"),
dict(name="prepare_data"),
dict(name="Callback.setup", args=(trainer, model), kwargs=dict(stage="fit")),
dict(name="setup", kwargs=dict(stage="fit")),
dict(name="on_load_checkpoint", args=(loaded_ckpt,)),
dict(name="Callback.on_load_checkpoint", args=(trainer, model, loaded_ckpt)),
dict(name="Callback.load_state_dict", args=({"foo": True},)),
dict(name="configure_sharded_model"),
dict(name="configure_optimizers"),
dict(name="Callback.on_fit_start", args=(trainer, model)),
dict(name="on_fit_start"),
dict(name="train", args=(True,)),
dict(name="train_dataloader"),
dict(name="Callback.on_train_start", args=(trainer, model)),
dict(name="on_train_start"),
dict(name="Callback.on_train_epoch_start", args=(trainer, model)),
dict(name="on_train_epoch_start"),
*model._train_batch(trainer, model, steps_after_reload, current_batch=1),
dict(name="training_epoch_end", args=([dict(loss=ANY)] * train_batches,)),
dict(name="Callback.on_train_epoch_end", args=(trainer, model)),
dict(name="Callback.state_dict"),
dict(name="Callback.on_save_checkpoint", args=(trainer, model, saved_ckpt)),
dict(name="on_save_checkpoint", args=(saved_ckpt,)),
dict(name="on_train_epoch_end"),
dict(name="Callback.on_train_end", args=(trainer, model)),
dict(name="on_train_end"),
dict(name="Callback.on_fit_end", args=(trainer, model)),
dict(name="on_fit_end"),
dict(name="Callback.teardown", args=(trainer, model), kwargs=dict(stage="fit")),
dict(name="teardown", kwargs=dict(stage="fit")),
]
assert called == expected
@pytest.mark.parametrize("batches", (0, 2))
@pytest.mark.parametrize(
["verb", "noun", "dataloader", "key"], [("validate", "validation", "val", "x"), ("test", "test", "test", "y")]
)
def test_trainer_model_hook_system_eval(tmpdir, batches, verb, noun, dataloader, key):
called = []
model = HookedModel(called)
callback = HookedCallback(called)
trainer = Trainer(
default_root_dir=tmpdir,
max_epochs=1,
limit_val_batches=batches,
limit_test_batches=batches,
enable_progress_bar=False,
enable_model_summary=False,
callbacks=[callback],
)
fn = getattr(trainer, verb)
fn(model, verbose=False)
hooks = [
dict(name=f"{dataloader}_dataloader"),
dict(name="train", args=(False,)),
dict(name=f"on_{noun}_model_eval"),
dict(name="zero_grad"),
dict(name=f"Callback.on_{noun}_start", args=(trainer, model)),
dict(name=f"on_{noun}_start"),
*model._eval_epoch(noun, trainer, model, batches, key),
dict(name=f"Callback.on_{noun}_end", args=(trainer, model)),
dict(name=f"on_{noun}_end"),
dict(name="train", args=(True,)),
dict(name=f"on_{noun}_model_train"),
]
expected = [
dict(name="configure_callbacks"),
dict(name="prepare_data"),
dict(name="Callback.setup", args=(trainer, model), kwargs=dict(stage=verb)),
dict(name="setup", kwargs=dict(stage=verb)),
dict(name="configure_sharded_model"),
*(hooks if batches else []),
dict(name="Callback.teardown", args=(trainer, model), kwargs=dict(stage=verb)),
dict(name="teardown", kwargs=dict(stage=verb)),
]
assert called == expected
def test_trainer_model_hook_system_predict(tmpdir):
called = []
model = HookedModel(called)
callback = HookedCallback(called)
batches = 2
trainer = Trainer(
default_root_dir=tmpdir, limit_predict_batches=batches, enable_progress_bar=False, callbacks=[callback]
)
trainer.predict(model)
expected = [
dict(name="configure_callbacks"),
dict(name="prepare_data"),
dict(name="Callback.setup", args=(trainer, model), kwargs=dict(stage="predict")),
dict(name="setup", kwargs=dict(stage="predict")),
dict(name="configure_sharded_model"),
dict(name="predict_dataloader"),
dict(name="train", args=(False,)),
dict(name="on_predict_model_eval"),
dict(name="zero_grad"),
dict(name="Callback.on_predict_start", args=(trainer, model)),
dict(name="on_predict_start"),
dict(name="Callback.on_predict_epoch_start", args=(trainer, model)),
dict(name="on_predict_epoch_start"),
*model._predict_batch(trainer, model, batches),
# TODO: `predict_epoch_end`
dict(name="Callback.on_predict_epoch_end", args=(trainer, model, [[ANY] * batches])),
dict(name="on_predict_epoch_end", args=([[ANY] * batches],)),
dict(name="Callback.on_predict_end", args=(trainer, model)),
dict(name="on_predict_end"),
# TODO: `on_predict_model_train`
dict(name="Callback.teardown", args=(trainer, model), kwargs=dict(stage="predict")),
dict(name="teardown", kwargs=dict(stage="predict")),
]
assert called == expected
# TODO: add test for tune
def test_hooks_with_different_argument_names(tmpdir):
"""Test that argument names can be anything in the hooks."""
class CustomBoringModel(BoringModel):
def assert_args(self, x, batch_nb):
assert isinstance(x, torch.Tensor)
assert x.size() == (1, 32)
assert isinstance(batch_nb, int)
def training_step(self, x1, batch_nb1):
self.assert_args(x1, batch_nb1)
return super().training_step(x1, batch_nb1)
def validation_step(self, x2, batch_nb2):
self.assert_args(x2, batch_nb2)
return super().validation_step(x2, batch_nb2)
def test_step(self, x3, batch_nb3, dl_idx3):
self.assert_args(x3, batch_nb3)
assert isinstance(dl_idx3, int)
return super().test_step(x3, batch_nb3)
def predict(self, x4, batch_nb4, dl_idx4):
self.assert_args(x4, batch_nb4)
assert isinstance(dl_idx4, int)
return super().predict(x4, batch_nb4, dl_idx4)
def test_dataloader(self):
return [DataLoader(RandomDataset(32, 64)), DataLoader(RandomDataset(32, 64))]
def predict_dataloader(self):
return [DataLoader(RandomDataset(32, 64)), DataLoader(RandomDataset(32, 64))]
model = CustomBoringModel()
model.test_epoch_end = None
trainer = Trainer(default_root_dir=tmpdir, fast_dev_run=5)
trainer.fit(model)
assert trainer.state.finished, f"Training failed with {trainer.state}"
trainer.test(model)
preds = trainer.predict(model)
assert len(preds) == 2
assert all(len(x) == 5 for x in preds)
def test_trainer_datamodule_hook_system(tmpdir):
"""Test the LightningDataModule hook system."""
model = BoringModel()
batches = 2
trainer = Trainer(
default_root_dir=tmpdir,
max_epochs=1,
limit_train_batches=batches,
limit_val_batches=batches,
limit_test_batches=batches,
limit_predict_batches=batches,
enable_progress_bar=False,
enable_model_summary=False,
reload_dataloaders_every_n_epochs=1,
)
called = []
dm = HookedDataModule(called)
trainer.fit(model, datamodule=dm)
expected = [
dict(name="prepare_data"),
dict(name="setup", kwargs=dict(stage="fit")),
dict(name="val_dataloader"),
dict(name="train_dataloader"),
dict(name="state_dict"),
dict(name="teardown", kwargs=dict(stage="fit")),
]
assert called == expected
called = []
dm = HookedDataModule(called)
trainer.validate(model, datamodule=dm, verbose=False)
expected = [
dict(name="prepare_data"),
dict(name="setup", kwargs=dict(stage="validate")),
dict(name="val_dataloader"),
dict(name="teardown", kwargs=dict(stage="validate")),
]
assert called == expected
called = []
dm = HookedDataModule(called)
trainer.test(model, datamodule=dm, verbose=False)
expected = [
dict(name="prepare_data"),
dict(name="setup", kwargs=dict(stage="test")),
dict(name="test_dataloader"),
dict(name="teardown", kwargs=dict(stage="test")),
]
assert called == expected
called = []
dm = HookedDataModule(called)
trainer.predict(model, datamodule=dm)
expected = [
dict(name="prepare_data"),
dict(name="setup", kwargs=dict(stage="predict")),
dict(name="predict_dataloader"),
dict(name="teardown", kwargs=dict(stage="predict")),
]
assert called == expected
def test_load_from_checkpoint_hook_calls(tmpdir):
class CustomHookedDataModule(HookedDataModule):
def state_dict(self):
return {"foo": "bar"}
lm_called, ldm_called = [], []
model = HookedModel(lm_called)
datamodule = CustomHookedDataModule(ldm_called)
trainer = Trainer()
trainer.strategy.connect(model)
trainer._data_connector.attach_data(model, datamodule=datamodule)
ckpt_path = str(tmpdir / "file.ckpt")
trainer.save_checkpoint(ckpt_path)
datamodule_state_dict_key = datamodule.__class__.__qualname__
saved_ckpt = {
"callbacks": ANY,
"epoch": 0,
"global_step": 0,
"lr_schedulers": ANY,
"optimizer_states": ANY,
"pytorch-lightning_version": __version__,
"state_dict": ANY,
"loops": ANY,
datamodule_state_dict_key: {"foo": "bar"},
}
assert lm_called == [dict(name="on_save_checkpoint", args=(saved_ckpt,))]
assert ldm_called == [dict(name="state_dict")]
lm_called, ldm_called = [], []
_ = HookedModel.load_from_checkpoint(ckpt_path, called=lm_called)
_ = CustomHookedDataModule.load_from_checkpoint(ckpt_path, called=ldm_called)
assert lm_called == [dict(name="on_load_checkpoint", args=({**saved_ckpt, "hyper_parameters": ANY},))]
assert ldm_called == [dict(name="load_state_dict", args=(saved_ckpt[datamodule_state_dict_key],))]