one of the variables needed for gradient computation has been modified by an inplace operation

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation

Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True)

使用torch在训练模型时候出现了这个问题, 按照提示使用
with torch.autograd.set_detect_anomaly(True) 来查找哪里出错.
在问题出现的函数中运行,例如是在运行forward是出错,可以按照下面的形式定位,其他情况也类似.

def forward(self, x):
	with torch.autograd.set_detect_anomaly(True) :
			'''
			代码
			'''

运行结果需要在一步分析, 运行结果如下所示:

RuntimeWarning: Traceback of forward call that caused the error:
'''
错误信息	(这里可以进一步发现inplace问题发生的位置, 和正常的debug定位一样)
'''
Traceback (most recent call last):
'''
错误信息
'''

你可能感兴趣的:(pytorch)