lightning/pytorch_lightning/profiler/pytorch.py

# 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.
"""Profiler to check if there are any bottlenecks in your code."""
import inspect
import logging
import os
from functools import partial
from pathlib import Path
from typing import Any, Callable, Dict, List, Optional, Set, Type, TYPE_CHECKING, Union

import torch
from torch import nn, Tensor
from torch.autograd.profiler import record_function

from pytorch_lightning.profiler.base import BaseProfiler
from pytorch_lightning.utilities import rank_zero_warn
from pytorch_lightning.utilities.exceptions import MisconfigurationException
from pytorch_lightning.utilities.imports import _KINETO_AVAILABLE

if TYPE_CHECKING:
    from torch.autograd.profiler import EventList
    from torch.utils.hooks import RemovableHandle

    from pytorch_lightning.core.lightning import LightningModule

if _KINETO_AVAILABLE:
    from torch.profiler import ProfilerAction, ProfilerActivity, tensorboard_trace_handler

log = logging.getLogger(__name__)

_PROFILER = Union[torch.autograd.profiler.profile, torch.cuda.profiler.profile, torch.autograd.profiler.emit_nvtx]


class RegisterRecordFunction:
    """
    While profiling autograd operations, this class will add labels for module names around the forward function.

    The Lightning PyTorch Profiler will activate this feature automatically. It can be deactivated as follows:

    Example::
        from pytorch_lightning.profilers import PyTorchProfiler
        profiler = PyTorchProfiler(record_module_names=False)
        Trainer(profiler=profiler)

    It can be used outside of Lightning as follows:

    Example::
        from pytorch_lightning import Trainer, seed_everything
        with RegisterRecordFunction(model):
            out = model(batch)
    """

    def __init__(self, model: nn.Module) -> None:
        self._model = model
        self._records: Dict[str, record_function] = {}
        self._handles: Dict[str, List["RemovableHandle"]] = {}

    def _start_recording_forward(self, _: nn.Module, input: Tensor, record_name: str) -> Tensor:
        record = record_function(record_name)
        record.__enter__()
        self._records[record_name] = record
        return input

    def _stop_recording_forward(self, _: nn.Module, __: Tensor, output: Tensor, record_name: str) -> Tensor:
        self._records[record_name].__exit__(None, None, None)
        return output

    def __enter__(self) -> None:
        for module_name, module in self._model.named_modules():
            if module_name:
                full_name = f"{type(module).__module__}.{type(module).__name__}"
                record_name = f"{full_name}: {module_name}"
                pre_forward_handle = module.register_forward_pre_hook(
                    partial(self._start_recording_forward, record_name=record_name)
                )
                post_forward_handle = module.register_forward_hook(
                    partial(self._stop_recording_forward, record_name=record_name)
                )

                self._handles[module_name] = [pre_forward_handle, post_forward_handle]

    def __exit__(self, type: Any, value: Any, traceback: Any) -> None:
        for handles in self._handles.values():
            for h in handles:
                h.remove()
        self._handles = {}


class ScheduleWrapper:
    """
    This class is used to override the schedule logic from the profiler and perform
    recording for both `training_step`, `validation_step`.
    """

    def __init__(self, schedule: Callable) -> None:
        if not _KINETO_AVAILABLE:
            raise ModuleNotFoundError("You are trying to use `ScheduleWrapper` which require kineto install.")
        self._schedule = schedule
        self.reset()

    def setup(self, start_action_name: str) -> None:
        self._start_action_name = start_action_name

    def pre_step(self, current_action: str) -> None:
        self._current_action = current_action

    def reset(self):
        self._num_optimizer_step_and_closure = 0
        self._num_validation_step = 0
        self._num_test_step = 0
        self._num_predict_step = 0
        self._optimizer_step_and_closure_reached_end = False
        self._validation_step_reached_end = False
        self._test_step_reached_end = False
        self._predict_step_reached_end = False
        # used to stop profiler when `ProfilerAction.RECORD_AND_SAVE` is reached.
        self._current_action: Optional[str] = None
        self._start_action_name: Optional[str] = None

    @property
    def num_step(self) -> int:
        if self._current_action is not None and self._current_action.startswith("optimizer_step_and_closure_"):
            return self._num_optimizer_step_and_closure
        if self._current_action == "validation_step":
            return self._num_validation_step
        if self._current_action == "test_step":
            return self._num_test_step
        if self._current_action == "predict_step":
            return self._num_predict_step
        return 0

    def _step(self) -> None:
        if self._current_action is not None and self._current_action.startswith("optimizer_step_and_closure_"):
            self._num_optimizer_step_and_closure += 1
        elif self._current_action == "validation_step":
            if self._start_action_name == "on_fit_start":
                if self._num_optimizer_step_and_closure > 0:
                    self._num_validation_step += 1
            else:
                self._num_validation_step += 1
        elif self._current_action == "test_step":
            self._num_test_step += 1
        elif self._current_action == "predict_step":
            self._num_predict_step += 1

    @property
    def has_finished(self) -> bool:
        if self._current_action is not None and self._current_action.startswith("optimizer_step_and_closure_"):
            return self._optimizer_step_and_closure_reached_end
        if self._current_action == "validation_step":
            return self._validation_step_reached_end
        if self._current_action == "test_step":
            return self._test_step_reached_end
        if self._current_action == "predict_step":
            return self._predict_step_reached_end
        return False

    def __call__(self, num_step: int) -> "ProfilerAction":
        # ignore the provided input. Keep internal state instead.
        if self.has_finished:
            return ProfilerAction.NONE

        self._step()
        action = self._schedule(self.num_step)
        if action == ProfilerAction.RECORD_AND_SAVE:
            if self._current_action is not None and self._current_action.startswith("optimizer_step_and_closure_"):
                self._optimizer_step_and_closure_reached_end = True
            elif self._current_action == "validation_step":
                self._validation_step_reached_end = True
            elif self._current_action == "test_step":
                self._test_step_reached_end = True
            elif self._current_action == "predict_step":
                self._predict_step_reached_end = True
        return action


class PyTorchProfiler(BaseProfiler):

    RECORD_FUNCTIONS = {
        "training_step_and_backward",
        "training_step",
        "backward",
        "validation_step",
        "test_step",
        "predict_step",
    }
    RECORD_FUNCTION_PREFIX = "optimizer_step_and_closure_"
    STEP_FUNCTIONS = {"validation_step", "test_step", "predict_step"}
    STEP_FUNCTION_PREFIX = "optimizer_step_and_closure_"
    AVAILABLE_SORT_KEYS = {
        "cpu_time",
        "cuda_time",
        "cpu_time_total",
        "cuda_time_total",
        "cpu_memory_usage",
        "cuda_memory_usage",
        "self_cpu_memory_usage",
        "self_cuda_memory_usage",
        "count",
    }
    START_RECORD_FUNCTIONS = {"on_fit_start", "on_validation_start", "on_test_start", "on_predict_start"}

    def __init__(
        self,
        dirpath: Optional[Union[str, Path]] = None,
        filename: Optional[str] = None,
        group_by_input_shapes: bool = False,
        emit_nvtx: bool = False,
        export_to_chrome: bool = True,
        row_limit: int = 20,
        sort_by_key: Optional[str] = None,
        record_functions: Set[str] = None,
        record_module_names: bool = True,
        **profiler_kwargs: Any,
    ) -> None:
        """
        This profiler uses PyTorch's Autograd Profiler and lets you inspect the cost of
        different operators inside your model - both on the CPU and GPU

        Args:
            dirpath: Directory path for the ``filename``. If ``dirpath`` is ``None`` but ``filename`` is present, the
                ``trainer.log_dir`` (from :class:`~pytorch_lightning.loggers.tensorboard.TensorBoardLogger`)
                will be used.

            filename: If present, filename where the profiler results will be saved instead of printing to stdout.
                The ``.txt`` extension will be used automatically.

            group_by_input_shapes: Include operator input shapes and group calls by shape.

            emit_nvtx: Context manager that makes every autograd operation emit an NVTX range
                Run::

                    nvprof --profile-from-start off -o trace_name.prof -- <regular command here>

                To visualize, you can either use::

                    nvvp trace_name.prof
                    torch.autograd.profiler.load_nvprof(path)

            export_to_chrome: Whether to export the sequence of profiled operators for Chrome.
                It will generate a ``.json`` file which can be read by Chrome.

            row_limit: Limit the number of rows in a table, ``-1`` is a special value that
                removes the limit completely.

            sort_by_key: Attribute used to sort entries. By default
                they are printed in the same order as they were registered.
                Valid keys include: ``cpu_time``, ``cuda_time``, ``cpu_time_total``,
                ``cuda_time_total``, ``cpu_memory_usage``, ``cuda_memory_usage``,
                ``self_cpu_memory_usage``, ``self_cuda_memory_usage``, ``count``.

            record_functions: Set of profiled functions which will create a context manager on.
                Any other will be pass through.

            record_module_names: Whether to add module names while recording autograd operation.

            profiler_kwargs: Keyword arguments for the PyTorch profiler. This depends on your PyTorch version

        Raises:
            MisconfigurationException:
                If arg ``sort_by_key`` is not present in ``AVAILABLE_SORT_KEYS``.
                If arg ``schedule`` is not a ``Callable``.
                If arg ``schedule`` does not return a ``torch.profiler.ProfilerAction``.
        """
        super().__init__(dirpath=dirpath, filename=filename)

        self._group_by_input_shapes = group_by_input_shapes and profiler_kwargs.get("record_shapes", False)
        self._emit_nvtx = emit_nvtx
        self._export_to_chrome = export_to_chrome
        self._row_limit = row_limit
        self._sort_by_key = sort_by_key or f"{'cuda' if profiler_kwargs.get('use_cuda', False) else 'cpu'}_time_total"
        self._user_record_functions = record_functions or set()
        self._record_functions_start = self._user_record_functions | self.START_RECORD_FUNCTIONS
        self._record_functions = self._user_record_functions | self.RECORD_FUNCTIONS
        self._record_module_names = record_module_names
        self._profiler_kwargs = profiler_kwargs

        self.profiler: Optional[_PROFILER] = None
        self.function_events: Optional["EventList"] = None
        self._lightning_module: Optional["LightningModule"] = None  # set by ProfilerConnector
        self._register: Optional[RegisterRecordFunction] = None
        self._parent_profiler: Optional[_PROFILER] = None
        self._recording_map: Dict[str, record_function] = {}
        self._start_action_name: Optional[str] = None
        self._schedule: Optional[ScheduleWrapper] = None

        if _KINETO_AVAILABLE:
            self._init_kineto(profiler_kwargs)

        if self._sort_by_key not in self.AVAILABLE_SORT_KEYS:
            raise MisconfigurationException(
                f"Found sort_by_key: {self._sort_by_key}. Should be within {self.AVAILABLE_SORT_KEYS}. "
            )

    def _init_kineto(self, profiler_kwargs: Any) -> None:
        has_schedule = "schedule" in profiler_kwargs
        self._has_on_trace_ready = "on_trace_ready" in profiler_kwargs

        schedule = profiler_kwargs.get("schedule", None)
        if schedule is not None:
            if not isinstance(schedule, Callable):
                raise MisconfigurationException(f"Schedule should be a callable. Found: {schedule}")
            action = schedule(0)
            if not isinstance(action, ProfilerAction):
                raise MisconfigurationException(
                    f"Schedule should return a `torch.profiler.ProfilerAction`. Found: {action}"
                )
        schedule = schedule if has_schedule else self._default_schedule()
        self._schedule = ScheduleWrapper(schedule) if schedule is not None else schedule
        self._profiler_kwargs["schedule"] = self._schedule

        activities = profiler_kwargs.get("activities", None)
        self._profiler_kwargs["activities"] = activities or self._default_activities()
        self._export_to_flame_graph = profiler_kwargs.get("export_to_flame_graph", False)
        self._metric = profiler_kwargs.get("metric", "self_cpu_time_total")
        with_stack = profiler_kwargs.get("with_stack", False) or self._export_to_flame_graph
        self._profiler_kwargs["with_stack"] = with_stack

    @staticmethod
    def _default_schedule() -> Optional[callable]:
        if _KINETO_AVAILABLE:
            # Those schedule defaults allow the profiling overhead to be negligible over training time.
            return torch.profiler.schedule(wait=1, warmup=1, active=3)

    def _default_activities(self) -> List["ProfilerActivity"]:
        activities = []
        if not _KINETO_AVAILABLE:
            return activities
        if self._profiler_kwargs.get("use_cpu", True):
            activities.append(ProfilerActivity.CPU)
        if self._profiler_kwargs.get("use_cuda", torch.cuda.is_available()):
            activities.append(ProfilerActivity.CUDA)
        return activities

    def start(self, action_name: str) -> None:
        if self.profiler is None and action_name in self._record_functions_start:

