flash mla kernal是否支持同一batch下不同query的动态token数？ #190

Open

opened

on Jun 17, 2026

q: (batch_size, seq_len_q, num_heads_q, head_dim)
flash mla算子的输入这个seq_len_q是定死的，那么是否不支持不同的seq_len_q进入kernal计算？

to join this conversation on GitHub. Already have an account? Sign in to comment

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