Skip to content

TDM3.0 beam search采样不会很耗时吗 #356

Open
@akafen

Description

@akafen

TDM3.0训练时是采用beam search得到样本,然后根据样本的子节点max得到该节点的拟合值,但是一个样本假设耗时0.3s,一个batch有2048个样本,那一个batch的耗时不就将近600s?这不会超级耗时吗

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions