Skip to content

Testing callbacks

AbstractStopAndGoCallback

Bases: ABC, BaseInterruptedVsContinuousCallback

Abstract base class for stop-and-go callback to compare metadata before pausing and after resuming training.

This base class provides utility methods to help streamline stop and go comparison.

Provided methods
  • init: initializes the callback with the given mode.
  • get_metadata: abstract method that should be overridden to get metadata from the trainer and pl_module.
Default behaviors
  • in stop mode, metadata is gotten and compared on_validation_epoch_end.
  • in go mode, metadata is gotten and saved on_train_epoch_start.

Override these behaviors if necessary.

Source code in bionemo/testing/testing_callbacks.py 
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
class AbstractStopAndGoCallback(ABC, BaseInterruptedVsContinuousCallback):
    """Abstract base class for stop-and-go callback to compare metadata before pausing and after resuming training.

    This base class provides utility methods to help streamline stop and go comparison.

    Provided methods:
        - __init__: initializes the callback with the given mode.
        - get_metadata: abstract method that should be overridden to get metadata from the trainer and pl_module.

    Default behaviors:
        - in stop mode, metadata is gotten and compared on_validation_epoch_end.
        - in go mode, metadata is gotten and saved on_train_epoch_start.

    Override these behaviors if necessary.
    """

    def __init__(self, mode: Mode = Mode.STOP):
        """Initialize StopAndGoCallback.

        Args:
            mode (str, optional): Mode to run in. Must be either Mode.STOP or Mode.RESUME. Defaults to Mode.STOP.

        Notes:
            User must override get_metadata to get metadata from the trainer and pl_module.
        """
        if mode not in [Mode.STOP, Mode.RESUME]:
            raise ValueError(f"mode must be 'stop' or 'go', got {mode}")
        self.mode = mode
        super().__init__()

    @abstractmethod
    def get_metadata(self, trainer: Trainer, pl_module: LightningModule) -> Any:
        """Get metadata from trainer and pl_module."""
        raise NotImplementedError

    def on_train_epoch_start(self, trainer: Trainer, pl_module: LightningModule):  # noqa: D102
        if self.mode == Mode.RESUME:
            self.data = self.get_metadata(trainer, pl_module)

    def on_validation_epoch_end(self, trainer: Trainer, pl_module: LightningModule):  # noqa: D102
        if not trainer.sanity_checking and self.mode == Mode.STOP:
            self.data = self.get_metadata(trainer, pl_module)

__init__(mode=Mode.STOP)

Initialize StopAndGoCallback.

Parameters:

Name Type Description Default
mode str

Mode to run in. Must be either Mode.STOP or Mode.RESUME. Defaults to Mode.STOP.

 STOP
Notes

User must override get_metadata to get metadata from the trainer and pl_module.

Source code in bionemo/testing/testing_callbacks.py 
221
222
223
224
225
226
227
228
229
230
231
232
233
def __init__(self, mode: Mode = Mode.STOP):
    """Initialize StopAndGoCallback.

    Args:
        mode (str, optional): Mode to run in. Must be either Mode.STOP or Mode.RESUME. Defaults to Mode.STOP.

    Notes:
        User must override get_metadata to get metadata from the trainer and pl_module.
    """
    if mode not in [Mode.STOP, Mode.RESUME]:
        raise ValueError(f"mode must be 'stop' or 'go', got {mode}")
    self.mode = mode
    super().__init__()

get_metadata(trainer, pl_module) abstractmethod

Get metadata from trainer and pl_module.

Source code in bionemo/testing/testing_callbacks.py 
235
236
237
238
@abstractmethod
def get_metadata(self, trainer: Trainer, pl_module: LightningModule) -> Any:
    """Get metadata from trainer and pl_module."""
    raise NotImplementedError

BaseInterruptedVsContinuousCallback

Bases: Callback, CallbackMethods, IOMixin

Base class for serializable stop-and-go callback to compare continuous to interrupted training.

This class is used by extending a callback and collecting data into the self.data attribute. This data is then compared between continuous and interrupted training.

See nemo.lightning.megatron_parallel.CallbackMethods for the available callback methods.

Source code in bionemo/testing/testing_callbacks.py 
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
class BaseInterruptedVsContinuousCallback(Callback, CallbackMethods, io.IOMixin):
    """Base class for serializable stop-and-go callback to compare continuous to interrupted training.

    This class is used by extending a callback and collecting data into the `self.data` attribute. This data is then
    compared between continuous and interrupted training.

    See nemo.lightning.megatron_parallel.CallbackMethods for the available callback methods.
    """

    def __init__(self):
        """Initializes the callback."""
        self.data = []

    def __deepcopy__(self, memo):
        """Don't actually attempt to copy this data when this callback is being serialized."""
        ...

__deepcopy__(memo)

Don't actually attempt to copy this data when this callback is being serialized.

Source code in bionemo/testing/testing_callbacks.py 
61
62
63
def __deepcopy__(self, memo):
    """Don't actually attempt to copy this data when this callback is being serialized."""
    ...

__init__()

Initializes the callback.

Source code in bionemo/testing/testing_callbacks.py 
57
58
59
def __init__(self):
    """Initializes the callback."""
    self.data = []

ConsumedSamplesCallback

Bases: BaseInterruptedVsContinuousCallback

Stop-and-go callback to check consumed samples before pausing and after resuming training.

Source code in bionemo/testing/testing_callbacks.py 
86
87
88
89
90
91
92
93
94
95
96
97
class ConsumedSamplesCallback(BaseInterruptedVsContinuousCallback):
    """Stop-and-go callback to check consumed samples before pausing and after resuming training."""

    def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
        """Get consumed samples as metadata."""
        if step.trainer.training:
            data_sampler = step.trainer.datamodule.data_sampler
            consumed_samples = data_sampler.compute_consumed_samples(
                step.trainer.global_step - step.trainer.datamodule.init_global_step
            )
            self.data.append(np.array(consumed_samples))
        return step

on_megatron_step_start(step)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
89
90
91
92
93
94
95
96
97
def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
    """Get consumed samples as metadata."""
    if step.trainer.training:
        data_sampler = step.trainer.datamodule.data_sampler
        consumed_samples = data_sampler.compute_consumed_samples(
            step.trainer.global_step - step.trainer.datamodule.init_global_step
        )
        self.data.append(np.array(consumed_samples))
    return step

GlobalStepStateCallback

Bases: BaseInterruptedVsContinuousCallback

Stop-and-go callback for global_step before pausing and after resuming training.

Source code in bionemo/testing/testing_callbacks.py 
76
77
78
79
80
81
82
83
class GlobalStepStateCallback(BaseInterruptedVsContinuousCallback):
    """Stop-and-go callback for global_step before pausing and after resuming training."""

    def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
        """Get learning rate as metadata."""
        if step.trainer.training:
            self.data.append(np.array(step.trainer.global_step))
        return step

on_megatron_step_start(step)

Get learning rate as metadata.

Source code in bionemo/testing/testing_callbacks.py 
79
80
81
82
83
def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
    """Get learning rate as metadata."""
    if step.trainer.training:
        self.data.append(np.array(step.trainer.global_step))
    return step

LearningRateCallback

Bases: BaseInterruptedVsContinuousCallback

Stop-and-go callback for learning rate before pausing and after resuming training.

Source code in bionemo/testing/testing_callbacks.py 
66
67
68
69
70
71
72
73
class LearningRateCallback(BaseInterruptedVsContinuousCallback):
    """Stop-and-go callback for learning rate before pausing and after resuming training."""

    def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
        """Get learning rate as metadata."""
        if step.trainer.training:
            self.data.append(np.array(step.trainer.optimizers[0].param_groups[0]["lr"]))
        return step

on_megatron_step_start(step)

Get learning rate as metadata.

Source code in bionemo/testing/testing_callbacks.py 
69
70
71
72
73
def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
    """Get learning rate as metadata."""
    if step.trainer.training:
        self.data.append(np.array(step.trainer.optimizers[0].param_groups[0]["lr"]))
    return step

OptimizerStateCallback

Bases: BaseInterruptedVsContinuousCallback

Stop-and-go callback to check optimizer states before pausing and after resuming training.

Source code in bionemo/testing/testing_callbacks.py 
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
class OptimizerStateCallback(BaseInterruptedVsContinuousCallback):
    """Stop-and-go callback to check optimizer states before pausing and after resuming training."""

    def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
        """Get optimizer states as metadata."""
        if step.trainer.training:
            self.data.append(
                recursive_detach(
                    [
                        optimizer.mcore_optimizer.optimizer.state_dict()["state"]
                        for optimizer in step.trainer.optimizers
                    ]
                )
            )
        return step

on_megatron_step_start(step)

Get optimizer states as metadata.

Source code in bionemo/testing/testing_callbacks.py 
191
192
193
194
195
196
197
198
199
200
201
202
def on_megatron_step_start(self, step: MegatronStep) -> MegatronStep:
    """Get optimizer states as metadata."""
    if step.trainer.training:
        self.data.append(
            recursive_detach(
                [
                    optimizer.mcore_optimizer.optimizer.state_dict()["state"]
                    for optimizer in step.trainer.optimizers
                ]
            )
        )
    return step

RaiseAfterMetadataCallback

Bases: Callback

A callback that raises a StopAndGoException after the validation epoch.

Use this callback for pytest based Stop and go tests.

Source code in bionemo/testing/testing_callbacks.py 
36
37
38
39
40
41
42
43
44
45
class RaiseAfterMetadataCallback(Callback):
    """A callback that raises a StopAndGoException after the validation epoch.

    Use this callback for pytest based Stop and go tests.
    """

    def on_validation_epoch_end(self, trainer: Trainer, pl_module: LightningModule):  # noqa: D102
        if trainer.sanity_checking:
            return
        raise StopAndGoException()

TrainInputCallback

Bases: BaseInterruptedVsContinuousCallback

Collect training input samples for comparison.

Source code in bionemo/testing/testing_callbacks.py 
100
101
102
103
104
105
106
107
108
109
110
111
112
class TrainInputCallback(BaseInterruptedVsContinuousCallback):
    """Collect training input samples for comparison."""

    def on_megatron_microbatch_end(
        self,
        step: MegatronStep,
        batch: DataT,
        forward_callback: "MegatronLossReduction",
        output: Any,
    ) -> None:
        """Get consumed samples as metadata."""
        if step.trainer.training:
            self.data.append(recursive_detach(batch))

on_megatron_microbatch_end(step, batch, forward_callback, output)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
103
104
105
106
107
108
109
110
111
112
def on_megatron_microbatch_end(
    self,
    step: MegatronStep,
    batch: DataT,
    forward_callback: "MegatronLossReduction",
    output: Any,
) -> None:
    """Get consumed samples as metadata."""
    if step.trainer.training:
        self.data.append(recursive_detach(batch))

TrainLossCallback

Bases: BaseInterruptedVsContinuousCallback

Collect training loss samples for comparison.

Source code in bionemo/testing/testing_callbacks.py 
160
161
162
163
164
165
166
167
168
169
170
171
class TrainLossCallback(BaseInterruptedVsContinuousCallback):
    """Collect training loss samples for comparison."""

    def on_megatron_step_end(
        self,
        step: MegatronStep,
        microbatch_outputs: List[Any],
        reduced: Optional[Union[torch.Tensor, Dict[str, torch.Tensor]]] = None,
    ) -> None:
        """Get consumed samples as metadata."""
        if step.trainer.training:
            self.data.append(recursive_detach(reduced))

on_megatron_step_end(step, microbatch_outputs, reduced=None)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
163
164
165
166
167
168
169
170
171
def on_megatron_step_end(
    self,
    step: MegatronStep,
    microbatch_outputs: List[Any],
    reduced: Optional[Union[torch.Tensor, Dict[str, torch.Tensor]]] = None,
) -> None:
    """Get consumed samples as metadata."""
    if step.trainer.training:
        self.data.append(recursive_detach(reduced))

TrainOutputCallback

Bases: BaseInterruptedVsContinuousCallback

Collect training output samples for comparison.

Source code in bionemo/testing/testing_callbacks.py 
130
131
132
133
134
135
136
137
138
139
140
141
142
class TrainOutputCallback(BaseInterruptedVsContinuousCallback):
    """Collect training output samples for comparison."""

    def on_megatron_microbatch_end(
        self,
        step: MegatronStep,
        batch: DataT,
        forward_callback: "MegatronLossReduction",
        output: Any,
    ) -> None:
        """Get consumed samples as metadata."""
        if step.trainer.training:
            self.data.append(recursive_detach(output))

on_megatron_microbatch_end(step, batch, forward_callback, output)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
133
134
135
136
137
138
139
140
141
142
def on_megatron_microbatch_end(
    self,
    step: MegatronStep,
    batch: DataT,
    forward_callback: "MegatronLossReduction",
    output: Any,
) -> None:
    """Get consumed samples as metadata."""
    if step.trainer.training:
        self.data.append(recursive_detach(output))

TrainValInitConsumedSamplesStopAndGoCallback

Bases: AbstractStopAndGoCallback

Stop-and-go callback to check consumed samples before pausing and after resuming training.

This is currently the only callback that doesn't fit with the new pattern of directly comparing continuous and interrupted training, since the dataloaders don't track their consumed_samples before and after checkpoint resumption.

Source code in bionemo/testing/testing_callbacks.py 
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
class TrainValInitConsumedSamplesStopAndGoCallback(AbstractStopAndGoCallback):
    """Stop-and-go callback to check consumed samples before pausing and after resuming training.

    This is currently the only callback that doesn't fit with the new pattern of directly comparing continuous and
    interrupted training, since the dataloaders don't track their consumed_samples before and after checkpoint
    resumption.
    """

    @override
    def get_metadata(self, trainer: Trainer, pl_module: LightningModule) -> Any:
        """Get consumed samples as metadata."""
        # return trainer.datamodule.state_dict()["consumed_samples"]  # TODO why state_dict can be empty despite working lines below
        train_data_sampler: MegatronPretrainingSampler = trainer.train_dataloader.batch_sampler
        val_data_sampler: MegatronPretrainingSampler = trainer.val_dataloaders.batch_sampler
        return train_data_sampler.consumed_samples, val_data_sampler.consumed_samples

get_metadata(trainer, pl_module)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
257
258
259
260
261
262
263
@override
def get_metadata(self, trainer: Trainer, pl_module: LightningModule) -> Any:
    """Get consumed samples as metadata."""
    # return trainer.datamodule.state_dict()["consumed_samples"]  # TODO why state_dict can be empty despite working lines below
    train_data_sampler: MegatronPretrainingSampler = trainer.train_dataloader.batch_sampler
    val_data_sampler: MegatronPretrainingSampler = trainer.val_dataloaders.batch_sampler
    return train_data_sampler.consumed_samples, val_data_sampler.consumed_samples

ValidInputCallback

Bases: BaseInterruptedVsContinuousCallback

Collect validation input samples for comparison.

Source code in bionemo/testing/testing_callbacks.py 
115
116
117
118
119
120
121
122
123
124
125
126
127
class ValidInputCallback(BaseInterruptedVsContinuousCallback):
    """Collect validation input samples for comparison."""

    def on_megatron_microbatch_end(
        self,
        step: MegatronStep,
        batch: DataT,
        forward_callback: "MegatronLossReduction",
        output: Any,
    ) -> None:
        """Get consumed samples as metadata."""
        if step.trainer.validating:
            self.data.append(recursive_detach(batch))

on_megatron_microbatch_end(step, batch, forward_callback, output)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
118
119
120
121
122
123
124
125
126
127
def on_megatron_microbatch_end(
    self,
    step: MegatronStep,
    batch: DataT,
    forward_callback: "MegatronLossReduction",
    output: Any,
) -> None:
    """Get consumed samples as metadata."""
    if step.trainer.validating:
        self.data.append(recursive_detach(batch))

ValidLossCallback

Bases: BaseInterruptedVsContinuousCallback

Collect training loss samples for comparison.

Source code in bionemo/testing/testing_callbacks.py 
174
175
176
177
178
179
180
181
182
183
184
185
class ValidLossCallback(BaseInterruptedVsContinuousCallback):
    """Collect training loss samples for comparison."""

    def on_megatron_step_end(
        self,
        step: MegatronStep,
        microbatch_outputs: List[Any],
        reduced: Optional[Union[torch.Tensor, Dict[str, torch.Tensor]]] = None,
    ) -> None:
        """Get consumed samples as metadata."""
        if step.trainer.validating:
            self.data.append(recursive_detach(reduced))

on_megatron_step_end(step, microbatch_outputs, reduced=None)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
177
178
179
180
181
182
183
184
185
def on_megatron_step_end(
    self,
    step: MegatronStep,
    microbatch_outputs: List[Any],
    reduced: Optional[Union[torch.Tensor, Dict[str, torch.Tensor]]] = None,
) -> None:
    """Get consumed samples as metadata."""
    if step.trainer.validating:
        self.data.append(recursive_detach(reduced))

ValidOutputCallback

Bases: BaseInterruptedVsContinuousCallback

Collect validation output samples for comparison.

Source code in bionemo/testing/testing_callbacks.py 
145
146
147
148
149
150
151
152
153
154
155
156
157
class ValidOutputCallback(BaseInterruptedVsContinuousCallback):
    """Collect validation output samples for comparison."""

    def on_megatron_microbatch_end(
        self,
        step: MegatronStep,
        batch: DataT,
        forward_callback: "MegatronLossReduction",
        output: Any,
    ) -> None:
        """Get consumed samples as metadata."""
        if step.trainer.validating:
            self.data.append(recursive_detach(output))

on_megatron_microbatch_end(step, batch, forward_callback, output)

Get consumed samples as metadata.

Source code in bionemo/testing/testing_callbacks.py 
148
149
150
151
152
153
154
155
156
157
def on_megatron_microbatch_end(
    self,
    step: MegatronStep,
    batch: DataT,
    forward_callback: "MegatronLossReduction",
    output: Any,
) -> None:
    """Get consumed samples as metadata."""
    if step.trainer.validating:
        self.data.append(recursive_detach(output))