MedicalSeg的问题

[lusichao]

trafficstars

训练的时候出现错误： Traceback (most recent call last): File "train.py", line 194, in main(args) File "train.py", line 173, in main train( File "/root/PaddleSeg/contrib/MedicalSeg/medicalseg/core/train.py", line 132, in train logits_list = model(images) File "/root/anaconda3/lib/python3.8/site-packages/paddle/fluid/dygraph/layers.py", line 930, in call return self._dygraph_call_func(*inputs, **kwargs) File "/root/anaconda3/lib/python3.8/site-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_c outputs = self.forward(*inputs, **kwargs) File "/root/PaddleSeg/contrib/MedicalSeg/medicalseg/models/vnet.py", line 257, in forward out16 = self.in_tr(x) # dropout cause a lot align problem File "/root/anaconda3/lib/python3.8/site-packages/paddle/fluid/dygraph/layers.py", line 930, in call return self._dygraph_call_func(*inputs, **kwargs) File "/root/anaconda3/lib/python3.8/site-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_c outputs = self.forward(*inputs, **kwargs) File "/root/PaddleSeg/contrib/MedicalSeg/medicalseg/models/vnet.py", line 77, in forward out = self.bn1(out) File "/root/anaconda3/lib/python3.8/site-packages/paddle/fluid/dygraph/layers.py", line 930, in call return self._dygraph_call_func(*inputs, **kwargs) File "/root/anaconda3/lib/python3.8/site-packages/paddle/fluid/dygraph/layers.py", line 915, in _dygraph_c outputs = self.forward(*inputs, **kwargs) File "/root/anaconda3/lib/python3.8/site-packages/paddle/nn/layer/norm.py", line 1095, in forward sync_batch_norm_out, _, _, _, _, _ = _C_ops.sync_batch_norm( AttributeError: module 'paddle.fluid.core_avx.ops' has no attribute 'sync_batch_norm'

找了下_C_ops.py文件，没有sync_batch_norm的数据，

PaddlePaddle版本：（PaddlePaddle 2.3） 操作系统信息：（Ubuntu 18） Python版本号：（Python3.8） CUDA版本：（ CUDA11.4）

[shiyutang]

I think this is the problem with the paddle version, try switching to paddle==2.2. Btw, did you switch the bn1 from BatchNorm3D to SyncBatchNorm? The error is located within SyncBatchNorm.If you did, I think changing back to BatchNorm3D will help as well.

But it is a bit odd that I cannot reproduce your error with different paddle versions or different BN. Maybe you can put forward an issue to Paddle as well.

[XuyangPan]

解决了吗？我也报了同样的错误，怎么解决的

[lusichao]

解决了吗？我也报了同样的错误，怎么解决的

没解决，后来一直没再用。

[shiyutang]

建议安装paddlepaddle2.4后重新尝试

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

[shiyutang]

_C_ops为paddle内部库，在paddleseg中bn1也就是BatchNorm3D自动转为SyncBN，从而forward计算落入syncbn函数中，但此处的报错说明syncbn对应的_C_ops函数不存在，说明使用的paddle版本中没有SyncBN的定义，可以1. 取消转换syncbn，或者2. 切换有syncbn的paddle版本。

转换syncbn的代码位置： https://github.com/PaddlePaddle/PaddleSeg/blob/19351bab9a824a8f96e1c1b527ec2d7db21309c9/contrib/MedicalSeg/medicalseg/cvlibs/config.py#L322