            # close profiler if it is already opened. might happen if 2 profilers
            # are created and the first one did not call `describe`
            try:
                torch.autograd._disable_profiler()
            except (AttributeError, RuntimeError):
                pass

            if self._schedule is not None:
                self._schedule.setup(action_name)

            self._create_profilers()

            profiler = self.profiler.__enter__()
            if profiler is not None:
                self.profiler = profiler

            if self._parent_profiler is not None:
                self._parent_profiler.__enter__()

            if self._register is not None:
                self._register.__enter__()

        if (
            self.profiler is not None
            and (action_name in self._record_functions or action_name.startswith(self.RECORD_FUNCTION_PREFIX))
            and action_name not in self._recording_map
        ):
            recording = record_function(action_name)
            recording.__enter__()
            self._recording_map[action_name] = recording

    def stop(self, action_name: str) -> None:
        if action_name in self._recording_map:
            self._recording_map[action_name].__exit__(None, None, None)
            del self._recording_map[action_name]

        if not _KINETO_AVAILABLE or self._emit_nvtx:
            return

        if self.profiler is not None and (
            action_name in self.STEP_FUNCTIONS or action_name.startswith(self.STEP_FUNCTION_PREFIX)
        ):
            if self._schedule is not None:
                self._schedule.pre_step(action_name)

            def on_trace_ready(profiler):
                if self.dirpath is not None:
                    if self._export_to_chrome:
                        handler = tensorboard_trace_handler(
                            self.dirpath, self._prepare_filename(action_name=action_name, extension="")
                        )
                        handler(profiler)

                    if self._export_to_flame_graph:
                        path = os.path.join(
                            self.dirpath, self._prepare_filename(action_name=action_name, extension=".stack")
                        )
                        profiler.export_stacks(path, metric=self._metric)
                else:
                    rank_zero_warn("The PyTorchProfiler failed to export trace as `dirpath` is None")

            if not self._has_on_trace_ready:
                self.profiler.on_trace_ready = on_trace_ready

            if self._schedule is not None:
                self.profiler.step_num = self._schedule.num_step
            self.profiler.step()

    def summary(self) -> str:
        if not self._profiler_kwargs.get("enabled", True) or self._emit_nvtx:
            return ""

        self._delete_profilers()

        if not self.function_events:
            return ""

        if self._export_to_chrome and not _KINETO_AVAILABLE:
            filename = f"{self.local_rank}_trace.json"
            path_to_trace = filename if self.dirpath is None else os.path.join(self.dirpath, filename)
            self.function_events.export_chrome_trace(path_to_trace)

        data = self.function_events.key_averages(group_by_input_shapes=self._group_by_input_shapes)
        table = data.table(sort_by=self._sort_by_key, row_limit=self._row_limit)

        recorded_stats = {"records": table}
        return self._stats_to_str(recorded_stats)

    def _create_profilers(self) -> None:
        if self._emit_nvtx:
            self._parent_profiler = self._create_profiler(torch.cuda.profiler.profile)
            self.profiler = self._create_profiler(torch.autograd.profiler.emit_nvtx)
        else:
            self._parent_profiler = None
            self.profiler = self._create_profiler(
                torch.profiler.profile if _KINETO_AVAILABLE else torch.autograd.profiler.profile
            )
        if self._record_module_names and self._lightning_module is not None:
            self._register = RegisterRecordFunction(self._lightning_module)

    def _create_profiler(self, profiler: Type[_PROFILER]) -> _PROFILER:
        init_parameters = inspect.signature(profiler.__init__).parameters
        kwargs = {k: v for k, v in self._profiler_kwargs.items() if k in init_parameters}
        return profiler(**kwargs)

    def _cache_functions_events(self) -> None:
        if self._emit_nvtx:
            return
        self.function_events = self.profiler.events() if _KINETO_AVAILABLE else self.profiler.function_events

    def _delete_profilers(self) -> None:
        if self.profiler is not None:
            self.profiler.__exit__(None, None, None)
            self._cache_functions_events()
            self.profiler = None

        if self._schedule is not None:
            self._schedule.reset()

        if self._parent_profiler is not None:
            self._parent_profiler.__exit__(None, None, None)
            self._parent_profiler = None

        if self._register is not None:
            self._register.__exit__(None, None, None)
            self._register = None

    def teardown(self, stage: Optional[str] = None) -> None:
        self._delete_profilers()

        for k in self._recording_map:
            self.stop(k)
        self._recording_map = {}

        super().teardown(stage=stage)