def main():
args = parse_args()
if args.amp:
import os
os.environ['AMP'] = "1"
cfg = get_config(args.config, overrides=args.override)
_, world_size = get_dist_info()
parallel = world_size != 1
if parallel:
paddle.distributed.init_parallel_env()
if args.test:
test_model(cfg, weights=args.weights, parallel=parallel)
elif args.train_dali:
train_dali(cfg, weights=args.weights, parallel=parallel)
elif args.multigrid:
train_model_multigrid(cfg, world_size, validate=args.validate)
else:
train_model(cfg,
weights=args.weights,
parallel=parallel,
validate=args.validate,
fleet=args.fleet,
amp=args.amp)
def __init__(self,
data_size,
batch_size,
world_size,
log_interval=1,
**kwargs):
self.data_size = data_size
self.batch_size = batch_size
_, self.world_size = get_dist_info()
self.log_interval = log_interval
def get_acc(self, scores, labels, valid_mode):
top1 = paddle.metric.accuracy(input=scores, label=labels, k=1)
top5 = paddle.metric.accuracy(input=scores, label=labels, k=5)
_, world_size = get_dist_info()
#NOTE(shipping): deal with multi cards validate
if world_size > 1 and valid_mode: #reduce sum when valid
top1 = paddle.distributed.all_reduce(
top1, op=paddle.distributed.ReduceOp.SUM) / world_size
top5 = paddle.distributed.all_reduce(
top5, op=paddle.distributed.ReduceOp.SUM) / world_size
return top1, top5
def main():
args = parse_args()
cfg = get_config(args.config, overrides=args.override)
dataset = build_dataset((cfg.DATASET.test, cfg.PIPELINE.test))
_, world_size = get_dist_info()
parallel = world_size != 1
if parallel:
paddle.distributed.init_parallel_env()
model = build_model(cfg.MODEL)
test_model(model, dataset, cfg, args.weights, world_size)
def loss(self, scores, labels, reduce_sum=False, **kwargs):
"""Calculate the loss accroding to the model output ```scores```,
and the target ```labels```.
Args:
scores (paddle.Tensor): The output of the model.
labels (paddle.Tensor): The target output of the model.
Returns:
losses (dict): A dict containing field 'loss'(mandatory) and 'top1_acc', 'top5_acc'(optional).
"""
if len(labels) == 1:
labels = labels[0]
elif len(labels) == 3:
labels_a, labels_b, lam = labels
return self.mixup_loss(scores, labels_a, labels_b, lam)
else:
raise NotImplemented
if self.ls_eps != 0.:
labels = F.one_hot(labels, self.num_classes)
labels = F.label_smooth(labels, epsilon=self.ls_eps)
# reshape [bs, 1, num_classes] to [bs, num_classes]
#NOTE: maybe squeeze is helpful for understanding.
labels = paddle.reshape(labels, shape=[-1, self.num_classes])
#labels.stop_gradient = True #XXX(shipping): check necessary
losses = dict()
#NOTE(shipping): F.crossentropy include logsoftmax and nllloss !
#NOTE(shipping): check the performance of F.crossentropy
loss = self.loss_func(scores, labels, **kwargs)
avg_loss = paddle.mean(loss)
top1 = paddle.metric.accuracy(input=scores, label=labels, k=1)
top5 = paddle.metric.accuracy(input=scores, label=labels, k=5)
_, world_size = get_dist_info()
#NOTE(shipping): deal with multi cards validate
if world_size > 1 and reduce_sum:
top1 = paddle.distributed.all_reduce(
top1, op=paddle.distributed.ReduceOp.SUM) / world_size
top5 = paddle.distributed.all_reduce(
top5, op=paddle.distributed.ReduceOp.SUM) / world_size
losses['top1'] = top1
losses['top5'] = top5
losses['loss'] = avg_loss
return losses
def main():
args = parse_args()
cfg = get_config(args.config, overrides=args.override)
_, world_size = get_dist_info()
parallel = world_size != 1
if parallel:
paddle.distributed.init_parallel_env()
if args.test:
test_model(cfg, weights=args.weights, parallel=parallel)
else:
train_model(cfg,
weights=args.weights,
parallel=parallel,
validate=args.validate)