Skip to content
This repository has been archived by the owner on Jan 24, 2024. It is now read-only.

The oneDNN segment fault error #230

Open
2742195759 opened this issue Jul 3, 2023 · 0 comments
Open

The oneDNN segment fault error #230

2742195759 opened this issue Jul 3, 2023 · 0 comments

Comments

@2742195759
Copy link
Collaborator

在单测验证场景中,发现在动转静的单测:test_resnet_v2.py 中,如果将

    def forward(self, inputs):
        y = self.conv0(inputs)
        conv1 = self.conv1(y)
        conv2 = self.conv2(conv1)

        if self.shortcut:
            short = inputs
        else:
            short = self.short(inputs)

        y = paddle.add(x=short, y=conv2)

        # TODO: uncomment this lines to reproduce the oneDNN segment fault error.
        #layer_helper = paddle.fluid.layer_helper.LayerHelper(
            #self.full_name(), act='relu'
        #)
        #return layer_helper.append_activation(y)
        return paddle.nn.functional.relu(y)

上述的 relu 改为 append op 的情况,那么会出现因为子图个数过多,而导致的 one dnn segment fault 的问题。这个问题目前是通过移出 legacy api 调用,使得子图个数为1规避了。后续还是需要基础查明这个问题的,应该是ProgramOp的一个bug。

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant