def build(cfg, registry, default_args=None):
if isinstance(cfg, list):
modules = [
build_from_cfg(cfg_, registry, default_args) for cfg_ in cfg
]
return nn.Sequential(*modules)
else:
return build_from_cfg(cfg, registry, default_args)
def build(cfg, registry, default_args=None):
if isinstance(cfg, list): # 传进来的cfg.model是一个字典，不走这条路
modules = [
build_from_cfg(cfg_, registry, default_args) for cfg_ in cfg
]
return nn.Sequential(*modules)
else: # 所以调用的是build_from_cfg这个函数
return build_from_cfg(cfg, registry, default_args)
def build(cfg, registry, default_args=None):
## 主干是一个判断结构，其实就是判断传进来的cfg是字典列表还是单独的字典，来分情况处理。
# 字典列表的话：挨个调用build_from_cfg()，将其加到注册表******的_module_dict中，然后再返回return nn.Sequential(*modules)
# 字典的话：直接调用build_from_cfg()，将其添加到注册表DETECTORS中（以DETECTORS为例）。
if isinstance(cfg, list):
modules = [
build_from_cfg(cfg_, registry, default_args) for cfg_ in cfg
]
return nn.Sequential(*modules)
else:
return build_from_cfg(cfg, registry, default_args)
def build(cfg, registry, default_args=None):#建立模型的所有几口函数，
'''
函数的作用是用于构建网络的所有模块，他会根据模型被分成多少个部分，然后根据给定的参数字典来构建相应的部分，
如果给定的参数是一个list，那么他就生成了序列模块，不然就是用整个字典进行构建
'''
if isinstance(cfg, list):#如果给定的参数是列表格式，这个是模型中一小部分的构建
modules = [
build_from_cfg(cfg_, registry, default_args) for cfg_ in cfg
]#生成模块序列列表
return nn.Sequential(*modules)#生成序列模型
else:
return build_from_cfg(cfg, registry, default_args)#利用给定的字典生成模型，其中的registry是一个定义的类
def build(cfg, registry, default_args=None, parallel=None):
# not sequence, but parallel
# pdb.set_trace()
if isinstance(cfg, list):
modules = [
build_from_cfg(cfg_, registry, default_args) for cfg_ in cfg
]
if parallel:
return nn.ModuleList(modules)
else:
return nn.Sequential(*modules)
else:
return build_from_cfg(cfg, registry, default_args)
def build(cfg, registry, default_args=None):
# 多次进入build，每次cfg给一部分进行搭建(即cfg和registry的print是交替进行的，而非cfg一次输出完)
# print(cfg) # cfg.model -> cfg.model.backbone(ResNeXt) -> .neck(FPN) -> .ReinaHead(head) -> ...
if isinstance(cfg, list):
modules = [
build_from_cfg(cfg_, registry, default_args) for cfg_ in cfg
]
# print('1')
return nn.Sequential(*modules)
else:
'''CORE'''
# print('2') # √
# print(registry) # detector -> backbone -> neck -> head -> loss -> loss
return build_from_cfg(cfg, registry, default_args)
def build_dataset(cfg):
if cfg['type'] == 'RepeatDataset':
dataset = RepeatDataset(build_dataset(cfg['dataset']), cfg['times'])
elif isinstance(cfg['ann_file'], (list, tuple)):
dataset = _concat_dataset(cfg)
else:
dataset = build_from_cfg(cfg, DATASETS)
return dataset
def build_dataset(cfg, default_args=None):
if cfg['type'] == 'RepeatDataset':
dataset = RepeatDataset(build_dataset(cfg['dataset'], default_args),
cfg['times'])
elif isinstance(cfg['ann_file'], (list, tuple)):
dataset = _concat_dataset(cfg, default_args)
else:
# √
dataset = build_from_cfg(cfg, DATASETS, default_args)
return dataset
def __init__(self, transforms):
assert isinstance(transforms, collections.abc.Sequence)
self.transforms = []
for transform in transforms:
if isinstance(transform, dict):
transform = build_from_cfg(transform, PIPELINES)
self.transforms.append(transform)
elif callable(transform):
self.transforms.append(transform)
else:
raise TypeError('transform must be callable or a dict')
def build_dataset(cfg, default_args=None):
if isinstance(cfg, (list, tuple)): # if the cfg compose of several
dataset = ConcatDataset([build_dataset(c, default_args) for c in cfg])
elif cfg['type'] == 'RepeatDataset':
dataset = RepeatDataset(
build_dataset(cfg['dataset'], default_args), cfg['times'])
elif isinstance(cfg['ann_file'], (list, tuple)): # 进入到这个if中
dataset = _concat_dataset(cfg, default_args)
else:
dataset = build_from_cfg(cfg, DATASETS, default_args)
return dataset # 整合不同源数据集之后的数据集list，在这里都是coco源
def build_dataset(cfg, default_args=None):
if isinstance(cfg, (list, tuple)):
dataset = ConcatDataset([build_dataset(c, default_args) for c in cfg])
elif cfg['type'] == 'RepeatDataset':
dataset = RepeatDataset(
build_dataset(cfg['dataset'], default_args), cfg['times'])
elif isinstance(cfg['ann_file'], (list, tuple)):
dataset = _concat_dataset(cfg, default_args)
else:
dataset = build_from_cfg(cfg, DATASETS, default_args)
return dataset
# import copy
# from mmdet.utils import build_from_cfg
# from .dataset_wrappers import ConcatDataset, RepeatDataset
# from .registry import DATASETS
# def _concat_dataset(cfg, default_args=None):
# ann_files = cfg['ann_file']
# img_prefixes = cfg.get('img_prefix', None)
# seg_prefixes = cfg.get('seg_prefixes', None)
# proposal_files = cfg.get('proposal_file', None)
# datasets = []
# num_dset = len(ann_files)
# for i in range(num_dset):
# data_cfg = copy.deepcopy(cfg)
# data_cfg['ann_file'] = ann_files[i]
# if isinstance(img_prefixes, (list, tuple)):
# data_cfg['img_prefix'] = img_prefixes[i]
# if isinstance(seg_prefixes, (list, tuple)):
# data_cfg['seg_prefix'] = seg_prefixes[i]
# if isinstance(proposal_files, (list, tuple)):
# data_cfg['proposal_file'] = proposal_files[i]
# datasets.append(build_dataset(data_cfg, default_args))
# return ConcatDataset(datasets)
# def build_dataset(cfg, default_args=None):
# if isinstance(cfg, (list, tuple)):
# dataset = ConcatDataset([build_dataset(c, default_args) for c in cfg])
# elif cfg['type'] == 'RepeatDataset':
# dataset = RepeatDataset(
# build_dataset(cfg['dataset'], default_args), cfg['times'])
# elif isinstance(cfg['ann_file'], (list, tuple)):
# dataset = _concat_dataset(cfg, default_args)
# else:
# dataset = build_from_cfg(cfg, DATASETS, default_args)
# return dataset
def build_dataset(cfg, default_args=None):
if isinstance(cfg, (list, tuple)):
dataset = ConcatDataset([build_dataset(c, default_args) for c in cfg])
elif cfg['type'] == 'RepeatDataset':
dataset = RepeatDataset(build_dataset(cfg['dataset'], default_args),
cfg['times'])
elif isinstance(cfg.get('ann_file'), (tuple, list)):
dataset = _concat_dataset(cfg, default_args)
else:
dataset = build_from_cfg(cfg, DATASETS, default_args)
return dataset
def build_dataset(cfg, default_args=None):
if isinstance(cfg, (list, tuple)):
dataset = ConcatDataset([build_dataset(c, default_args) for c in cfg])
elif cfg["type"] == "RepeatDataset":
dataset = RepeatDataset(build_dataset(cfg["dataset"], default_args),
cfg["times"])
elif isinstance(cfg.get("ann_file"), (list, tuple)):
dataset = _concat_dataset(cfg, default_args)
else:
dataset = build_from_cfg(cfg, DATASETS, default_args)
return dataset
def build_dataset(cfg, default_args=None):
if isinstance(cfg, (list, tuple)):
dataset = ConcatDataset([build_dataset(c, default_args) for c in cfg])
elif cfg['type'] == 'RepeatDataset':
dataset = RepeatDataset(build_dataset(cfg['dataset'], default_args),
cfg['times'])
elif cfg['type'] == 'ClassBalancedDataset':
dataset = ClassBalancedDataset(
build_dataset(cfg['dataset'], default_args), cfg['oversample_thr'])
elif isinstance(cfg['ann_file'], (list, tuple)):
dataset = _concat_dataset(cfg, default_args)
else:
dataset = build_from_cfg(cfg, DATASETS, default_args)
return dataset
def __init__(self, transforms=None):
# You should load overlay object here.
print("%s __init__ called" % self.__class__.__name__)
assert transforms is None or isinstance(transforms,
collections.abc.Sequence)
self.transforms = []
if transforms is not None:
for transform in transforms:
if isinstance(transform, dict):
transform = build_from_cfg(transform, PIPELINES)
self.transforms.append(transform)
elif callable(transform):
self.transforms.append(transform)
else:
raise TypeError('transform must be callable or a dict')
def build_dataset(cfg, default_args=None):
if isinstance(cfg, (list, tuple)):
dataset = ConcatDataset([build_dataset(c, default_args) for c in cfg])
elif cfg['type'] == 'RepeatDataset':
dataset = RepeatDataset(
build_dataset(cfg['dataset'], default_args), cfg['times'])
elif isinstance(cfg['ann_file'], (list, tuple)):
dataset = _concat_dataset(cfg, default_args)
else:
matches = glob.glob(cfg['ann_file'], recursive=True)
if not matches:
raise RuntimeError(f'Failed to find annotation files that match pattern: '
f'{cfg["ann_file"]}')
cfg['ann_file'] = matches
if len(cfg['ann_file']) == 1:
cfg['ann_file'] = cfg['ann_file'][0]
dataset = build_from_cfg(cfg, DATASETS, default_args)
else:
dataset = _concat_dataset(cfg, default_args)
return dataset
def build_iou_calculator(cfg, default_args=None):
"""Builder of IoU calculator."""
return build_from_cfg(cfg, IOU_CALCULATORS, default_args)
def build_optimizer(model, optimizer_cfg):
"""Build optimizer from configs.
Args:
model (:obj:`nn.Module`): The model with parameters to be optimized.
optimizer_cfg (dict): The config dict of the optimizer.
Positional fields are:
- type: class name of the optimizer.
- lr: base learning rate.
Optional fields are:
- any arguments of the corresponding optimizer type, e.g.,
weight_decay, momentum, etc.
- paramwise_options: a dict with 4 accepted fileds
(bias_lr_mult, bias_decay_mult, norm_decay_mult,
dwconv_decay_mult).
`bias_lr_mult` and `bias_decay_mult` will be multiplied to
the lr and weight decay respectively for all bias parameters
(except for the normalization layers), and
`norm_decay_mult` will be multiplied to the weight decay
for all weight and bias parameters of normalization layers.
`dwconv_decay_mult` will be multiplied to the weight decay
for all weight and bias parameters of depthwise conv layers.
Returns:
torch.optim.Optimizer: The initialized optimizer.
Example:
>>> import torch
>>> model = torch.nn.modules.Conv1d(1, 1, 1)
>>> optimizer_cfg = dict(type='SGD', lr=0.01, momentum=0.9,
>>> weight_decay=0.0001)
>>> optimizer = build_optimizer(model, optimizer_cfg)
"""
if hasattr(model, 'module'):
model = model.module
optimizer_cfg = optimizer_cfg.copy()
paramwise_options = optimizer_cfg.pop('paramwise_options', None)
# if no paramwise option is specified, just use the global setting
if paramwise_options is None:
params = model.parameters()
else:
assert isinstance(paramwise_options, dict)
# get base lr and weight decay
base_lr = optimizer_cfg['lr']
base_wd = optimizer_cfg.get('weight_decay', None)
# weight_decay must be explicitly specified if mult is specified
if ('bias_decay_mult' in paramwise_options
or 'norm_decay_mult' in paramwise_options
or 'dwconv_decay_mult' in paramwise_options):
assert base_wd is not None
# get param-wise options
bias_lr_mult = paramwise_options.get('bias_lr_mult', 1.)
bias_decay_mult = paramwise_options.get('bias_decay_mult', 1.)
norm_decay_mult = paramwise_options.get('norm_decay_mult', 1.)
dwconv_decay_mult = paramwise_options.get('dwconv_decay_mult', 1.)
named_modules = dict(model.named_modules())
# set param-wise lr and weight decay
params = []
for name, param in model.named_parameters():
param_group = {'params': [param]}
if not param.requires_grad:
# FP16 training needs to copy gradient/weight between master
# weight copy and model weight, it is convenient to keep all
# parameters here to align with model.parameters()
params.append(param_group)
continue
# for norm layers, overwrite the weight decay of weight and bias
# TODO: obtain the norm layer prefixes dynamically
if re.search(r'(bn|gn)(\d+)?.(weight|bias)', name):
if base_wd is not None:
param_group['weight_decay'] = base_wd * norm_decay_mult
# for other layers, overwrite both lr and weight decay of bias
elif name.endswith('.bias'):
param_group['lr'] = base_lr * bias_lr_mult
if base_wd is not None:
param_group['weight_decay'] = base_wd * bias_decay_mult
module_name = name.replace('.weight', '').replace('.bias', '')
if module_name in named_modules and base_wd is not None:
module = named_modules[module_name]
# if this Conv2d is depthwise Conv2d
if isinstance(module, torch.nn.Conv2d) and \
module.in_channels == module.groups:
param_group['weight_decay'] = base_wd * dwconv_decay_mult
# otherwise use the global settings
params.append(param_group)
optimizer_cfg['params'] = params
return build_from_cfg(optimizer_cfg, OPTIMIZERS)
def build_sampler(cfg, **default_args):
"""Builder of box sampler."""
return build_from_cfg(cfg, BBOX_SAMPLERS, default_args)
def build_bbox_coder(cfg, **default_args):
"""Builder of box coder."""
return build_from_cfg(cfg, BBOX_CODERS, default_args)
def build_assigner(cfg, **default_args):
"""Builder of box assigner."""
return build_from_cfg(cfg, BBOX_ASSIGNERS, default_args)
def build_dataset(cfg, default_args=None):
dataset = build_from_cfg(cfg, DATASETS, default_args)
return dataset
