* [Feature] Add Runner and Loop

* refine docstring

* refine docstring

* refine __init__ of Runner

* add Runner.save_checkpoint

* add comment for data_batch

* rename validation_cfg to val_cfg

* update branch

* update branch

* discard wear reference

* set seed for torch in dataloader worker

* refine comments

* reorder methods

* import module in __init__.py

* add runner.load_checkpoint

* rename attributes

* fix indent

* fix indent

* add comment for LOOPS

* add example for methods

* add example for methods

* refine examples

* add Runner.resume and refactor register_hooks

* remove type hint of runner in CheckpointHook

* rename before_run to before_train

* remove collate

* add before_run after_run for loop.run

* dummy collate_fn

* add more unit tests

* move call_hook('before_run') to runner

* minor fix

* add comments for collate_fn

* fix type hint

* refactor logic to load or resume checkpoint

* fix typo

* add comments for IterBasedTrainLoop

* refactor unit tests

* fix unit tests

* refactor

* fix setup_env

* refine

* minor fix

* Merge branch 'main' of github.com:open-mmlab/mmengine into zzd/add-runner

* minor fix

* minor fix

* refine

* refactor unit tests

* fix unit tests

* add launcher property

* add logger, message_hub and writer unit tests

* add cur_dataloader attribute for runner

* fix IterBasedTrainLoop

* add torch.no_grad decorator

* add master_only decorator for save_checkpoint

* add before_test_epoch in TestLoop

* refactor load_or_resume method

* refine docstring

* fix typo

* build_dataloader can handle dataset object

* add dump_config for runner

* collate_fn should not a lambda function

* minor fix

* refine docstring

* fix unit tests

* improve comments

* refine unit tests

* rename collate_fn to pseduo_collate

* add 'TODO' tag

* [Refactor] Refactor interface of checkpointhook

* fix print format

* minor ifx

* Messagehub support update dict values

* Messagehub support update dict values

* fix assertion error message

* [Refactor] Use mmengine distributed in evaluator

* remove 'TODO' comment

* use master_only to decorator _log_train and _log_val

* fix resoloved TODO

fix resoloved TODO

fix resoloved TODO

* fix raise error typo

* ensure log item is python scalar

* Fix hook

* Fix

* Fix docs

* FIx

* Fix

* Fix as comment

* update

* Fix hook

* Fix hook

* Fix hook

* Fix itertimerhook

* Fix iter_timer_hook

* Fix

* Fix

* fix logger hook

* Fix loggerhook

* update cur_dataloader

* Fix docstring

* Fix docstring

* Fix as commet

* Fix as commet

* Fix as comment

* rename is_last_epoch, enhance and add after_val before_val .etc

* fix typo in docstring

* remove resolved TODO

* refactor docstring

* [Refactor]: Change scheduler to param_scheduler

* [Fix]: Fix UT of param scheduler hook

Co-authored-by: Your <you@example.com>

* update evaluator prefix

* update docstring and comments

* update doc

* [WIP] testvisualizationhook

* add TestNaiveVisualizationHook

* fix comment

* unpad

* batch imdenormalize

* fix comment

* fix comment

* Config support deep copy

* Fix end of line

* Fix hook

* Fix

* Fix docs

* FIx

* Fix

* Fix as comment

* add required lib for vis

* fix comment

* refine visualization

* refine visualization

* epdate

* update

* update

* update

* fix comment

* add logger hook

* update

* update

* update test

* update

* update test

* update

* update

* update

* update

* update

* Add logger hook

* Fix pre-commit

* Fix as comment

* Fix as comment

* Fix as comment

* Fix as comment

* Fix as comment

* Fix bytes

* update

* Fix as comment

* Fix as comment

* Update runner

* Fix as comment

* Fix as comment

* Fix as comment

* Fix as comment

* Support default_scope when building optimizer and evaluator.

* add docstring

* fix

* fix

* update type hint and ut for evaluator

* update doc

* fix

* support metric prefix in BaseEvaluator

* update docs

* use class attribute default_prefix

* add writer

* update

* update

* update docstr

* update unittest

* update unittest

* fix comment

* update unittest

* fix comment

* fix comment

* fix comment

* fix comment

* update

* [WIP] Visualizer

* add draw featmap

* update docstring

* update docstring

* update docstring

* fix comment

* fix comment

* fix comment

* fix comment

* fix comment

* [Fix]: Change after inter and epoch to after train iter and epoch

* [Fix]: Add new UT to param scheduler hook

Co-authored-by: Your <you@example.com>

* [Fix]: Change after inter and epoch to after train iter and epoch

* [Fix]: Add new UT to param scheduler hook

* [Fix]: Change the type of runner in docstring to Runner

Co-authored-by: Your <you@example.com>

* [Fix]: Fix data batch type in base hook

* [Fix]: Fix the type hint bug in checkpoint, optimizer, param scheduler hooks

Co-authored-by: Your <you@example.com>

* [Feature] Add checkpoint helper

* minor fix

* add comments and fix format

* add hub directory

* add missing docstring

* first commit

* update logging

* update MessageHub unit test and LogBuffer unitest

* update logging docs

* update logging impl and test

* update test logging

* update test

* update logging test and impl

* Fix logging test

* Update log_buffer.py

* rename statistics argumentes

* fix as comment

* Fix as comment

* Fix as comment

* Fix as comment

* Fix meta class

* Fix as comment

* Fix as comment

* Fix as comment

* Fix name declare

* Fix as comment

* Fix as comment

* Fix as comment

* Fix docstring

* Fix as comment

* Fix as comment

* [Feature] Add distributed module

* fix IS_DIST error

* all_reduce_dict does operations in-place

* support 'mean' operation

* provide local group process

* add tmpdir argument for collect_results

* add unit tests

* refactor unit tests

* simplify steps to create multiple processes

* minor fix

* describe the different of *gather* in mmengine and pytorch

* minor fix

* add unit tests for nccl

* test nccl backend in multiple gpu

* add get_default_group function to handle different torch versions

* minor fix

* [Feature] Add distributed module

* fix IS_DIST error

* all_reduce_dict does operations in-place

* support 'mean' operation

* provide local group process

* add tmpdir argument for collect_results

* add unit tests

* refactor unit tests

* simplify steps to create multiple processes

* minor fix

* describe the different of *gather* in mmengine and pytorch

* minor fix

* add unit tests for nccl

* test nccl backend in multiple gpu

* add get_default_group function to handle different torch versions

* minor fix

* minor fix

* handle torch1.5

* handle torch1.5

* minor fix

* fix typo

* refactor unit tests

* nccl does not support gather and gather_object

* fix gather

* fix collect_results_cpu

* fix collect_results and refactor unit tests

* fix collect_results unit tests

* handle torch.cat in torch1.5

* refine docstring

* refine docstring

* fix comments

* fix comments

* add user docs for get_config/get_models

* resolve comments