def main():
args = parse_args()
if len(args.shape) == 1:
input_shape = (3, args.shape[0], args.shape[0])
elif len(args.shape) == 2:
input_shape = (3, ) + tuple(args.shape)
elif len(args.shape) == 3:
input_shape = (1, ) + tuple(args.shape)
else:
raise ValueError('invalid input shape')
cfg = Config.fromfile(args.config)
model = build_classifier(cfg.model)
model.eval()
if hasattr(model, 'extract_feat'):
model.forward = model.extract_feat
else:
raise NotImplementedError(
'FLOPs counter is currently not currently supported with {}'.
format(model.__class__.__name__))
flops, params = get_model_complexity_info(model, input_shape)
split_line = '=' * 30
print(f'{split_line}\nInput shape: {input_shape}\n'
f'Flops: {flops}\nParams: {params}\n{split_line}')
print('!!!Please be cautious if you use the results in papers. '
'You may need to check if all ops are supported and verify that the '
'flops computation is correct.')
def init_model(config, checkpoint=None, device='cuda:0'):
"""Initialize a classifier from config file.
Args:
config (str or :obj:`mmcv.Config`): Config file path or the config
object.
checkpoint (str, optional): Checkpoint path. If left as None, the model
will not load any weights.
Returns:
nn.Module: The constructed classifier.
"""
if isinstance(config, str):
config = mmcv.Config.fromfile(config)
elif not isinstance(config, mmcv.Config):
raise TypeError('config must be a filename or Config object, '
f'but got {type(config)}')
config.model.pretrained = None
model = build_classifier(config.model)
if checkpoint is not None:
map_loc = 'cpu' if device == 'cpu' else None
checkpoint = load_checkpoint(model, checkpoint, map_location=map_loc)
if 'CLASSES' in checkpoint['meta']:
model.CLASSES = checkpoint['meta']['CLASSES']
else:
from mmcls.datasets import ImageNet
warnings.simplefilter('once')
warnings.warn('Class names are not saved in the checkpoint\'s '
'meta data, use imagenet by default.')
model.CLASSES = ImageNet.CLASSES
model.cfg = config # save the config in the model for convenience
model.to(device)
model.eval()
return model
def main():
args = parse_args()
# load test results
outputs = mmcv.load(args.result)
assert ('pred_score' in outputs and 'pred_class' in outputs
and 'pred_label' in outputs), \
'No "pred_label", "pred_score" or "pred_class" in result file, ' \
'please set "--out-items" in test.py'
cfg = mmcv.Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
model = build_classifier(cfg.model)
# build the dataloader
dataset = build_dataset(cfg.data.test)
filenames = list()
for info in dataset.data_infos:
if info['img_prefix'] is not None:
filename = osp.join(info['img_prefix'],
info['img_info']['filename'])
else:
filename = info['img_info']['filename']
filenames.append(filename)
gt_labels = list(dataset.get_gt_labels())
gt_classes = [dataset.CLASSES[x] for x in gt_labels]
outputs['filename'] = filenames
outputs['gt_label'] = gt_labels
outputs['gt_class'] = gt_classes
outputs_list = list()
for i in range(len(gt_labels)):
output = dict()
for k in outputs.keys():
output[k] = outputs[k][i]
outputs_list.append(output)
# sort result
outputs_list = sorted(outputs_list, key=lambda x: x['pred_score'])
success = list()
fail = list()
for output in outputs_list:
if output['pred_label'] == output['gt_label']:
success.append(output)
else:
fail.append(output)
success = success[:args.topk]
fail = fail[:args.topk]
save_imgs(args.out_dir, 'success', success, model)
save_imgs(args.out_dir, 'fail', fail, model)
def test_classifier_loss(self):
data = self.read_random_data()
data = scatter(data, ["cuda:0"])
data = data[0]
classifier = build_classifier(self.cfg.model)
classifier = MMDataParallel(module=classifier.cuda(0), device_ids=[0])
loss = classifier(**data)
for k, v in loss.items():
print(k)
print(v, v.dtype)
print("---")
def main():
args = parse_args()
if len(args.shape) == 1:
input_shape = (3, args.shape[0], args.shape[0])
elif len(args.shape) == 2:
input_shape = (3, ) + tuple(args.shape)
else:
raise ValueError('invalid input shape')
cfg = Config.fromfile(args.config)
if args.net_params:
tag, input_channels, block1, block2, block3, block4, last_channel = args.net_params.split(
'-')
input_channels = [int(item) for item in input_channels.split('_')]
block1 = [int(item) for item in block1.split('_')]
block2 = [int(item) for item in block2.split('_')]
block3 = [int(item) for item in block3.split('_')]
block4 = [int(item) for item in block4.split('_')]
last_channel = int(last_channel)
inverted_residual_setting = []
for item in [block1, block2, block3, block4]:
for _ in range(item[0]):
inverted_residual_setting.append([
item[1], item[2:-int(len(item) / 2 - 1)],
item[-int(len(item) / 2 - 1):]
])
cfg.model.backbone.input_channel = input_channels
cfg.model.backbone.inverted_residual_setting = inverted_residual_setting
cfg.model.backbone.last_channel = last_channel
cfg.model.head.in_channels = last_channel
model = build_classifier(cfg.model)
model.eval()
if hasattr(model, 'extract_feat'):
model.forward = model.extract_feat
else:
raise NotImplementedError(
'FLOPs counter is currently not currently supported with {}'.
format(model.__class__.__name__))
flops, params = get_model_complexity_info(model, input_shape)
split_line = '=' * 30
print(f'{split_line}\nInput shape: {input_shape}\n'
f'Flops: {flops}\nParams: {params}\n{split_line}')
print('!!!Please be cautious if you use the results in papers. '
'You may need to check if all ops are supported and verify that the '
'flops computation is correct.')
def load_model(checkpoint, config):
if isinstance(config, str):
config = mmcv.Config.fromfile(config)
elif not isinstance(config, mmcv.Config):
raise TypeError('config must be a filename or Config object, '
f'but got {type(config)}')
config.model.pretrained = None
model = build_classifier(config.model)
checkpoint = load_checkpoint(model, checkpoint, map_location="cpu")
model.cfg = config # save the config in the model for convenience
model.eval()
return model
def get_flops(cfg, input_shape):
model = build_classifier(cfg.model)
model.eval()
if hasattr(model, 'extract_feat'):
model.forward = model.extract_feat
else:
raise NotImplementedError(
'FLOPs counter is currently not currently supported with {}'.
format(model.__class__.__name__))
buf = io.StringIO()
all_flops, params = get_model_complexity_info(model, input_shape, print_per_layer_stat=True, as_strings=False, ost=buf)
buf = buf.getvalue()
return all_flops/1e9
def load_model(checkpoint, config):
if isinstance(config, str):
config = mmcv.Config.fromfile(config)
elif not isinstance(config, mmcv.Config):
raise TypeError('config must be a filename or Config object, '
f'but got {type(config)}')
config.model.pretrained = None
model = build_classifier(config.model)
checkpoint = load_checkpoint(model, checkpoint)
model.cfg = config # save the config in the model for convenience
if args.do_plda:
print("---INFO---: Far layer selected for plda")
model.neck.embedding[1] = torch.nn.Identity()
model.eval()
return model
def convert(config_file,
ckp_file,
onnx_file,
input_size=(224, 224, 3),
device='cpu'):
# Read mmdetection model
config = mmcv.Config.fromfile(config_file)
mmcls_model = build_classifier(config.model)
# Init weights:
# if ckp_file is None:
# mmcls_model.init_weights()
# Set forward function to dummy
if hasattr(mmcls_model, 'forward_test'):
mmcls_model.forward = mmcls_model.forward_test
# Generate dummy data
if len(input_size) == 0:
input_size = (224, 224, 3)
h, w, c = input_size
x = torch.rand(1, c, h, w, device=torch.device(device))
# Convert to onnx
with torch.no_grad():
torch.onnx.export(mmcls_model,
x,
onnx_file,
verbose=False,
input_names=['input'],
opset_version=9) # 9
# Deal with shared weights
rename_share_weights(onnx_file)
# Simplify nodes
ox_model = onnx.load(onnx_file)
model_simp, check = simplify(ox_model)
assert check, 'Simplified ONNX model could not be validated'
onnx.save(model_simp, onnx_file)
if isinstance(config_file, str):
print('Convert {} to {} success'.format(config_file, onnx_file))
else:
print('Convert to {} success'.format(onnx_file))
return (config_file, onnx_file)
def init_model(config, checkpoint=None, device='cuda:0', options=None):
"""Initialize a classifier from config file.
Args:
config (str or :obj:`mmcv.Config`): Config file path or the config
object.
checkpoint (str, optional): Checkpoint path. If left as None, the model
will not load any weights.
options (dict): Options to override some settings in the used config.
Returns:
nn.Module: The constructed classifier.
"""
if isinstance(config, str):
config = mmcv.Config.fromfile(config)
elif not isinstance(config, mmcv.Config):
raise TypeError('config must be a filename or Config object, '
f'but got {type(config)}')
if options is not None:
config.merge_from_dict(options)
config.model.pretrained = None
model = build_classifier(config.model)
if checkpoint is not None:
# Mapping the weights to GPU may cause unexpected video memory leak
# which refers to https://github.com/open-mmlab/mmdetection/pull/6405
checkpoint = load_checkpoint(model, checkpoint, map_location='cpu')
if 'CLASSES' in checkpoint.get('meta', {}):
model.CLASSES = checkpoint['meta']['CLASSES']
else:
from mmcls.datasets import ImageNet
warnings.simplefilter('once')
warnings.warn('Class names are not saved in the checkpoint\'s '
'meta data, use imagenet by default.')
model.CLASSES = ImageNet.CLASSES
model.cfg = config # save the config in the model for convenience
model.to(device)
model.eval()
return model
stem_channels=32,
base_channels=32,
depth=34,
num_stages=4,
out_indices=(3, ),
style='pytorch',
with_cp=True,
norm_cfg=dict(type='BN',
requires_grad=True,
momentum=0.5)),
neck=dict(type='ECAPA_ASP',
emb_dim=256,
in_plane=2816,
emb_bn=True,
activation_type="relu"),
head=dict(type='AMSoftmaxClsHead',
num_classes=5994,
in_channels=256,
loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
topk=(1, 5)))
if __name__ == "__main__":
import torch
from mmcls.models import build_backbone, build_neck, build_classifier
with torch.no_grad():
model = build_classifier(model)
#print(model)
input = torch.rand(2, 200, 81)
gt_label = torch.LongTensor([1, 2])
print(model(input, gt_label=gt_label))
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# set multi-process settings
setup_multi_processes(cfg)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids[0:1]
warnings.warn('`--gpu-ids` is deprecated, please use `--gpu-id`. '
'Because we only support single GPU mode in '
'non-distributed testing. Use the first GPU '
'in `gpu_ids` now.')
else:
cfg.gpu_ids = [args.gpu_id]
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
dataset = build_dataset(cfg.data.test, default_args=dict(test_mode=True))
# the extra round_up data will be removed during gpu/cpu collect
data_loader = build_dataloader(dataset,
samples_per_gpu=cfg.data.samples_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False,
round_up=True)
# build the model and load checkpoint
model = build_classifier(cfg.model)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
if 'CLASSES' in checkpoint.get('meta', {}):
CLASSES = checkpoint['meta']['CLASSES']
else:
from mmcls.datasets import ImageNet
warnings.simplefilter('once')
warnings.warn('Class names are not saved in the checkpoint\'s '
'meta data, use imagenet by default.')
CLASSES = ImageNet.CLASSES
if not distributed:
if args.device == 'cpu':
model = model.cpu()
else:
model = MMDataParallel(model, device_ids=cfg.gpu_ids)
if not model.device_ids:
assert mmcv.digit_version(mmcv.__version__) >= (1, 4, 4), \
'To test with CPU, please confirm your mmcv version ' \
'is not lower than v1.4.4'
model.CLASSES = CLASSES
show_kwargs = {} if args.show_options is None else args.show_options
outputs = single_gpu_test(model, data_loader, args.show, args.show_dir,
**show_kwargs)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
results = {}
logger = get_root_logger()
if args.metrics:
eval_results = dataset.evaluate(results=outputs,
metric=args.metrics,
metric_options=args.metric_options,
logger=logger)
results.update(eval_results)
for k, v in eval_results.items():
if isinstance(v, np.ndarray):
v = [round(out, 2) for out in v.tolist()]
elif isinstance(v, Number):
v = round(v, 2)
else:
raise ValueError(f'Unsupport metric type: {type(v)}')
print(f'\n{k} : {v}')
if args.out:
if 'none' not in args.out_items:
scores = np.vstack(outputs)
pred_score = np.max(scores, axis=1)
pred_label = np.argmax(scores, axis=1)
pred_class = [CLASSES[lb] for lb in pred_label]
res_items = {
'class_scores': scores,
'pred_score': pred_score,
'pred_label': pred_label,
'pred_class': pred_class
}
if 'all' in args.out_items:
results.update(res_items)
else:
for key in args.out_items:
results[key] = res_items[key]
print(f'\ndumping results to {args.out}')
mmcv.dump(results, args.out)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.aml:
data_store = os.environ['AZUREML_DATAREFERENCE_{}'.format(
args.aml_data_store)]
parse(cfg, data_store)
if cfg.resume_from is not None:
cfg.resume_from = os.path.join(data_store,
args.aml_work_dir_prefix,
cfg.resume_from)
cfg.work_dir = os.path.join(data_store, args.aml_work_dir_prefix,
cfg.work_dir)
print('work_dir: ', cfg.work_dir)
if 'data' in cfg.model.pretrained:
cfg.model.pretrained = os.path.join(data_store,
cfg.model.pretrained)
# if not args.aml:
# # work_dir is determined in this priority: CLI > segment in file > filename
# if args.work_dir is not None:
# # update configs according to CLI args if args.work_dir is not None
# cfg.work_dir = args.work_dir
# elif cfg.get('work_dir', None) is None:
# # use config filename as default work_dir if cfg.work_dir is None
# cfg.work_dir = osp.join('./work_dirs',
# osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# update gpu num
if dist.is_initialized():
cfg.gpus = dist.get_world_size()
else:
cfg.gpus = args.gpus
if args.autoscale_lr:
# apply the linear scaling rule (https://arxiv.org/abs/1706.02677)
cfg.optimizer['lr'] = cfg.optimizer[
'lr'] * cfg.gpus / 8 * cfg.data.samples_per_gpu / 32
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
model = build_classifier(cfg.model)
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmcls version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(mmcls_version=__version__,
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES)
# add an attribute for visualization convenience
train_model(model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
meta=meta)
def main():
print("--------> 1")
args = parse_args()
print("--------> 2")
cfg = Config.fromfile(args.config)
print("--------> 3")
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
print("--------> 4")
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
print("--------> 5")
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
_, world_size = get_dist_info()
cfg.gpu_ids = range(world_size)
print("--------> 6")
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
print("--------> 7")
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
cfg_pretty_text = cfg.pretty_text
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg_pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
model = build_classifier(cfg.model)
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmcls version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(mmcls_version=__version__,
config=cfg_pretty_text,
CLASSES=datasets[0].CLASSES)
model.CLASSES = datasets[0].CLASSES
# add an attribute for visualization convenience
train_model(model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
device='cpu' if args.device == 'cpu' else 'cuda',
meta=meta)
if __name__ == '__main__':
args = parse_args()
if len(args.shape) == 1:
input_shape = (1, 3, args.shape[0], args.shape[0])
elif len(args.shape) == 2:
input_shape = (
1,
3,
) + tuple(args.shape)
else:
raise ValueError('invalid input shape')
cfg = mmcv.Config.fromfile(args.config)
cfg.model.pretrained = None
# build the model and load checkpoint
classifier = build_classifier(cfg.model)
if args.checkpoint:
load_checkpoint(classifier, args.checkpoint, map_location='cpu')
# conver model to TorchScript file
pytorch2torchscript(
classifier,
input_shape,
output_file=args.output_file,
verify=args.verify)
def train_single_fold(args, cfg, fold, distributed, seed):
# create the work_dir for the fold
work_dir = osp.join(cfg.work_dir, f'fold{fold}')
cfg.work_dir = work_dir
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# wrap the dataset cfg
train_dataset = dict(
type='KFoldDataset',
fold=fold,
dataset=cfg.data.train,
num_splits=args.num_splits,
seed=seed,
)
val_dataset = dict(
type='KFoldDataset',
fold=fold,
# Use the same dataset with training.
dataset=copy.deepcopy(cfg.data.train),
num_splits=args.num_splits,
seed=seed,
test_mode=True,
)
val_dataset['dataset']['pipeline'] = cfg.data.val.pipeline
cfg.data.train = train_dataset
cfg.data.val = val_dataset
cfg.data.test = val_dataset
# dump config
stem, suffix = osp.basename(args.config).rsplit('.', 1)
cfg.dump(osp.join(cfg.work_dir, f'{stem}_fold{fold}.{suffix}'))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
logger.info(
f'-------- Cross-validation: [{fold+1}/{args.num_splits}] -------- ')
# set random seeds
# Use different seed in different folds
logger.info(f'Set random seed to {seed + fold}, '
f'deterministic: {args.deterministic}')
set_random_seed(seed + fold, deterministic=args.deterministic)
cfg.seed = seed + fold
meta['seed'] = seed + fold
model = build_classifier(cfg.model)
model.init_weights()
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
meta.update(
dict(mmcls_version=__version__,
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES,
kfold=dict(fold=fold, num_splits=args.num_splits)))
# add an attribute for visualization convenience
train_model(model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
device='cpu' if args.device == 'cpu' else 'cuda',
meta=meta)
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
samples_per_gpu=cfg.data.samples_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False,
round_up=False)
# build the model and load checkpoint
model = build_classifier(cfg.model)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
if args.metrics:
results = dataset.evaluate(outputs, args.metrics,
args.metric_options)
for k, v in results.items():
print(f'\n{k} : {v:.2f}')
else:
warnings.warn('Evaluation metrics are not specified.')
scores = np.vstack(outputs)
pred_score = np.max(scores, axis=1)
pred_label = np.argmax(scores, axis=1)
if 'CLASSES' in checkpoint['meta']:
CLASSES = checkpoint['meta']['CLASSES']
else:
from mmcls.datasets import ImageNet
warnings.simplefilter('once')
warnings.warn('Class names are not saved in the checkpoint\'s '
'meta data, use imagenet by default.')
CLASSES = ImageNet.CLASSES
pred_class = [CLASSES[lb] for lb in pred_label]
results = {
'pred_score': pred_score,
'pred_label': pred_label,
'pred_class': pred_class
}
if not args.out:
print('\nthe predicted result for the first element is '
f'pred_score = {pred_score[0]:.2f}, '
f'pred_label = {pred_label[0]} '
f'and pred_class = {pred_class[0]}. '
'Specify --out to save all results to files.')
if args.out and rank == 0:
print(f'\nwriting results to {args.out}')
mmcv.dump(results, args.out)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# set multi-process settings
setup_multi_processes(cfg)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpus is not None:
cfg.gpu_ids = range(1)
warnings.warn('`--gpus` is deprecated because we only support '
'single GPU mode in non-distributed training. '
'Use `gpus=1` now.')
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids[0:1]
warnings.warn('`--gpu-ids` is deprecated, please use `--gpu-id`. '
'Because we only support single GPU mode in '
'non-distributed training. Use the first GPU '
'in `gpu_ids` now.')
if args.gpus is None and args.gpu_ids is None:
cfg.gpu_ids = [args.gpu_id]
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
_, world_size = get_dist_info()
cfg.gpu_ids = range(world_size)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# dump config
cfg.dump(osp.join(cfg.work_dir, osp.basename(args.config)))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
seed = init_random_seed(args.seed)
logger.info(f'Set random seed to {seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(seed, deterministic=args.deterministic)
cfg.seed = seed
meta['seed'] = seed
model = build_classifier(cfg.model)
model.init_weights()
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
# save mmcls version, config file content and class names in
# runner as meta data
meta.update(
dict(mmcls_version=__version__,
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES))
# add an attribute for visualization convenience
train_model(model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
device='cpu' if args.device == 'cpu' else 'cuda',
meta=meta)
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
dataset = build_dataset(cfg.data.test)
data_loader = build_dataloader(dataset,
samples_per_gpu=cfg.data.samples_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False,
round_up=False)
# build the model and load checkpoint
model = build_classifier(cfg.model)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
_ = load_checkpoint(model, args.checkpoint, map_location='cpu')
if not distributed:
model = MMDataParallel(model, device_ids=[0])
outputs = single_gpu_test(model, data_loader)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
nums = []
results = {}
for output in outputs:
nums.append(output['num_samples'].item())
for topk, v in output['accuracy'].items():
if topk not in results:
results[topk] = []
results[topk].append(v.item())
assert sum(nums) == len(dataset)
for topk, accs in results.items():
avg_acc = np.average(accs, weights=nums)
print(f'\n{topk} accuracy: {avg_acc:.2f}')
if args.out and rank == 0:
print(f'\nwriting results to {args.out}')
mmcv.dump(outputs, args.out)
def test_classifier(self):
classifier = build_classifier(self.cfg.model)
img_w, img_h = 224, 224
input_x = torch.randn(4, 3, img_h, img_w)
yaw, pitch, roll = classifier.simple_test(input_x)
print(F"yaw: {yaw}, pitch: {pitch}, roll: {roll}")
def main():
args = parse_args()
cfg = mmcv.Config.fromfile(args.config)
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
cfg.model.pretrained = None
cfg.data.test.test_mode = True
assert args.metrics or args.out, \
'Please specify at least one of output path and evaluation metrics.'
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# build the dataloader
dataset = build_dataset(cfg.data.test)
# the extra round_up data will be removed during gpu/cpu collect
data_loader = build_dataloader(dataset,
samples_per_gpu=cfg.data.samples_per_gpu,
workers_per_gpu=cfg.data.workers_per_gpu,
dist=distributed,
shuffle=False,
round_up=True)
# build the model and load checkpoint
model = build_classifier(cfg.model)
fp16_cfg = cfg.get('fp16', None)
if fp16_cfg is not None:
wrap_fp16_model(model)
checkpoint = load_checkpoint(model, args.checkpoint, map_location='cpu')
if 'CLASSES' in checkpoint.get('meta', {}):
CLASSES = checkpoint['meta']['CLASSES']
else:
from mmcls.datasets import ImageNet
warnings.simplefilter('once')
warnings.warn('Class names are not saved in the checkpoint\'s '
'meta data, use imagenet by default.')
