* fix wrong epoch andd iter when save bbest ckpt

* fix ut

* fix resume best ckpt unexpectedly

* minor refine

* fix unit test

* support training with data without metainfo

* clean the code

* clean the code

* [Enhance] Add `preprocess_cfg` as an argument of runner.

* Rename `preprocess_cfg` to `data_preprocessor`

* Fix docstring

* Support log enviroment information when initiate runner

* Fix unit test

* fix as comment, save world_size

* log gpu num

* clear code and reformat log

* minor refine

* fix as comment

* minor refine

* clean the code

* clean the code

* remove save world_size in meta

* add build function for scheduler

* add unit test

add unit test

* handle convert_to_iter in build_scheduler_from_cfg

* restore deleted code

* format import

* fix lint

* fix build multiple scheduler

* add new unit test

* fix comment and error message

* fix comment and error message

* extract _parse_scheduler_cfg

* always call build_param_scheduler during train and resume. If there is only one optimizer, the defaut value for sheduler will be a list, otherwise there is multiple optimizer, the default value of sheduler will be a dict

* minor refine

* rename runner test exp name

* fix as comment

* minor refine

* fix ut

* only check parameter scheduler

* minor refine

* make scheduler default to None

* fix bc breaking

* refine warning message

* fix as comment

* fix as comment

* fix lint

* [Fix] Fix weight initializing and registry logging.

* sync params

* resolve comments

* [Enhance] Auto set the `end` of param schedulers.

* Add log output and unit test

* Update docstring

* Update unit tests of `CosineAnnealingParamScheduler`.

* fix resume message_hub

* add unit test

* support resume from messagehub

* minor refine

* add comment

* fix typo

* update docstring

* fix missing device ids in wrap_model

* clean the code

* use default broadcast_buffers

* refine MMSeparateDistributedDataParallel

* rename tmp variable

* refine docstring

* add type hints

* refactor docstring of ddp model

* add  arg in docstring

* minor refine

* better ddp link

* fix build multiple runner error

* fix comments

* fix cpu ci

* [Feature]: Set different seed for diff rank

* [Feature]: Add log

* [Fix]: Fix lint

* [Fix]: Fix docstring

* [Fix]: Fix sampler seed

* [Fix]: Fix log bug

* [Fix]: Change diff_seed to diff_rank_seed

* [Fix]: Fix lint

* [Feat] Support revert syncbn

* use logger.info but not warning

* fix info string

* add autocast wrapper

* fix docstring

* fix docstring

* fix compare version

* fix unit test

* fix incompatible arguments

* fix as comment

* fix unit test

* rename auto_cast to autocast

* [Feat] Support save best ckpt

* reformat code

* rename function and reformat code

* fix logging info

* modify cuda() to to()

* rollback load_checkpoint

* refine runner

* add TODO

* move import resource

* move import resource

* [Enhance] dump messagehub in runner.resume

* delete unnecessary code

* delete debugging code

Co-authored-by: imabackstabber <312276423@qq.com>

* support resume from ceph

* move func and refine

* delete symlink

* fix unittest

* perserve _allow_symlink and symlink

* support mlu

* add ut and refine docstring

* merge context

* update unit test

* add docstring

* fix bug in AmpOptimWrapper

* add docstring for backward

* add warning and docstring for accumuate gradient

* fix docstring

* fix docstring

* add params_group method

* fix as comment

* fix as comment

* make default_value of loss_scale to dynamic

* Fix docstring

* decouple should update and should no sync

* rename attribute in OptimWrapper

* fix docstring

* fix comment

* fix comment

* fix as comment

* fix as comment and add unit test

* [Feat]:support customizing evaluator

* fix keyname of determining using default evaluator or not

* add assertion

* fix typo

* Replace auto_scale_lr_cfg to auto_scale_lr

* Update

* add base model, ddp model and unit test

* add unit test

* fix unit test

* fix docstring

* fix cpu unit test

* refine base data preprocessor

* refine base data preprocessor

* refine interface of ddp module

* remove optimizer hook

* add forward

* fix as comment

* fix unit test

* fix as comment

* fix build optimizer wrapper

* rebase main and fix unit test

* stack_batch support stacking ndim tensor, add docstring for merge dict

* fix lint

* fix test loop

* make precision_context effective to data_preprocessor

* fix as comment

* fix as comment

* refine docstring

* change collate_data output typehints

* rename to_rgb to bgr_to_rgb and rgb_to_bgr

* support build basemodel with built DataPreprocessor

* fix as comment

* fix docstring

* Add auto scale lr fucntion

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

Co-authored-by: wangjiabao1.vendor <wangjiabao@pjlab.org.cn>

* support custom runner

* change build_runner_from_cfg

* refine docstring

* refine docstring

* Modify val_interval and val_begin to be the attributes of TrainLoop.

* update doc

* fix lint

* type hint

* fix build optimizer wrapper without type

* refine logic

* fix as comment

* fix optim_wrapper config error in docstring and unit test

* refine docstring of build_optim_wrapper

* Support multiple optimizers

* minor refinement

* improve unit tests

* minor fix

* Update unit tests for resuming or saving ckpt for multiple optimizers

* refine docstring

* refine docstring

* fix typo

* update docstring

* refactor the logic to build multiple optimizers

* resolve comments

* ParamSchedulers spports multiple optimizers

* add optimizer_wrapper

* fix comment and docstirng

* fix unit test

* add unit test

* refine docstring

* RuntimeInfoHook supports printing multi learning rates

* resolve comments

* add optimizer_wrapper

* fix mypy

* fix lint

* fix OptimizerWrapperDict docstring and add unit test

* rename OptimizerWrapper to OptimWrapper, OptimWrapperDict inherit OptimWrapper, and fix as comment

* Fix AmpOptimizerWrapper

* rename build_optmizer_wrapper to build_optim_wrapper

* refine optimizer wrapper

* fix AmpOptimWrapper.step, docstring

* resolve confict

* rename DefaultOptimConstructor

* fix as comment

* rename clig grad auguments

* refactor optim_wrapper config

* fix docstring of DefaultOptimWrapperConstructor

fix docstring of DefaultOptimWrapperConstructor

* add get_lr method to OptimWrapper and OptimWrapperDict

* skip some amp unit test

* fix unit test

* fix get_lr, get_momentum docstring

* refactor get_lr, get_momentum, fix as comment

* fix error message

Co-authored-by: zhouzaida <zhouzaida@163.com>

* Support multiple optimizers

* minor refinement

* improve unit tests

* minor fix

* Update unit tests for resuming or saving ckpt for multiple optimizers

* refine docstring

* refine docstring

* fix typo

* update docstring

* refactor the logic to build multiple optimizers

* resolve comments

* ParamSchedulers spports multiple optimizers

* refine docstring

* RuntimeInfoHook supports printing multi learning rates

* resolve comments

* fix typo