Skip to content

[3/5] refactor(attention): add provider backends and explicit selection - #1719

Draft
sufubao wants to merge 1 commit into
refactor/attention-02-kernelsfrom
refactor/attention-03-backends
Draft

sufubao wants to merge 1 commit into
refactor/attention-02-kernelsfrom
refactor/attention-03-backends

Conversation

@sufubao

@sufubao sufubao commented Nov 12, 2026 •

Copy link
Copy Markdown
Collaborator

新增dense、MLA、sparse MLA、GDN/KDA native/replay provider实现和显式selector/probe。backend保存组配置,context保存批次metadata,MTP cursor由层参数持有。旧LLM路由继续使用原wrapper,新接口由独立provider测试执行。

验证:H200新provider数值/契约、跨State cursor Graph与原attention兼容性215项通过;后续第4步包含此层并完成真实模型MTP/Graph缓存恢复验证。

由已关闭的 #1700 拆分,本PR为第3/5步。依赖顺序:

  1. [1/5] refactor(attention): define execution contracts and device workspace #1717 — API与设备/工作区
  2. [2/5] refactor(attention): relocate kernels without changing execution #1718 — kernel目录迁移
  3. [3/5] refactor(attention): add provider backends and explicit selection #1719 — provider backend与选择
  4. [4/5] refactor(attention): use varlen operators for vision and audio #1720 — 视觉/音频Varlen
  5. [5/5] refactor(attention): integrate model bindings and state lifecycle #1721 — 模型/State/Graph运行时

当前base为#1718 的分支,Files changed只展示本步增量。请按序合入;前序合入main后,将后续PR的base依次改为main并更新分支。所有步骤保持Draft。

每层已运行仓库固定版本Black/Flake8。仓库CI仅监听base=main,因此第2–5步当前以本地lint和实验记录为准,retarget后再检查CI。实验记录位于/data6/exp_logs/runs。

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant