Repository navigation
Flow2Code is a dataset for generating code from flowcharts. Each task includes reference code, original task and test fields, three PNG diagrams (code flowchart, UML and pseudocode), and two DOT representations.
| Download | Programs | Images |
|---|---|---|
flow2code-main-5622.tar.gz |
5,622 | 16,866 |
flow2code-supplementary-humaneval-x-javascript-164.tar.gz |
164 | 492 |
The supplementary package contains HumanEval-X JavaScript; its task IDs do not overlap the main dataset. Each archive is complete and can be extracted into an empty directory. Python 3.10+ is sufficient; no additional packages are required.
After extracting the main archive, run python3 scripts/validate_dataset.py. For the supplementary archive, run python3 scripts/validate_dataset.py --bundle supplementary. Use --bundle all when both are installed. SHA256SUMS and packages.json provide archive hashes, sizes and counts.
No train/test split is assigned. The release includes data and an integrity checker; it does not include a model inference or evaluation harness. See dataset format and source-specific licenses.
The 50 MCEval HTML tasks retain empty original test fields; task descriptions, reference code and all diagrams are included.