Support fa3/fa4 flashmask benchmarking and testing with versioned ground truth and multi-row radar plots#25
Open
baoqiwen wants to merge 2 commits into
Open
Support fa3/fa4 flashmask benchmarking and testing with versioned ground truth and multi-row radar plots#25baoqiwen wants to merge 2 commits into
baoqiwen wants to merge 2 commits into
Conversation
baoqiwen
force-pushed
the
bqw_sm90
branch
5 times, most recently
from
July 24, 2026 09:36
8aec933 to
fa8c67b
Compare
…und truth and multi-row radar plots
baoqiwen
commented
Jul 24, 2026
| DV, | ||
| dtype = 'bf16', | ||
| use_sink = False, | ||
| backend = 'cpp', |
Contributor
Author
There was a problem hiding this comment.
允许 fa3 性能测试用原来的 cpp 版本以及新增实现 cutedsl 版本
| } | ||
|
|
||
| # Global Sliding Window is enabled for fa3, but disabled for fa4. | ||
| if fm_version == 3: |
Contributor
Author
There was a problem hiding this comment.
fa3 新增 global sliding window 测试
baoqiwen
force-pushed
the
bqw_sm90
branch
2 times, most recently
from
July 24, 2026 09:50
a2e936d to
ac861b5
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
benchmark_flashmask.py：新增 --backend 参数（cpp/cutedsl），仅 fa3 允许切换后端（fa4 恒 cutedsl、fa1/fa2 仅 cpp，非 fa3 传 cutedsl 会报错）；Global Sliding Window 用例改为 fa3 启用、fa4 关闭。
test_fwd_md5sum.py / test_bwd_md5sum.py：ground truth 文件名按 fa_versions 带版本后缀（flashmask_{fwd,bwd}_gt_fa3.json / _fa4.json），录制、加载、存在性检查统一走 get_gt_filename()，使 fa3 和 fa4 的 gt 可共存。
gt json 文件：原 fa4 的 flashmask_{fwd,bwd}_gt.json 重命名为 fa4.json；新增 fa3 的 flashmask{fwd,bwd}_gt_fa3.json。
test_flashmask.py：(2, 7600, 7600, 32, 8) fa3 跳过（避免 OOM）。
plot_radar.py：把单行雷达图重构为多行网格（plot_radar_grid，fwd/bwd/total 一张图）；类别按 Operation 名称对齐取交集，避免不同 benchmark 掩码集合错位；新增 total 提升百分比区间统计与 suptitle；字体改 DejaVu Sans、屏蔽 matplotlib 警告、plt.close 收尾。