scripts/trainer.py

# Copyright (c) MONAI Consortium
# 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 __future__ import annotations

from typing import TYPE_CHECKING, Any, Callable, Dict, Iterable, List, Optional, Sequence, Tuple, Union

import torch
from monai.config import IgniteInfo
from monai.engines.trainer import Trainer
from monai.engines.utils import IterationEvents, default_metric_cmp_fn
from monai.inferers import Inferer
from monai.transforms import Transform
from monai.utils import min_version, optional_import
from monai.utils.enums import CommonKeys as Keys
from torch.optim.optimizer import Optimizer
from torch.utils.data import DataLoader

if TYPE_CHECKING:
    from ignite.engine import Engine, EventEnum
    from ignite.metrics import Metric
else:
    Engine, _ = optional_import("ignite.engine", IgniteInfo.OPT_IMPORT_VERSION, min_version, "Engine")
    Metric, _ = optional_import("ignite.metrics", IgniteInfo.OPT_IMPORT_VERSION, min_version, "Metric")
    EventEnum, _ = optional_import("ignite.engine", IgniteInfo.OPT_IMPORT_VERSION, min_version, "EventEnum")

__all__ = ["DetectionTrainer"]


def detection_prepare_batch(
    batchdata: List[Dict[str, torch.Tensor]],
    device: Optional[Union[str, torch.device]] = None,
    non_blocking: bool = False,
    **kwargs,
) -> Union[Tuple[torch.Tensor, Optional[torch.Tensor]], torch.Tensor]:
    """
    Default function to prepare the data for current iteration.
    Args `batchdata`, `device`, `non_blocking` refer to the ignite API:
    https://pytorch.org/ignite/v0.4.8/generated/ignite.engine.create_supervised_trainer.html.
    `kwargs` supports other args for `Tensor.to()` API.
    Returns:
        image, label(optional).
    """
    inputs = [
        batch_data_ii["image"].to(device=device, non_blocking=non_blocking, **kwargs)
        for batch_data_i in batchdata
        for batch_data_ii in batch_data_i
    ]

    if isinstance(batchdata[0][0].get(Keys.LABEL), torch.Tensor):
        targets = [
            dict(
                label=batch_data_ii["label"].to(device=device, non_blocking=non_blocking, **kwargs),
                box=batch_data_ii["box"].to(device=device, non_blocking=non_blocking, **kwargs),
            )
            for batch_data_i in batchdata
            for batch_data_ii in batch_data_i
        ]
        return (inputs, targets)
    return inputs, None


class DetectionTrainer(Trainer):
    """
    Supervised detection training method with image and label, inherits from ``Trainer`` and ``Workflow``.
    Args:
        device: an object representing the device on which to run.
        max_epochs: the total epoch number for trainer to run.
        train_data_loader: Ignite engine use data_loader to run, must be Iterable or torch.DataLoader.
        detector: detector to train in the trainer, should be regular PyTorch `torch.nn.Module`.
        optimizer: the optimizer associated to the detector, should be regular PyTorch optimizer from `torch.optim`
            or its subclass.
        epoch_length: number of iterations for one epoch, default to `len(train_data_loader)`.
        non_blocking: if True and this copy is between CPU and GPU, the copy may occur asynchronously
            with respect to the host. For other cases, this argument has no effect.
        prepare_batch: function to parse expected data (usually `image`,`box`, `label` and other detector args)
            from `engine.state.batch` for every iteration, for more details please refer to:
            https://pytorch.org/ignite/generated/ignite.engine.create_supervised_trainer.html.
        iteration_update: the callable function for every iteration, expect to accept `engine`
            and `engine.state.batch` as inputs, return data will be stored in `engine.state.output`.
            if not provided, use `self._iteration()` instead. for more details please refer to:
            https://pytorch.org/ignite/generated/ignite.engine.engine.Engine.html.
        inferer: inference method that execute model forward on input data, like: SlidingWindow, etc.
        postprocessing: execute additional transformation for the model output data.
            Typically, several Tensor based transforms composed by `Compose`.
        key_train_metric: compute metric when every iteration completed, and save average value to
            engine.state.metrics when epoch completed. key_train_metric is the main metric to compare and save the
            checkpoint into files.
        additional_metrics: more Ignite metrics that also attach to Ignite Engine.
        metric_cmp_fn: function to compare current key metric with previous best key metric value,
            it must accept 2 args (current_metric, previous_best) and return a bool result: if `True`, will update
            `best_metric` and `best_metric_epoch` with current metric and epoch, default to `greater than`.
        train_handlers: every handler is a set of Ignite Event-Handlers, must have `attach` function, like:
            CheckpointHandler, StatsHandler, etc.
        amp: whether to enable auto-mixed-precision training, default is False.
        event_names: additional custom ignite events that will register to the engine.
            new events can be a list of str or `ignite.engine.events.EventEnum`.
        event_to_attr: a dictionary to map an event to a state attribute, then add to `engine.state`.
            for more details, check: https://pytorch.org/ignite/generated/ignite.engine.engine.Engine.html
            #ignite.engine.engine.Engine.register_events.
        decollate: whether to decollate the batch-first data to a list of data after model computation,
            recommend `decollate=True` when `postprocessing` uses components from `monai.transforms`.
            default to `True`.
        optim_set_to_none: when calling `optimizer.zero_grad()`, instead of setting to zero, set the grads to None.
            more details: https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html.
        to_kwargs: dict of other args for `prepare_batch` API when converting the input data, except for
            `device`, `non_blocking`.
        amp_kwargs: dict of the args for `torch.cuda.amp.autocast()` API, for more details:
            https://pytorch.org/docs/stable/amp.html#torch.cuda.amp.autocast.
    """

    def __init__(
        self,
        device: torch.device,
        max_epochs: int,
        train_data_loader: Iterable | DataLoader,
        detector: torch.nn.Module,
        optimizer: Optimizer,
        epoch_length: int | None = None,
        non_blocking: bool = False,
        prepare_batch: Callable = detection_prepare_batch,
        iteration_update: Callable[[Engine, Any], Any] | None = None,
        inferer: Inferer | None = None,
        postprocessing: Transform | None = None,
        key_train_metric: dict[str, Metric] | None = None,
        additional_metrics: dict[str, Metric] | None = None,
        metric_cmp_fn: Callable = default_metric_cmp_fn,
        train_handlers: Sequence | None = None,
        amp: bool = False,
        event_names: list[str | EventEnum] | None = None,
        event_to_attr: dict | None = None,
        decollate: bool = True,
        optim_set_to_none: bool = False,
        to_kwargs: dict | None = None,
        amp_kwargs: dict | None = None,
    ) -> None:
        super().__init__(
            device=device,
            max_epochs=max_epochs,
            data_loader=train_data_loader,
            epoch_length=epoch_length,
            non_blocking=non_blocking,
            prepare_batch=prepare_batch,
            iteration_update=iteration_update,
            postprocessing=postprocessing,
            key_metric=key_train_metric,
            additional_metrics=additional_metrics,
            metric_cmp_fn=metric_cmp_fn,
            handlers=train_handlers,
            amp=amp,
            event_names=event_names,
            event_to_attr=event_to_attr,
            decollate=decollate,
            to_kwargs=to_kwargs,
            amp_kwargs=amp_kwargs,
        )

        self.detector = detector
        self.optimizer = optimizer
        self.optim_set_to_none = optim_set_to_none

    def _iteration(self, engine, batchdata: dict[str, torch.Tensor]):
        """
        Callback function for the Supervised Training processing logic of 1 iteration in Ignite Engine.
        Return below items in a dictionary:
            - IMAGE: image Tensor data for model input, already moved to device.
            - BOX: box regression loss corresponding to the image, already moved to device.
            - LABEL: classification loss corresponding to the image, already moved to device.
            - LOSS: weighted sum of loss values computed by loss function.
        Args:
            engine: `DetectionTrainer` to execute operation for an iteration.
            batchdata: input data for this iteration, usually can be dictionary or tuple of Tensor data.
        Raises:
            ValueError: When ``batchdata`` is None.
        """

        if batchdata is None:
            raise ValueError("Must provide batch data for current iteration.")

        batch = engine.prepare_batch(batchdata, engine.state.device, engine.non_blocking, **engine.to_kwargs)
        if len(batch) == 2:
            inputs, targets = batch
            args: tuple = ()
            kwargs: dict = {}
        else:
            inputs, targets, args, kwargs = batch
        # put iteration outputs into engine.state
        engine.state.output = {Keys.IMAGE: inputs, Keys.LABEL: targets}

        def _compute_pred_loss(w_cls: float = 1.0, w_box_reg: float = 1.0):
            """
            Args:
                w_cls: weight of classification loss
                w_box_reg: weight of box regression loss
            """
            outputs = engine.detector(inputs, targets)
            engine.state.output[engine.detector.cls_key] = outputs[engine.detector.cls_key]
            engine.state.output[engine.detector.box_reg_key] = outputs[engine.detector.box_reg_key]
            engine.state.output[Keys.LOSS] = (
                w_cls * outputs[engine.detector.cls_key] + w_box_reg * outputs[engine.detector.box_reg_key]
            )
            engine.fire_event(IterationEvents.LOSS_COMPLETED)

        engine.detector.train()
        engine.optimizer.zero_grad(set_to_none=engine.optim_set_to_none)

        if engine.amp and engine.scaler is not None:
            with torch.cuda.amp.autocast(**engine.amp_kwargs):
                inputs = [img.to(torch.float16) for img in inputs]
                _compute_pred_loss()
            engine.scaler.scale(engine.state.output[Keys.LOSS]).backward()
            engine.fire_event(IterationEvents.BACKWARD_COMPLETED)
            engine.scaler.step(engine.optimizer)
            engine.scaler.update()
        else:
            _compute_pred_loss()
            engine.state.output[Keys.LOSS].backward()
            engine.fire_event(IterationEvents.BACKWARD_COMPLETED)
            engine.optimizer.step()

        return engine.state.output