# -*- coding: utf-8 -*- # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved import contextlib import logging import numpy as np import time import weakref import torch import detectron2.utils.comm as comm from detectron2.utils.events import EventStorage __all__ = ["HookBase", "TrainerBase", "SimpleTrainer"] try: _nullcontext = contextlib.nullcontext # python 3.7+ except AttributeError: @contextlib.contextmanager def _nullcontext(enter_result=None): yield enter_result class HookBase: """ Base class for hooks that can be registered with :class:`TrainerBase`. Each hook can implement 4 methods. The way they are called is demonstrated in the following snippet: :: hook.before_train() for iter in range(start_iter, max_iter): hook.before_step() trainer.run_step() hook.after_step() iter += 1 hook.after_train() Notes: 1. In the hook method, users can access ``self.trainer`` to access more properties about the context (e.g., model, current iteration, or config if using :class:`DefaultTrainer`). 2. A hook that does something in :meth:`before_step` can often be implemented equivalently in :meth:`after_step`. If the hook takes non-trivial time, it is strongly recommended to implement the hook in :meth:`after_step` instead of :meth:`before_step`. The convention is that :meth:`before_step` should only take negligible time. Following this convention will allow hooks that do care about the difference between :meth:`before_step` and :meth:`after_step` (e.g., timer) to function properly. Attributes: trainer (TrainerBase): A weak reference to the trainer object. Set by the trainer when the hook is registered. """ def before_train(self): """ Called before the first iteration. """ pass def after_train(self): """ Called after the last iteration. """ pass def before_step(self): """ Called before each iteration. """ pass def after_step(self): """ Called after each iteration. """ pass class TrainerBase: """ Base class for iterative trainer with hooks. The only assumption we made here is: the training runs in a loop. A subclass can implement what the loop is. We made no assumptions about the existence of dataloader, optimizer, model, etc. Attributes: iter(int): the current iteration. start_iter(int): The iteration to start with. By convention the minimum possible value is 0. max_iter(int): The iteration to end training. storage(EventStorage): An EventStorage that's opened during the course of training. """ def __init__(self): self._hooks = [] def register_hooks(self, hooks): """ Register hooks to the trainer. The hooks are executed in the order they are registered. Args: hooks (list[Optional[HookBase]]): list of hooks """ hooks = [h for h in hooks if h is not None] for h in hooks: assert isinstance(h, HookBase) # To avoid circular reference, hooks and trainer cannot own each other. # This normally does not matter, but will cause memory leak if the # involved objects contain __del__: # See http://engineering.hearsaysocial.com/2013/06/16/circular-references-in-python/ h.trainer = weakref.proxy(self) self._hooks.extend(hooks) def train(self, start_iter: int, max_iter: int): """ Args: start_iter, max_iter (int): See docs above """ logger = logging.getLogger(__name__) logger.info("Starting training from iteration {}".format(start_iter)) self.iter = self.start_iter = start_iter self.max_iter = max_iter with EventStorage(start_iter) as self.storage: try: self.before_train() for self.iter in range(start_iter, max_iter): self.before_step() self.run_step() self.after_step() # self.iter == max_iter can be used by `after_train` to # tell whether the training successfully finished or failed # due to exceptions. self.iter += 1 except Exception: logger.exception("Exception during training:") raise finally: self.after_train() def before_train(self): for h in self._hooks: h.before_train() def after_train(self): self.storage.iter = self.iter for h in self._hooks: h.after_train() def before_step(self): # Maintain the invariant that storage.iter == trainer.iter # for the entire execution of each step self.storage.iter = self.iter for h in self._hooks: h.before_step() def after_step(self): for h in self._hooks: h.after_step() def run_step(self): raise NotImplementedError class SimpleTrainer(TrainerBase): """ A simple trainer for the most common type of task: single-cost single-optimizer single-data-source iterative optimization. It assumes that every step, you: 1. Compute the loss with a data from the data_loader. 2. Compute the gradients with the above loss. 3. Update the model with the optimizer. All other tasks during training (checkpointing, logging, evaluation, LR schedule) are maintained by hooks, which can be registered by :meth:`TrainerBase.register_hooks`. If you want to do anything fancier than this, either subclass TrainerBase and implement your own `run_step`, or write your own training loop. """ def __init__(self, model, data_loader, optimizer): """ Args: model: a torch Module. Takes a data from data_loader and returns a dict of losses. data_loader: an iterable. Contains data to be used to call model. optimizer: a torch optimizer. """ super().__init__() """ We set the model to training mode in the trainer. However it's valid to train a model that's in eval mode. If you want your model (or a submodule of it) to behave like evaluation during training, you can overwrite its train() method. """ model.train() self.model = model self.data_loader = data_loader self._data_loader_iter = iter(data_loader) self.optimizer = optimizer def run_step(self): """ Implement the standard training logic described above. """ assert self.model.training, "[SimpleTrainer] model was changed to eval mode!" start = time.perf_counter() """ If you want to do something with the data, you can wrap the dataloader. """ data = next(self._data_loader_iter) data_time = time.perf_counter() - start """ If you want to do something with the losses, you can wrap the model. """ loss_dict = self.model(data) losses = sum(loss_dict.values()) """ If you need to accumulate gradients or do something similar, you can wrap the optimizer with your custom `zero_grad()` method. """ self.optimizer.zero_grad() losses.backward() # use a new stream so the ops don't wait for DDP with torch.cuda.stream( torch.cuda.Stream() ) if losses.device.type == "cuda" else _nullcontext(): metrics_dict = loss_dict metrics_dict["data_time"] = data_time self._write_metrics(metrics_dict) self._detect_anomaly(losses, loss_dict) """ If you need gradient clipping/scaling or other processing, you can wrap the optimizer with your custom `step()` method. But it is suboptimal as explained in https://arxiv.org/abs/2006.15704 Sec 3.2.4 """ self.optimizer.step() def _detect_anomaly(self, losses, loss_dict): if not torch.isfinite(losses).all(): raise FloatingPointError( "Loss became infinite or NaN at iteration={}!\nloss_dict = {}".format( self.iter, loss_dict ) ) def _write_metrics(self, metrics_dict: dict): """ Args: metrics_dict (dict): dict of scalar metrics """ metrics_dict = { k: v.detach().cpu().item() if isinstance(v, torch.Tensor) else float(v) for k, v in metrics_dict.items() } # gather metrics among all workers for logging # This assumes we do DDP-style training, which is currently the only # supported method in detectron2. all_metrics_dict = comm.gather(metrics_dict) if comm.is_main_process(): if "data_time" in all_metrics_dict[0]: # data_time among workers can have high variance. The actual latency # caused by data_time is the maximum among workers. data_time = np.max([x.pop("data_time") for x in all_metrics_dict]) self.storage.put_scalar("data_time", data_time) # average the rest metrics metrics_dict = { k: np.mean([x[k] for x in all_metrics_dict]) for k in all_metrics_dict[0].keys() } total_losses_reduced = sum(loss for loss in metrics_dict.values()) self.storage.put_scalar("total_loss", total_losses_reduced) if len(metrics_dict) > 1: self.storage.put_scalars(**metrics_dict)