Skip to content

About the efficiency of CAM implementation #55

@qqtang-code

Description

@qqtang-code
for token_index in range(
                start_budget + padding_length + recent_budget, seq_length
            ):

When merging with CAM, if seq_length is long, such as 8k, and start_budget_ratio=0.1, then the for loop will be very long and inefficient.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions