Dao-AILab/flash-attention has 105,794 source lines of code out of 134,403 total lines across 999 files, counted with tokei from the main ref at commit 328dd7b5a690 on November 16, 2026. Tests, docs and generated files are excluded under OctoCounts' canonical public options.
As of 2026-11-16 (commit 328dd7b5a690), Dao-AILab/flash-attention contains 134,403 total lines: 105,794 code, 15,470 comments, 13,139 blank, across 999 files in 13 languages (top language: Python 74.5%). Canonical public options (tests, docs and generated files excluded). Counted with tokei via OctoCounts.
OctoCounts produced this report by resolving Dao-AILab/flash-attention to commit 328dd7b5a690, downloading the repository source archive, and counting files selected by the effective rules with tokei, the open-source line counter written in Rust. The table below breaks the count down by programming language into files, total lines, code lines, comment lines, and blank lines, so the figures can be compared across languages and projects. Results are cached by commit, tokei version, and analysis options; an exact reproduction needs the same commit, engine version, and effective counting rules.
This is a large codebase by counted code lines. Code represents 78.7% of all lines, comments represent 11.5%, and the repository averages 106 code lines per file. Python accounts for 74.5% of counted code.
| Language | Files | Lines | Code | Comments | Blanks |
|---|---|---|---|---|---|
| Python | 184 | 94525 | 78836 | 7866 | 7823 |
| C Header | 43 | 11559 | 8574 | 1557 | 1428 |
| C++ | 12 | 9008 | 7497 | 613 | 898 |
| C++ Header | 12 | 6780 | 5406 | 688 | 686 |
| CUDA | 606 | 7453 | 4199 | 1940 | 1314 |
| YAML | 107 | 1440 | 1042 | 209 | 189 |
| Shell | 5 | 139 | 84 | 38 | 17 |
| TOML | 2 | 82 | 72 | 0 | 10 |
| Dockerfile | 2 | 122 | 56 | 46 | 20 |
| Autoconf | 2 | 17 | 16 | 0 | 1 |
| Makefile | 1 | 9 | 6 | 0 | 3 |
| Module-Definition | 1 | 7 | 6 | 0 | 1 |
Top language: Python 74.5%. Generated on .
Dao-AILab/flash-attention: tokei-15.0.0, policy directories-v1; docs excluded, tests excluded, generated excluded. Excluded directory names: .cache, .generated, .git, .next, __tests__, build, dist, doc, docs, documentation, fixtures, gen, generated, node_modules, target, test, tests, vendor. Ignored languages: none. Repository declarations: off. Exclusions are rule-based; they do not prove that every test, document or generated file was removed.
Saved reports remain subject to cache cleanup and replacement. Export the JSON or text for durable citation; live repository URLs can change.
Dao-AILab/flash-attention has 105,794 lines of code, plus 15,470 comment lines and 13,139 blank lines, for 134,403 total lines at commit 328dd7b5a690 on November 16, 2026. Tests, docs and generated files are excluded under OctoCounts' canonical public options.
OctoCounts resolved the public GitHub repository to commit 328dd7b5a690, downloaded the source archive, counted it with tokei, and cached the report by commit, tokei version, and analysis options.
This OctoCounts report was generated from main at commit 328dd7b5a690 on November 16, 2026.
Other public repositories with an OctoCounts report, ranked by top language and code size similarity:
Play the growth animation for Dao-AILab/flash-attention on this page — 105,794 code lines across 13 languages, counted by tokei.