报错:Default process group is not initialized
报错代码,yolov5训练时不报错, 参数配置错了,
with torch_distributed_zero_first(rank):
dataset = LoadImagesAndLabels(path, imgsz, batch_size, augment=augment, # augment images
hyp=hyp, # augmentation hyperparameters
rect=rect, # rectangular training
cache_images=cache, single_cls=opt.single_cls, stride=int(stride), pad=pad, image_weights=image_weights, prefix=prefix,data_type=data_type)
@contextmanager
def torch_distributed_zero_first(local_rank: int):
"""
Decorator to make all processes in distributed training wait for each local_master to do something.
"""
if local_rank not in [-1, 0]:
torch.distributed.barrier()
yield
if local_rank == 0:
版权说明 : 本文为转载文章, 版权归原作者所有 版权申明
原文链接 : https://blog.csdn.net/jacke121/article/details/125476898
内容来源于网络,如有侵权,请联系作者删除!