cfg_from_list(args.set_cfgs)
if cfg.MODEL.IDENTITY_TRAINING and cfg.MODEL.IDENTITY_REPLACE_CLASS:
cfg.MODEL.NUM_CLASSES = 145
cfg.MODEL.IDENTITY_TRAINING = False
cfg.MODEL.ADD_UNKNOWN_CLASS = False
#Add unknow class type if necessary.
if cfg.MODEL.IDENTITY_TRAINING:
cfg.MODEL.TOTAL_INSTANCE_NUM = 145
if cfg.MODEL.ADD_UNKNOWN_CLASS is True:
cfg.MODEL.NUM_CLASSES += 1
assert bool(args.load_ckpt)
assert_and_infer_cfg()
maskRCNN = vos_model_builder.Generalized_VOS_RCNN()
if args.cuda:
maskRCNN.cuda()
if args.load_ckpt:
load_name = args.load_ckpt
print("loading checkpoint %s" % (load_name))
checkpoint = torch.load(load_name,
map_location=lambda storage, loc: storage)
net_utils.load_ckpt_no_mapping(maskRCNN, checkpoint['model'])
maskRCNN = mynn.DataParallel(maskRCNN,
cpu_keywords=['im_info', 'roidb'],
minibatch=True,
device_ids=[0])
def main():
"""main function"""
if not torch.cuda.is_available():
sys.exit("Need a CUDA device to run the code.")
args = parse_args()
print('Called with args:')
print(args)
assert args.image_dir or args.images
assert bool(args.image_dir) ^ bool(args.images)
if args.dataset.startswith("coco"):
dataset = datasets.get_coco_dataset()
cfg.MODEL.NUM_CLASSES = len(dataset.classes)
elif args.dataset.startswith("keypoints_coco"):
dataset = datasets.get_coco_dataset()
cfg.MODEL.NUM_CLASSES = 2
else:
raise ValueError('Unexpected dataset name: {}'.format(args.dataset))
print('load cfg from file: {}'.format(args.cfg_file))
cfg_from_file(args.cfg_file)
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs)
assert bool(args.load_ckpt) ^ bool(args.load_detectron), \
'Exactly one of --load_ckpt and --load_detectron should be specified.'
cfg.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS = False # Don't need to load imagenet pretrained weights
assert_and_infer_cfg()
maskRCNN = vos_model_builder.Generalized_VOS_RCNN()
if args.cuda:
maskRCNN.cuda()
if args.load_ckpt:
load_name = args.load_ckpt
print("loading checkpoint %s" % (load_name))
checkpoint = torch.load(load_name, map_location=lambda storage, loc: storage)
net_utils.load_ckpt_no_mapping(maskRCNN, checkpoint['model'])
if args.load_detectron:
print("loading detectron weights %s" % args.load_detectron)
load_detectron_weight(maskRCNN, args.load_detectron)
maskRCNN = mynn.DataParallel(maskRCNN, cpu_keywords=['im_info', 'roidb'],
minibatch=True, device_ids=[0]) # only support single GPU
maskRCNN.eval()
if args.image_dir:
imglist = misc_utils.get_imagelist_from_dir(args.image_dir)
else:
imglist = args.images
num_images = len(imglist)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
for i in xrange(num_images):
print('img', i)
im = cv2.imread(imglist[i])
assert im is not None
timers = defaultdict(Timer)
cls_boxes, cls_segms, cls_keyps = im_detect_all(maskRCNN, im, timers=timers)
im_name, _ = os.path.splitext(os.path.basename(imglist[i]))
vis_utils.vis_one_image(
im[:, :, ::-1], # BGR -> RGB for visualization
im_name,
args.output_dir,
cls_boxes,
cls_segms,
cls_keyps,
dataset=dataset,
box_alpha=0.3,
show_class=True,
thresh=0.7,
kp_thresh=2
)
if args.merge_pdfs and num_images > 1:
merge_out_path = '{}/results.pdf'.format(args.output_dir)
if os.path.exists(merge_out_path):
os.remove(merge_out_path)
command = "pdfunite {}/*.pdf {}".format(args.output_dir,
merge_out_path)
subprocess.call(command, shell=True)
def main():
"""Main function"""
args = parse_args()
print('Called with args:')
print(args)
if not torch.cuda.is_available():
sys.exit("Need a CUDA device to run the code.")
if args.cuda or cfg.NUM_GPUS > 0:
cfg.CUDA = True
else:
raise ValueError("Need Cuda device to run !")
if args.dataset == "coco2017":
cfg.TRAIN.DATASETS = ('coco_2017_train', )
cfg.MODEL.NUM_CLASSES = 81 #80 foreground + 1 background
elif args.dataset == "keypoints_coco2017":
cfg.TRAIN.DATASETS = ('keypoints_coco_2017_train', )
cfg.MODEL.NUM_CLASSES = 2
else:
raise ValueError("Unexpected args.dataset: {}".format(args.dataset))
#Add unknow class type if necessary.
if cfg.MODEL.ADD_UNKNOWN_CLASS is True:
cfg.MODEL.NUM_CLASSES += 1
cfg_from_file(args.cfg_file)
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs)
### Adaptively adjust some configs ###
original_batch_size = cfg.NUM_GPUS * cfg.TRAIN.IMS_PER_BATCH
original_ims_per_batch = cfg.TRAIN.IMS_PER_BATCH
original_num_gpus = cfg.NUM_GPUS
if args.batch_size is None:
args.batch_size = original_batch_size
cfg.NUM_GPUS = torch.cuda.device_count()
assert (args.batch_size % cfg.NUM_GPUS) == 0, \
'batch_size: %d, NUM_GPUS: %d' % (args.batch_size, cfg.NUM_GPUS)
cfg.TRAIN.IMS_PER_BATCH = args.batch_size // cfg.NUM_GPUS
effective_batch_size = args.iter_size * args.batch_size
print('effective_batch_size = batch_size * iter_size = %d * %d' %
(args.batch_size, args.iter_size))
print('Adaptive config changes:')
print(' effective_batch_size: %d --> %d' %
(original_batch_size, effective_batch_size))
print(' NUM_GPUS: %d --> %d' %
(original_num_gpus, cfg.NUM_GPUS))
print(' IMS_PER_BATCH: %d --> %d' %
(original_ims_per_batch, cfg.TRAIN.IMS_PER_BATCH))
### Adjust learning based on batch size change linearly
# For iter_size > 1, gradients are `accumulated`, so lr is scaled based
# on batch_size instead of effective_batch_size
old_base_lr = cfg.SOLVER.BASE_LR
cfg.SOLVER.BASE_LR *= args.batch_size / original_batch_size
print('Adjust BASE_LR linearly according to batch_size change:\n'
' BASE_LR: {} --> {}'.format(old_base_lr, cfg.SOLVER.BASE_LR))
### Adjust solver steps
step_scale = original_batch_size / effective_batch_size
old_solver_steps = cfg.SOLVER.STEPS
old_max_iter = cfg.SOLVER.MAX_ITER
cfg.SOLVER.STEPS = list(
map(lambda x: int(x * step_scale + 0.5), cfg.SOLVER.STEPS))
cfg.SOLVER.MAX_ITER = int(cfg.SOLVER.MAX_ITER * step_scale + 0.5)
print(
'Adjust SOLVER.STEPS and SOLVER.MAX_ITER linearly based on effective_batch_size change:\n'
' SOLVER.STEPS: {} --> {}\n'
' SOLVER.MAX_ITER: {} --> {}'.format(old_solver_steps,
cfg.SOLVER.STEPS, old_max_iter,
cfg.SOLVER.MAX_ITER))
# Scale FPN rpn_proposals collect size (post_nms_topN) in `collect` function
# of `collect_and_distribute_fpn_rpn_proposals.py`
#
# post_nms_topN = int(cfg[cfg_key].RPN_POST_NMS_TOP_N * cfg.FPN.RPN_COLLECT_SCALE + 0.5)
if cfg.FPN.FPN_ON and cfg.MODEL.FASTER_RCNN:
cfg.FPN.RPN_COLLECT_SCALE = cfg.TRAIN.IMS_PER_BATCH / original_ims_per_batch
print(
'Scale FPN rpn_proposals collect size directly propotional to the change of IMS_PER_BATCH:\n'
' cfg.FPN.RPN_COLLECT_SCALE: {}'.format(
cfg.FPN.RPN_COLLECT_SCALE))
if args.num_workers is not None:
cfg.DATA_LOADER.NUM_THREADS = args.num_workers
print('Number of data loading threads: %d' % cfg.DATA_LOADER.NUM_THREADS)
### Overwrite some solver settings from command line arguments
if args.optimizer is not None:
cfg.SOLVER.TYPE = args.optimizer
if args.lr is not None:
cfg.SOLVER.BASE_LR = args.lr
if args.lr_decay_gamma is not None:
cfg.SOLVER.GAMMA = args.lr_decay_gamma
assert_and_infer_cfg()
timers = defaultdict(Timer)
### Dataset ###
timers['roidb'].tic()
roidb, ratio_list, ratio_index = combined_roidb_for_training(
cfg.TRAIN.DATASETS, cfg.TRAIN.PROPOSAL_FILES)
timers['roidb'].toc()
roidb_size = len(roidb)
logger.info('{:d} roidb entries'.format(roidb_size))
logger.info('Takes %.2f sec(s) to construct roidb',
timers['roidb'].average_time)
# Effective training sample size for one epoch
train_size = roidb_size // args.batch_size * args.batch_size
batchSampler = BatchSampler(sampler=MinibatchSampler(
ratio_list, ratio_index),
batch_size=args.batch_size,
drop_last=True)
dataset = RoiDataLoader(roidb, cfg.MODEL.NUM_CLASSES, training=True)
dataloader = torch.utils.data.DataLoader(
dataset,
batch_sampler=batchSampler,
num_workers=cfg.DATA_LOADER.NUM_THREADS,
collate_fn=collate_minibatch)
dataiterator = iter(dataloader)
### Model ###
maskRCNN = vos_model_builder.Generalized_VOS_RCNN()
if cfg.CUDA:
maskRCNN.cuda()
### Optimizer ###
gn_param_nameset = set()
for name, module in maskRCNN.named_modules():
if isinstance(module, nn.GroupNorm):
gn_param_nameset.add(name + '.weight')
gn_param_nameset.add(name + '.bias')
gn_params = []
gn_param_names = []
bias_params = []
bias_param_names = []
nonbias_params = []
nonbias_param_names = []
nograd_param_names = []
for key, value in maskRCNN.named_parameters():
if value.requires_grad:
if 'bias' in key:
bias_params.append(value)
bias_param_names.append(key)
elif key in gn_param_nameset:
gn_params.append(value)
gn_param_names.append(key)
else:
nonbias_params.append(value)
nonbias_param_names.append(key)
else:
nograd_param_names.append(key)
assert (gn_param_nameset - set(nograd_param_names) -
set(bias_param_names)) == set(gn_param_names)
# Learning rate of 0 is a dummy value to be set properly at the start of training
params = [{
'params': nonbias_params,
'lr': 0,
'weight_decay': cfg.SOLVER.WEIGHT_DECAY
}, {
'params':
bias_params,
'lr':
0 * (cfg.SOLVER.BIAS_DOUBLE_LR + 1),
'weight_decay':
cfg.SOLVER.WEIGHT_DECAY if cfg.SOLVER.BIAS_WEIGHT_DECAY else 0
}, {
'params': gn_params,
'lr': 0,
'weight_decay': cfg.SOLVER.WEIGHT_DECAY_GN
}]
# names of paramerters for each paramter
param_names = [nonbias_param_names, bias_param_names, gn_param_names]
if cfg.SOLVER.TYPE == "SGD":
optimizer = torch.optim.SGD(params, momentum=cfg.SOLVER.MOMENTUM)
elif cfg.SOLVER.TYPE == "Adam":
optimizer = torch.optim.Adam(params)
### Load checkpoint
if args.load_ckpt:
load_name = args.load_ckpt
logging.info("loading checkpoint %s", load_name)
checkpoint = torch.load(load_name,
map_location=lambda storage, loc: storage)
#TODO: load_ckpt_no_mapping(maskRCNN, checkpoint['model'])
net_utils.load_ckpt_no_mapping(maskRCNN, checkpoint['model'])
if args.resume:
args.start_step = checkpoint['step'] + 1
if 'train_size' in checkpoint: # For backward compatibility
if checkpoint['train_size'] != train_size:
print(
'train_size value: %d different from the one in checkpoint: %d'
% (train_size, checkpoint['train_size']))
# reorder the params in optimizer checkpoint's params_groups if needed
# misc_utils.ensure_optimizer_ckpt_params_order(param_names, checkpoint)
# There is a bug in optimizer.load_state_dict on Pytorch 0.3.1.
# However it's fixed on master.
optimizer.load_state_dict(checkpoint['optimizer'])
# misc_utils.load_optimizer_state_dict(optimizer, checkpoint['optimizer'])
del checkpoint
torch.cuda.empty_cache()
if args.load_detectron: #TODO resume for detectron weights (load sgd momentum values)
logging.info("loading Detectron weights %s", args.load_detectron)
load_detectron_weight(maskRCNN,
args.load_detectron,
force_load_all=False)
lr = optimizer.param_groups[0][
'lr'] # lr of non-bias parameters, for commmand line outputs.
maskRCNN = mynn.DataParallel(maskRCNN,
cpu_keywords=['im_info', 'roidb'],
minibatch=True)
### Training Setups ###
args.run_name = misc_utils.get_run_name() + '_step'
output_dir = misc_utils.get_output_dir(args, args.run_name)
args.cfg_filename = os.path.basename(args.cfg_file)
if not args.no_save:
if not os.path.exists(output_dir):
os.makedirs(output_dir)
blob = {'cfg': yaml.dump(cfg), 'args': args}
with open(os.path.join(output_dir, 'config_and_args.pkl'), 'wb') as f:
pickle.dump(blob, f, pickle.HIGHEST_PROTOCOL)
if args.use_tfboard:
from tensorboardX import SummaryWriter
# Set the Tensorboard logger
tblogger = SummaryWriter(output_dir)
### Training Loop ###
maskRCNN.train()
CHECKPOINT_PERIOD = int(cfg.TRAIN.SNAPSHOT_ITERS / cfg.NUM_GPUS)
# Set index for decay steps
decay_steps_ind = None
for i in range(1, len(cfg.SOLVER.STEPS)):
if cfg.SOLVER.STEPS[i] >= args.start_step:
decay_steps_ind = i
break
if decay_steps_ind is None:
decay_steps_ind = len(cfg.SOLVER.STEPS)
training_stats = TrainingStats(
args, args.disp_interval,
tblogger if args.use_tfboard and not args.no_save else None)
try:
logger.info('Training starts !')
step = args.start_step
for step in range(args.start_step, cfg.SOLVER.MAX_ITER):
# Warm up
if step < cfg.SOLVER.WARM_UP_ITERS:
method = cfg.SOLVER.WARM_UP_METHOD
if method == 'constant':
warmup_factor = cfg.SOLVER.WARM_UP_FACTOR
elif method == 'linear':
alpha = step / cfg.SOLVER.WARM_UP_ITERS
warmup_factor = cfg.SOLVER.WARM_UP_FACTOR * (1 -
alpha) + alpha
else:
raise KeyError(
'Unknown SOLVER.WARM_UP_METHOD: {}'.format(method))
lr_new = cfg.SOLVER.BASE_LR * warmup_factor
net_utils.update_learning_rate(optimizer, lr, lr_new)
lr = optimizer.param_groups[0]['lr']
assert lr == lr_new
elif step == cfg.SOLVER.WARM_UP_ITERS:
net_utils.update_learning_rate(optimizer, lr,
cfg.SOLVER.BASE_LR)
lr = optimizer.param_groups[0]['lr']
assert lr == cfg.SOLVER.BASE_LR
# Learning rate decay
if decay_steps_ind < len(cfg.SOLVER.STEPS) and \
step == cfg.SOLVER.STEPS[decay_steps_ind]:
logger.info('Decay the learning on step %d', step)
lr_new = lr * cfg.SOLVER.GAMMA
net_utils.update_learning_rate(optimizer, lr, lr_new)
lr = optimizer.param_groups[0]['lr']
assert lr == lr_new
decay_steps_ind += 1
training_stats.IterTic()
optimizer.zero_grad()
for inner_iter in range(args.iter_size):
try:
input_data = next(dataiterator)
except StopIteration:
dataiterator = iter(dataloader)
input_data = next(dataiterator)
for key in input_data:
if key != 'roidb': # roidb is a list of ndarrays with inconsistent length
input_data[key] = list(map(Variable, input_data[key]))
net_outputs = maskRCNN(**input_data)
training_stats.UpdateIterStats(net_outputs, inner_iter)
loss = net_outputs['total_loss']
loss.backward()
optimizer.step()
training_stats.IterToc()
training_stats.LogIterStats(step, lr)
if (step + 1) % CHECKPOINT_PERIOD == 0:
save_ckpt(output_dir, args, step, train_size, maskRCNN,
optimizer)
# ---- Training ends ----
# Save last checkpoint
save_ckpt(output_dir, args, step, train_size, maskRCNN, optimizer)
except (RuntimeError, KeyboardInterrupt):
del dataiterator
#logger.info('Save ckpt on exception ...')
#save_ckpt(output_dir, args, step, train_size, maskRCNN, optimizer)
#logger.info('Save ckpt done.')
stack_trace = traceback.format_exc()
print(stack_trace)
finally:
if args.use_tfboard and not args.no_save:
tblogger.close()
def main():
"""Main function"""
args = parse_args()
print('Called with args:')
print(args)
if not torch.cuda.is_available():
sys.exit("Need a CUDA device to run the code.")
if args.cuda or cfg.NUM_GPUS > 0:
cfg.CUDA = True
else:
raise ValueError("Need Cuda device to run !")
if args.dataset == "davis2017":
cfg.TRAIN.DATASETS = ('davis_train', )
#For davis, coco category is used.
cfg.MODEL.NUM_CLASSES = 81 #80 foreground + 1 background
else:
raise ValueError("Unexpected args.dataset: {}".format(args.dataset))
cfg_from_file(args.cfg_file)
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs)
if cfg.MODEL.IDENTITY_TRAINING and cfg.MODEL.IDENTITY_REPLACE_CLASS:
cfg.MODEL.NUM_CLASSES = 145
cfg.MODEL.IDENTITY_TRAINING = False
cfg.MODEL.ADD_UNKNOWN_CLASS = False
#Add unknow class type if necessary.
if cfg.MODEL.ADD_UNKNOWN_CLASS is True:
cfg.MODEL.NUM_CLASSES += 1
### Adaptively adjust some configs ###
original_batch_size = cfg.NUM_GPUS * cfg.TRAIN.IMS_PER_BATCH
original_ims_per_batch = cfg.TRAIN.IMS_PER_BATCH
original_num_gpus = cfg.NUM_GPUS
if args.batch_size is None:
args.batch_size = original_batch_size
cfg.NUM_GPUS = torch.cuda.device_count()
assert (args.batch_size % cfg.NUM_GPUS) == 0, \
'batch_size: %d, NUM_GPUS: %d' % (args.batch_size, cfg.NUM_GPUS)
cfg.TRAIN.IMS_PER_BATCH = args.batch_size // cfg.NUM_GPUS
effective_batch_size = args.iter_size * args.batch_size
print('effective_batch_size = batch_size * iter_size = %d * %d' %
(args.batch_size, args.iter_size))
print('Adaptive config changes:')
print(' effective_batch_size: %d --> %d' %
(original_batch_size, effective_batch_size))
print(' NUM_GPUS: %d --> %d' %
(original_num_gpus, cfg.NUM_GPUS))
print(' IMS_PER_BATCH: %d --> %d' %
(original_ims_per_batch, cfg.TRAIN.IMS_PER_BATCH))
### Adjust learning based on batch size change linearly
# For iter_size > 1, gradients are `accumulated`, so lr is scaled based
# on batch_size instead of effective_batch_size
old_base_lr = cfg.SOLVER.BASE_LR
cfg.SOLVER.BASE_LR *= args.batch_size / original_batch_size
cfg.SOLVER.BASE_LR *= 1.0 / cfg.MODEL.SEQUENCE_LENGTH
print(
'Adjust BASE_LR linearly according to batch_size change and sequence length change:\n'
' BASE_LR: {} --> {}'.format(old_base_lr, cfg.SOLVER.BASE_LR))
### Adjust solver steps
step_scale = original_batch_size / effective_batch_size
old_solver_steps = cfg.SOLVER.STEPS
old_max_iter = cfg.SOLVER.MAX_ITER
cfg.SOLVER.STEPS = list(
map(lambda x: int(x * step_scale + 0.5), cfg.SOLVER.STEPS))
cfg.SOLVER.MAX_ITER = int(cfg.SOLVER.MAX_ITER * step_scale + 0.5)
print(
'Adjust SOLVER.STEPS and SOLVER.MAX_ITER linearly based on effective_batch_size change:\n'
' SOLVER.STEPS: {} --> {}\n'
' SOLVER.MAX_ITER: {} --> {}'.format(old_solver_steps,
cfg.SOLVER.STEPS, old_max_iter,
cfg.SOLVER.MAX_ITER))
# Scale FPN rpn_proposals collect size (post_nms_topN) in `collect` function
# of `collect_and_distribute_fpn_rpn_proposals.py`
#
# post_nms_topN = int(cfg[cfg_key].RPN_POST_NMS_TOP_N * cfg.FPN.RPN_COLLECT_SCALE + 0.5)
if cfg.FPN.FPN_ON and cfg.MODEL.FASTER_RCNN:
cfg.FPN.RPN_COLLECT_SCALE = cfg.TRAIN.IMS_PER_BATCH / original_ims_per_batch
print(
'Scale FPN rpn_proposals collect size directly propotional to the change of IMS_PER_BATCH:\n'
' cfg.FPN.RPN_COLLECT_SCALE: {}'.format(
cfg.FPN.RPN_COLLECT_SCALE))
if args.num_workers is not None:
cfg.DATA_LOADER.NUM_THREADS = args.num_workers
print('Number of data loading threads: %d' % cfg.DATA_LOADER.NUM_THREADS)
### Overwrite some solver settings from command line arguments
if args.optimizer is not None:
cfg.SOLVER.TYPE = args.optimizer
if args.lr is not None:
cfg.SOLVER.BASE_LR = args.lr
if args.lr_decay_gamma is not None:
cfg.SOLVER.GAMMA = args.lr_decay_gamma
assert_and_infer_cfg()
timers = defaultdict(Timer)
### Dataset ###
timers['roidb'].tic()
merged_roidb, seq_num, seq_start_end = sequenced_roidb_for_training(
cfg.TRAIN.DATASETS, cfg.TRAIN.PROPOSAL_FILES, load_inv_db=True)
timers['roidb'].toc()
roidb_size = len(merged_roidb)
logger.info('{:d} roidbs sequences.'.format(roidb_size))
logger.info('Takes %.2f sec(s) to construct roidbs',
timers['roidb'].average_time)
# Effective training sample size for one epoch, number of sequences.
train_size = roidb_size // args.batch_size * args.batch_size
### Model ###
maskRCNN = vos_model_builder.Generalized_VOS_RCNN()
if cfg.CUDA:
maskRCNN.cuda()
### Optimizer ###
gn_param_nameset = set()
for name, module in maskRCNN.named_modules():
if isinstance(module, nn.GroupNorm):
gn_param_nameset.add(name + '.weight')
gn_param_nameset.add(name + '.bias')
gn_params = []
gn_param_names = []
bias_params = []
bias_param_names = []
nonbias_params = []
nonbias_param_names = []
nograd_param_names = []
for key, value in maskRCNN.named_parameters():
if value.requires_grad:
if 'bias' in key:
bias_params.append(value)
bias_param_names.append(key)
elif key in gn_param_nameset:
gn_params.append(value)
gn_param_names.append(key)
else:
nonbias_params.append(value)
nonbias_param_names.append(key)
else:
nograd_param_names.append(key)
assert (gn_param_nameset - set(nograd_param_names) -
set(bias_param_names)) == set(gn_param_names)
# Learning rate of 0 is a dummy value to be set properly at the start of training
params = [{
'params': nonbias_params,
'lr': 0,
'weight_decay': cfg.SOLVER.WEIGHT_DECAY
}, {
'params':
bias_params,
'lr':
0 * (cfg.SOLVER.BIAS_DOUBLE_LR + 1),
'weight_decay':
cfg.SOLVER.WEIGHT_DECAY if cfg.SOLVER.BIAS_WEIGHT_DECAY else 0
}, {
'params': gn_params,
'lr': 0,
'weight_decay': cfg.SOLVER.WEIGHT_DECAY_GN
}]
# names of paramerters for each paramter
param_names = [nonbias_param_names, bias_param_names, gn_param_names]
if cfg.SOLVER.TYPE == "SGD":
optimizer = torch.optim.SGD(params, momentum=cfg.SOLVER.MOMENTUM)
elif cfg.SOLVER.TYPE == "Adam":
optimizer = torch.optim.Adam(params)
### Load checkpoint
if args.load_ckpt:
load_name = args.load_ckpt
logging.info("loading checkpoint %s", load_name)
checkpoint = torch.load(load_name,
map_location=lambda storage, loc: storage)
net_utils.load_ckpt_no_mapping(maskRCNN, checkpoint['model'])
if args.resume:
args.start_step = checkpoint['step'] + 1
if 'train_size' in checkpoint: # For backward compatibility
if checkpoint['train_size'] != train_size:
print(
'train_size value: %d different from the one in checkpoint: %d'
% (train_size, checkpoint['train_size']))
# reorder the params in optimizer checkpoint's params_groups if needed
# misc_utils.ensure_optimizer_ckpt_params_order(param_names, checkpoint)
# There is a bug in optimizer.load_state_dict on Pytorch 0.3.1.
# However it's fixed on master.
optimizer.load_state_dict(checkpoint['optimizer'])
# misc_utils.load_optimizer_state_dict(optimizer, checkpoint['optimizer'])
del checkpoint
torch.cuda.empty_cache()
if args.load_detectron: #TODO resume for detectron weights (load sgd momentum values)
logging.info("loading Detectron weights %s", args.load_detectron)
load_detectron_weight(maskRCNN,
args.load_detectron,
force_load_all=False)
lr = optimizer.param_groups[0][
'lr'] # lr of non-bias parameters, for commmand line outputs.
maskRCNN = mynn.DataParallel(maskRCNN,
cpu_keywords=['im_info', 'roidb'],
minibatch=True)
### Training Setups ###
args.run_name = misc_utils.get_run_name() + '_step'
output_dir = misc_utils.get_output_dir(args, args.run_name)
args.cfg_filename = os.path.basename(args.cfg_file)
if not args.no_save:
if not os.path.exists(output_dir):
os.makedirs(output_dir)
blob = {'cfg': yaml.dump(cfg), 'args': args}
with open(os.path.join(output_dir, 'config_and_args.pkl'), 'wb') as f:
pickle.dump(blob, f, pickle.HIGHEST_PROTOCOL)
if args.use_tfboard:
from tensorboardX import SummaryWriter
# Set the Tensorboard logger
tblogger = SummaryWriter(output_dir)
### Training Loop ###
maskRCNN.train()
CHECKPOINT_PERIOD = int(cfg.TRAIN.SNAPSHOT_ITERS / cfg.NUM_GPUS)
# Set index for decay steps
decay_steps_ind = None
for i in range(1, len(cfg.SOLVER.STEPS)):
if cfg.SOLVER.STEPS[i] >= args.start_step:
decay_steps_ind = i
break
if decay_steps_ind is None:
decay_steps_ind = len(cfg.SOLVER.STEPS)
training_stats = TrainingStats(
args, args.disp_interval,
tblogger if args.use_tfboard and not args.no_save else None)
dataloader, dataiterator, warmup_length = gen_sequence_data_sampler(
merged_roidb, seq_num, seq_start_end, use_seq_warmup=False)
try:
logger.info('Training starts !')
step = args.start_step
for step in range(args.start_step, cfg.SOLVER.MAX_ITER):
# Warm up
if step < cfg.SOLVER.WARM_UP_ITERS:
method = cfg.SOLVER.WARM_UP_METHOD
if method == 'constant':
warmup_factor = cfg.SOLVER.WARM_UP_FACTOR
elif method == 'linear':
alpha = step / cfg.SOLVER.WARM_UP_ITERS
warmup_factor = cfg.SOLVER.WARM_UP_FACTOR * (1 -
alpha) + alpha
else:
raise KeyError(
'Unknown SOLVER.WARM_UP_METHOD: {}'.format(method))
lr_new = cfg.SOLVER.BASE_LR * warmup_factor
net_utils.update_learning_rate(optimizer, lr, lr_new)
lr = optimizer.param_groups[0]['lr']
assert lr == lr_new
elif step == cfg.SOLVER.WARM_UP_ITERS:
net_utils.update_learning_rate(optimizer, lr,
cfg.SOLVER.BASE_LR)
lr = optimizer.param_groups[0]['lr']
assert lr == cfg.SOLVER.BASE_LR
# Learning rate decay
if decay_steps_ind < len(cfg.SOLVER.STEPS) and \
step == cfg.SOLVER.STEPS[decay_steps_ind]:
logger.info('Decay the learning on step %d', step)
lr_new = lr * cfg.SOLVER.GAMMA
net_utils.update_learning_rate(optimizer, lr, lr_new)
lr = optimizer.param_groups[0]['lr']
assert lr == lr_new
decay_steps_ind += 1
training_stats.IterTic()
optimizer.zero_grad()
if cfg.TRAIN.USE_SEQ_WARMUP is True and step >= cfg.TRAIN.ITER_BEFORE_USE_SEQ_WARMUP and (
(step - cfg.TRAIN.ITER_BEFORE_USE_SEQ_WARMUP) %
cfg.TRAIN.WARMUP_LENGTH_CHANGE_STEP) == 0:
#TODO better dataiterator creator.
raise NotImplementedError('not able to delete.')
dataloader, dataiterator, warmup_length = gen_sequence_data_sampler(
merged_roidb, seq_num, seq_start_end, use_seq_warmup=True)
print('update warmup length:', warmup_length)
try:
input_data_sequence = next(dataiterator)
except StopIteration:
dataiterator = iter(dataloader)
input_data_sequence = next(dataiterator)
# clean hidden states before training.
maskRCNN.module.clean_hidden_states()
maskRCNN.module.clean_flow_features()
assert len(input_data_sequence['data']
) == cfg.MODEL.SEQUENCE_LENGTH + warmup_length, print(
len(input_data_sequence['data']), '!=',
cfg.MODEL.SEQUENCE_LENGTH + warmup_length)
# if train_part == 0: train backbone.
# else train_part == 1: train heads.
train_part = 1
if cfg.TRAIN.ALTERNATE_TRAINING:
if step % 2 == 0:
maskRCNN.module.freeze_conv_body_only()
train_part = 1
else:
maskRCNN.module.train_conv_body_only()
train_part = 0
# this is used for longer sequence training.
# when reach maximum trainable length, detach the hidden states.
cnter_for_detach_hidden_states = 0
for inner_iter in range(cfg.MODEL.SEQUENCE_LENGTH + warmup_length):
#get input_data
input_data = {}
for key in input_data_sequence.keys():
input_data[key] = input_data_sequence[key][
inner_iter:inner_iter + 1]
for key in input_data:
if key != 'roidb' and key != 'data_flow': # roidb is a list of ndarrays with inconsistent length
input_data[key] = list(map(Variable, input_data[key]))
if key == 'data_flow':
if inner_iter != 0 and input_data[key][0][0][
0] is not None: # flow is not None.
input_data[key] = [
Variable(
torch.tensor(
np.expand_dims(
np.squeeze(
np.array(input_data[key],
dtype=np.float32)),
0),
device=input_data['data'][0].device))
]
assert input_data['data'][0].shape[
-2:] == input_data[key][0].shape[
-2:], "Spatial shape of image and flow are not equal."
else:
input_data[key] = [None]
if cfg.TRAIN.USE_SEQ_WARMUP and inner_iter < warmup_length:
maskRCNN.module.set_stop_after_hidden_states(stop=True)
net_outputs = maskRCNN(**input_data)
assert net_outputs is None
maskRCNN.module.detach_hidden_states()
maskRCNN.module.set_stop_after_hidden_states(stop=False)
continue
net_outputs = maskRCNN(**input_data)
training_stats.UpdateIterStats(net_outputs, inner_iter)
loss = net_outputs['total_loss']
if train_part == 0 or cnter_for_detach_hidden_states >= cfg.TRAIN.MAX_TRAINABLE_SEQ_LENGTH or inner_iter == cfg.MODEL.SEQUENCE_LENGTH + warmup_length - 1:
# if reach the max trainable length or end of the sequence. free the graph and detach the hidden states.
loss.backward()
maskRCNN.module.detach_hidden_states()
cnter_for_detach_hidden_states = 0
else:
loss.backward(retain_graph=True)
cnter_for_detach_hidden_states += 1
#TODO step every time?
if cnter_for_detach_hidden_states == 0:
optimizer.step()
optimizer.zero_grad()
training_stats.IterToc()
training_stats.LogIterStats(step, lr)
if (step + 1) % CHECKPOINT_PERIOD == 0:
save_ckpt(output_dir, args, step, train_size, maskRCNN,
optimizer)
# ---- Training ends ----
# Save last checkpoint
save_ckpt(output_dir, args, step, train_size, maskRCNN, optimizer)
except (RuntimeError):
del dataiterator
logger.info('Save ckpt on exception ...')
save_ckpt(output_dir, args, step, train_size, maskRCNN, optimizer)
logger.info('Save ckpt done.')
stack_trace = traceback.format_exc()
print(stack_trace)
except (KeyboardInterrupt):
del dataiterator
logger.info('Save ckpt on Keyboard Interrupt ...')
save_ckpt(output_dir, args, step, train_size, maskRCNN, optimizer)
logger.info('Save ckpt done.')
stack_trace = traceback.format_exc()
print(stack_trace)
finally:
if args.use_tfboard and not args.no_save:
tblogger.close()