STDC-Seg icon indicating copy to clipboard operation
STDC-Seg copied to clipboard

在训练时出现了TypeError: __init__() got an unexpected keyword argument 'find_unused_parameters‘

Open ding-Chen opened this issue 3 years ago • 1 comments

我在单GPU环境下执行: export CUDA_VISIBLE_DEVICES=0 python -m torch.distributed.launch
--nproc_per_node=1 train.py
--respath checkpoints/train_STDC1-Seg/
--backbone STDCNet813
--mode train
--n_workers_train 12
--n_workers_val 1
--max_iter 60000
--use_boundary_8 True
--pretrain_path checkpoints/STDCNet813M_73.91.tar 时

出现了以下错误:

Traceback (most recent call last): File "train.py", line 425, in train() File "train.py", line 220, in train find_unused_parameters=True TypeError: init() got an unexpected keyword argument 'find_unused_parameters’

请问大家有遇到过这个问题吗

ding-Chen avatar Aug 04 '21 08:08 ding-Chen

我也遇到了这个问题,好像是融合不同stride的边界标签的问题,只使用单独stride的边界标签进行训练就没问题,不知道作者是什么处理的

ydhongHIT avatar Nov 13 '21 15:11 ydhongHIT