def test_model(model_file, multi_gpu_testing, opts=None):
"""Test a model."""
# Clear memory before inference
workspace.ResetWorkspace()
# Run inference
run_inference(
model_file, multi_gpu_testing=multi_gpu_testing,
check_expected_results=True,
)
def main(ind_range=None, multi_gpu_testing=False):
all_results = run_inference(ind_range=ind_range,
multi_gpu_testing=multi_gpu_testing)
if not ind_range:
task_evaluation.check_expected_results(all_results,
atol=cfg.EXPECTED_RESULTS_ATOL,
rtol=cfg.EXPECTED_RESULTS_RTOL)
task_evaluation.log_copy_paste_friendly_results(all_results)
def main(ind_range=None, multi_gpu_testing=False):
all_results = run_inference(
ind_range=ind_range, multi_gpu_testing=multi_gpu_testing
)
if not ind_range:
task_evaluation.check_expected_results(
all_results,
atol=cfg.EXPECTED_RESULTS_ATOL,
rtol=cfg.EXPECTED_RESULTS_RTOL
)
task_evaluation.log_copy_paste_friendly_results(all_results)
def test_net_routine(args):
if not torch.cuda.is_available():
sys.exit("Need a CUDA device to run the code.")
logger = utils.logging.setup_logging(__name__)
logger.info('Called with args:')
logger.info(args)
assert (torch.cuda.device_count() == 1) ^ bool(args.multi_gpu_testing)
assert bool(args.load_ckpt) ^ bool(args.load_detectron), \
'Exactly one of --load_ckpt and --load_detectron should be specified.'
if args.output_dir is None:
ckpt_path = args.load_ckpt if args.load_ckpt else args.load_detectron
args.output_dir = os.path.join(
os.path.dirname(os.path.dirname(ckpt_path)), 'test')
logger.info('Automatically set output directory to %s', args.output_dir)
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
if args.cfg_file is not None:
merge_cfg_from_file(args.cfg_file)
if args.set_cfgs is not None:
merge_cfg_from_list(args.set_cfgs)
assert_and_infer_cfg()
logger.info('Testing with config:')
logger.info(pprint.pformat(cfg))
# For test_engine.multi_gpu_test_net_on_dataset
args.test_net_file, _ = os.path.splitext(__file__)
# manually set args.cuda
args.cuda = True
run_inference(
args,
ind_range=args.range,
multi_gpu_testing=args.multi_gpu_testing,
check_expected_results=True)
def main(ind_range=None, multi_gpu_testing=False):
output_dir = get_output_dir(training=False)
all_results = run_inference(output_dir,
ind_range=ind_range,
multi_gpu_testing=multi_gpu_testing)
if not ind_range:
task_evaluation.check_expected_results(all_results,
atol=cfg.EXPECTED_RESULTS_ATOL,
rtol=cfg.EXPECTED_RESULTS_RTOL)
import json
json.dump(all_results,
open(os.path.join(output_dir, 'bbox_results_all.json'), 'w'))
task_evaluation.log_copy_paste_friendly_results(all_results)
def main(args):
workspace.GlobalInit(['caffe2', '--caffe2_log_level=2'])
logger = utils.logging.setup_logging(__name__)
logger.info('Called with args:')
logger.info(args)
if args.cfg_file is not None:
merge_cfg_from_file(args.cfg_file)
if args.opts is not None:
merge_cfg_from_list(args.opts)
assert_and_infer_cfg(make_immutable=False) # Necessary but not recommended
logger.info('Testing with config:')
logger.info(pprint.pformat(cfg))
while not os.path.exists(cfg.TEST.WEIGHTS) and args.wait:
logger.info('Waiting for \'{}\' to exist...'.format(cfg.TEST.WEIGHTS))
time.sleep(10)
all_results = run_inference(
cfg.TEST.WEIGHTS,
ind_range=args.range,
multi_gpu_testing=args.multi_gpu_testing,
check_expected_results=True,
)
return all_results
cfg.TEST.DATASETS = ('coco_2017_val',)
cfg.MODEL.NUM_CLASSES = 81
elif args.dataset == "keypoints_coco2017":
cfg.TEST.DATASETS = ('keypoints_coco_2017_val',)
cfg.MODEL.NUM_CLASSES = 2
elif args.dataset == "pascal_voc":
cfg.TEST.DATASETS = ('voc_2007_test',)
cfg.MODEL.NUM_CLASSES = 21
elif args.dataset == "pascal_voc_0712":
cfg.TEST.DATASETS = ('voc_2007_test',)
cfg.MODEL.NUM_CLASSES = 21
elif args.dataset.startswith("vg"):
cfg.TEST.DATASETS = ('%s_test' %args.dataset,)
else: # For subprocess call
assert cfg.TEST.DATASETS, 'cfg.TEST.DATASETS shouldn\'t be empty'
assert_and_infer_cfg()
logger.info('Testing with config:')
logger.info(pprint.pformat(cfg))
# For test_engine.multi_gpu_test_net_on_dataset
args.test_net_file, _ = os.path.splitext(__file__)
# manually set args.cuda
args.cuda = True
run_inference(
args,
ind_range=args.range,
multi_gpu_testing=args.multi_gpu_testing,
check_expected_results=True)
logger.info('Waiting for \'{}\' to exist...'.format(cfg.TEST.WEIGHTS))
time.sleep(10)
l = cfg.TEST.WEIGHTS
lst = l.strip().split("/")
modelPath = "/".join(lst[:len(lst) - 1]) + "/"
print(modelPath)
files = [
x for x in os.listdir(modelPath)
if os.path.isfile(modelPath + x) and x.endswith('.pkl')
]
modelNumLst = []
for l in files:
ll = l[10:len(l) - 4]
#print ( ll )
modelNumLst.append(int(ll))
modelNumLst.sort()
for ll in modelNumLst:
l = "model_iter" + str(ll) + ".pkl"
run_inference(
modelPath + l,
ind_range=args.range,
multi_gpu_testing=args.multi_gpu_testing,
check_expected_results=True,
)
test_result(l, "testzj.log", cfg.TEST.DATASETS)
#os.remove( "testzj.log" )
cfg.VIS = args.vis
if args.cfg_file is not None:
merge_cfg_from_file(args.cfg_file)
if args.set_cfgs is not None:
merge_cfg_from_list(args.set_cfgs)
if args.dataset == "coco2017":
cfg.TEST.DATASETS = ('coco_2017_val',)
cfg.MODEL.NUM_CLASSES = 81
elif args.dataset == "keypoints_coco2017":
cfg.TEST.DATASETS = ('keypoints_coco_2017_val',)
cfg.MODEL.NUM_CLASSES = 2
else: # For subprocess call
assert cfg.TEST.DATASETS, 'cfg.TEST.DATASETS shouldn\'t be empty'
assert_and_infer_cfg()
logger.info('Testing with config:')
logger.info(pprint.pformat(cfg))
# For test_engine.multi_gpu_test_net_on_dataset
args.test_net_file, _ = os.path.splitext(__file__)
# manually set args.cuda
args.cuda = True
run_inference(
args,
ind_range=args.range,
multi_gpu_testing=args.multi_gpu_testing,
check_expected_results=True)
if len(sys.argv) == 1:
parser.print_help()
sys.exit(1)
return parser.parse_args()
if __name__ == '__main__':
workspace.GlobalInit(['caffe2', '--caffe2_log_level=0'])
logger = utils.logging.setup_logging(__name__)
args = parse_args()
logger.info('Called with args:')
logger.info(args)
if args.cfg_file is not None:
merge_cfg_from_file(args.cfg_file)
if args.opts is not None:
merge_cfg_from_list(args.opts)
assert_and_infer_cfg()
logger.info('Testing with config:')
logger.info(pprint.pformat(cfg))
while not os.path.exists(cfg.TEST.WEIGHTS) and args.wait:
logger.info('Waiting for \'{}\' to exist...'.format(cfg.TEST.WEIGHTS))
time.sleep(10)
run_inference(
cfg.TEST.WEIGHTS,
ind_range=args.range,
multi_gpu_testing=args.multi_gpu_testing,
check_expected_results=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 == "custom_dataset" and args.num_classes is None:
raise ValueError(
"Need number of classes in your custom dataset to run!")
if args.dataset == "coco2017":
cfg.TRAIN.DATASETS = ('coco_2017_train', )
cfg.TEST.DATASETS = ('coco_2017_val', )
cfg.MODEL.NUM_CLASSES = 81
elif args.dataset == "keypoints_coco2017":
cfg.TRAIN.DATASETS = ('keypoints_coco_2017_train', )
cfg.TEST.DATASETS = ('keypoints_coco_2017_val', )
cfg.MODEL.NUM_CLASSES = 2
elif args.dataset == "voc2007":
cfg.TRAIN.DATASETS = ('voc_2007_train', )
cfg.TEST.DATASETS = ('voc_2007_test', )
cfg.MODEL.NUM_CLASSES = 21
elif args.dataset == "voc2012":
cfg.TRAIN.DATASETS = ('voc_2012_train', )
cfg.MODEL.NUM_CLASSES = 21
elif args.dataset == "custom_dataset":
cfg.TRAIN.DATASETS = ('custom_data_train', )
cfg.TEST.DATASETS = ('custom_data_trainval', )
cfg.MODEL.NUM_CLASSES = args.num_classes
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)
cfg.VIS = args.vis
# For test_engine.multi_gpu_test_net_on_dataset
args.test_net_file, _ = os.path.splitext(__file__)
# manually set args.cuda
### 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 = Generalized_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(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)
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)
args.output_dir = output_dir
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)
maskRCNN.module.eval()
results = run_inference(
args,
ind_range=args.range,
multi_gpu_testing=args.multi_gpu_testing,
check_expected_results=True,
model=maskRCNN)
maskRCNN.module.train()
training_stats.UpdateValStats(results, epoch=step)
# ---- 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()