CLASSES = ImageNet.CLASSES
if not distributed:
if args.device == 'cpu':
model = model.cpu()
else:
model = MMDataParallel(model, device_ids=[0])
model.CLASSES = CLASSES
show_kwargs = {} if args.show_options is None else args.show_options
outputs = single_gpu_test(model, data_loader, args.show, args.show_dir,
**show_kwargs)
else:
model = MMDistributedDataParallel(
model.cuda(),
device_ids=[torch.cuda.current_device()],
broadcast_buffers=False)
outputs = multi_gpu_test(model, data_loader, args.tmpdir,
args.gpu_collect)
rank, _ = get_dist_info()
if rank == 0:
results = {}
if args.metrics:
eval_results = dataset.evaluate(outputs, args.metrics,
args.metric_options)
results.update(eval_results)
for k, v in eval_results.items():
print(f'\n{k} : {v:.2f}')
if args.out:
scores = np.vstack(outputs)
pred_score = np.max(scores, axis=1)
pred_label = np.argmax(scores, axis=1)
pred_class = [CLASSES[lb] for lb in pred_label]
results.update({
'class_scores': scores,
'pred_score': pred_score,
'pred_label': pred_label,
'pred_class': pred_class
})
print(f'\ndumping results to {args.out}')
mmcv.dump(results, args.out)
def main():
args = parse_args()
cfg = Config.fromfile(args.config)
if args.options is not None:
cfg.merge_from_dict(args.options)
# set cudnn_benchmark
if cfg.get('cudnn_benchmark', False):
torch.backends.cudnn.benchmark = True
if args.net_params:
tag, input_channels, block1, block2, block3, block4, last_channel = args.net_params.split('-')
input_channels = [int(item) for item in input_channels.split('_')]
block1 = [int(item) for item in block1.split('_')]
block2 = [int(item) for item in block2.split('_')]
block3 = [int(item) for item in block3.split('_')]
block4 = [int(item) for item in block4.split('_')]
last_channel = int(last_channel)
inverted_residual_setting = []
for item in [block1, block2, block3, block4]:
for _ in range(item[0]):
inverted_residual_setting.append([item[1], item[2:-int(len(item)/2-1)], item[-int(len(item)/2-1):]])
cfg.model.backbone.input_channel = input_channels
cfg.model.backbone.inverted_residual_setting = inverted_residual_setting
cfg.model.backbone.last_channel = last_channel
cfg.model.head.in_channels = last_channel
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
if args.resume_from is not None:
cfg.resume_from = args.resume_from
if args.gpu_ids is not None:
cfg.gpu_ids = args.gpu_ids
else:
cfg.gpu_ids = range(1) if args.gpus is None else range(args.gpus)
if args.autoscale_lr:
# apply the linear scaling rule (https://arxiv.org/abs/1706.02677)
cfg.optimizer['lr'] = cfg.optimizer['lr'] * len(cfg.gpu_ids) / 8
# init distributed env first, since logger depends on the dist info.
if args.launcher == 'none':
distributed = False
else:
distributed = True
init_dist(args.launcher, **cfg.dist_params)
# create work_dir
mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
# init the logger before other steps
timestamp = time.strftime('%Y%m%d_%H%M%S', time.localtime())
if args.net_params:
log_file = osp.join(cfg.work_dir, f'{args.net_params}.log')
else:
log_file = osp.join(cfg.work_dir, f'{timestamp}.log')
logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
# init the meta dict to record some important information such as
# environment info and seed, which will be logged
meta = dict()
# log env info
env_info_dict = collect_env()
env_info = '\n'.join([(f'{k}: {v}') for k, v in env_info_dict.items()])
dash_line = '-' * 60 + '\n'
logger.info('Environment info:\n' + dash_line + env_info + '\n' +
dash_line)
meta['env_info'] = env_info
# log some basic info
logger.info(f'Distributed training: {distributed}')
logger.info(f'Config:\n{cfg.pretty_text}')
# set random seeds
if args.seed is not None:
logger.info(f'Set random seed to {args.seed}, '
f'deterministic: {args.deterministic}')
set_random_seed(args.seed, deterministic=args.deterministic)
cfg.seed = args.seed
meta['seed'] = args.seed
model = build_classifier(cfg.model)
datasets = [build_dataset(cfg.data.train)]
if len(cfg.workflow) == 2:
val_dataset = copy.deepcopy(cfg.data.val)
val_dataset.pipeline = cfg.data.train.pipeline
datasets.append(build_dataset(val_dataset))
if cfg.checkpoint_config is not None:
# save mmcls version, config file content and class names in
# checkpoints as meta data
cfg.checkpoint_config.meta = dict(
mmcls_version=__version__,
config=cfg.pretty_text,
CLASSES=datasets[0].CLASSES)
# add an attribute for visualization convenience
train_model(
model,
datasets,
cfg,
distributed=distributed,
validate=(not args.no_validate),
timestamp=timestamp,
meta=meta)
