Skip to content

MiniGPT-5 inference error #6

@hlchen23

Description

@hlchen23

When generating image tokens, the attention mask can not be correctly generated.

the image has 8 tokens, but the attention mask only increases by 1, causing a torch size error.

When I run task-2 evaluation (text seq to image seq), it has error

ValueError('Attention mask should be of size (1, 1, 1, 380), but is torch.Size([1, 1, 1, 373])')

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