Closed pdoublerainbow closed 2 years ago
最近我用fast-reid实现了Deep Mutual Learning,但是代码遇到了一点问题。 首先,我先build了多个model; 然后就在train loop里面实现了loss计算过程; 但是结果发现不同模型forward多次之后,在梯度的反传会报错;如果多个模型分别forward一次就反传一次就没有问题。 报错如下: 工程是直接放在projects目录下面的 FastDML.zip 希望廖老师可以帮忙解决一下,谢谢!
This issue is stale because it has been open for 30 days with no activity.
This issue was closed because it has been inactive for 14 days since being marked as stale.
最近我用fast-reid实现了Deep Mutual Learning,但是代码遇到了一点问题。 首先,我先build了多个model; 然后就在train loop里面实现了loss计算过程; 但是结果发现不同模型forward多次之后,在梯度的反传会报错;如果多个模型分别forward一次就反传一次就没有问题。 报错如下: 工程是直接放在projects目录下面的 FastDML.zip 希望廖老师可以帮忙解决一下,谢谢!