| license: mit | |
| language: | |
| - en | |
| pretty_name: SLM Synthetic Task Code | |
| tags: | |
| - synthetic | |
| - llm | |
| - pretraining | |
| - reasoning | |
| - code | |
| # SLM Synthetic Task Code | |
| Beginner/intermediate Python tasks with short plans and code snippets. | |
| Generated date: 2026-05-30 | |
| Total records: 599,136 | |
| Storage size: 458.1 MB | |
| ## Files | |
| | File | Records | Size | | |
| |---|---:|---:| | |
| | `task_code.jsonl` | 599,136 | 458.1 MB | | |
| ## Record Format | |
| Each row is a JSON object stored as one line in JSONL format. | |
| Schema fields: | |
| `type`, `task`, `plan`, `code` | |
| ## Deduplication | |
| The uploaded split is exact-deduplicated. Fuzzy MinHash deduplication is not used for synthetic signals because synthetic examples often share useful structure by design. | |
| ## Intended Use | |
| This dataset is intended for SLM data experiments, pretraining/continued-pretraining mixtures, pretraining, and behavior evaluation. | |
| ## Limitations | |
| The data is synthetic and should be inspected before use in training or evaluation. It may contain simple, repetitive, or imperfect examples. It should not be treated as a source of authoritative factual knowledge. | |
| ## License | |
| This dataset is released under the MIT License. | |