Shpigford/cron-mini
Shpigford/cron-mini is a 1.5 billion parameter language model fine-tuned from Qwen/Qwen2.5-1.5B-Instruct, specifically designed for converting natural-language schedules into cron expressions and systemd OnCalendar strings. This model excels at handling complex scheduling requests, including holidays, ordinal weekdays, negative specifications, and time zones. It provides a specialized solution for automating task scheduling based on human-readable inputs.
Loading preview...
Overview
Shpigford/cron-mini is a specialized 1.5 billion parameter language model, fine-tuned from Qwen/Qwen2.5-1.5B-Instruct, to accurately convert natural-language schedule descriptions into machine-readable cron expressions and systemd OnCalendar strings. It processes a wide array of scheduling nuances, outputting results in a structured JSON format, and includes notes for limitations or approximations.
Key Capabilities
- Comprehensive Schedule Interpretation: Handles standard daily, weekly, monthly schedules, as well as complex patterns like holidays (e.g., Christmas, Thanksgiving), casual time references ("lunchtime"), and ordinal weekdays ("second Tuesday").
- Negative Specifications: Accurately processes requests like "every day except Sunday" or "all months except December."
- Advanced Time Handling: Supports sub-minute intervals (for systemd), awkward intervals (e.g., every 90 minutes), and compound schedules requiring multiple cron lines. It also manages time zones by setting
TZ=for cron or usingAsia/Tokyo-style for systemd. - Robustness: Tolerates typos and informal phrasings, and can incorporate systemd-specific features such as
OnBootSec=,Persistent=, andRandomizedDelaySec=. - Evaluation: Achieves 80.2% exact match for cron and 78.0% for systemd on a held-out test set of 91 cases, with 95.6% syntactic validity for cron expressions.
Good For
- Developers needing to automate the conversion of user-provided natural language into cron or systemd schedules.
- Applications requiring precise scheduling logic from flexible text inputs.
- Use cases where a small, specialized model is preferred for efficiency and targeted performance on schedule conversion tasks.