Skip to content

[TRITON] Fix Assert in Triton fused_kv_cache#3601

Open
leonling-ll wants to merge 1 commit into
mainfrom
liyang/fix_fused_kv_cache_assert
Open

[TRITON] Fix Assert in Triton fused_kv_cache#3601
leonling-ll wants to merge 1 commit into
mainfrom
liyang/fix_fused_kv_cache_assert

Conversation

@leonling-ll
Copy link
Copy Markdown

Assertion in triton kernel fused_qk_rope_reshape_and_cache: t == tk == tv and t_slot <= tk would be inconsistant the logic in _fused_qk_rope_reshape_and_cache_kernel.
In which pid could be greater than TxQH, while the assertion assume the launch grid is smaller or equal to TxQH (n_pid = t * qh + (t_slot - t) * kh)

This makes CI failing: https://github.com/ROCm/ATOM/actions/runs/27060334232/job/79872120485#step:22:537

This PR aims to fix it and fix some typo.

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Jun 8, 2026

🏷️ CI Guide

Runs automatically on every PR:

  • ✅ Pre-checks (submodule verification, code formatting)
  • ✅ Aiter op tests (gfx942 + gfx950)
  • ✅ Triton tests on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label Tests
ci:triton-300x Run an additional Triton test job on MI300X in PRs; main branch always runs both MI35X and MI300X
ci:sglang SGLang integration tests: DeepSeek-R1-MXFP4 accuracy, Qwen 3.5 accuracy
ci:atom ATOM benchmark: DeepSeek-R1-0528, GPT-OSS-120B
ci:atom_full ATOM accuracy suite for PR and main models from ATOM models_accuracy.json
ci:vllm vLLM benchmark: GPT-OSS-120B, DeepSeek-R1-0528, Kimi-K2.5
ci:all All standard extended tests (excludes ci:atom_full)

Only add ci:atom_full for FlyDSL or Triton upgrades.
Add labels via the sidebar or gh pr edit 3601 --add-label <label>

@JiaoliangYu
Copy link
Copy Markdown

looks good to merge

Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

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.

3 participants