---
license: mit
language:
- en
pretty_name: 1B-SFT — High-Quality Instruction Dataset for ~1B LLMs
tags:
- sft
- instruction-tuning
- synthetic
- sharegpt
---

# 1B-SFT — High-Quality Instruction Dataset for a ~1B LLM

A curated, deduplicated instruction-tuning dataset in **ShareGPT format**, built for fine-tuning a 1B-class model (Qwen2.5-1.5B, Llama-3.2-1B, Gemma-2-2B).

**5,235 samples** — train 4,919 / val 158 / test 158.

## Quality guarantees
- **Math**: every answer is computed by the generator (arithmetic, percentages, word problems, multi-step problems, unit conversions, linear equations, sequences, fractions) with step-by-step working.
- **Coding**: every code sample is executed and unit-tested before inclusion (33 task types + 10 bug-fix tasks); the shipped code blocks were re-executed against the final files (1,047/1,047 pass).
- **Reasoning**: ordering puzzles, liar-guard puzzles, syllogisms, seating puzzles — all consistent by construction.
- **Hand-written core**: ~260 manually authored seeds across knowledge, explain, creative writing, roleplay (flirty companion, boxing coach, motorcycle buddy, DM), advice, summarization, rewriting, and multi-turn chat.
- Exact + SimHash near-duplicate removal; artifact filtering; deterministic split (94/3/3).

## Format (ShareGPT, one JSON per line)
```json
{"system": "optional persona...",
 "conversations": [
   {"from": "human", "value": "..."},
   {"from": "gpt", "value": "..."}]}
```

## Category breakdown
| Category | Count |
|---|---|
| math | 3,251 |
| coding | 1,055 |
| reasoning | 549 |
| trivia | 140 |
| knowledge | 52 |
| roleplay | 47 |
| creative | 42 |
| explain | 36 |
| advice | 17 |
| rewrite | 17 |
| chat | 16 |
| summarization | 13 |

## Loading
```python
from datasets import load_dataset
ds = load_dataset("andro124543/1b-sft-dataset", split="train")
```
or use `train.jsonl` / `val.jsonl` / `test.jsonl` directly with LLaMA-Factory (sharegpt formatting) or axolotl.

## Suggested recipe for a 1B model (SFT)
- Base: `Qwen/Qwen2.5-1.5B` (ChatML) or `meta-llama/Llama-3.2-1B`
- lr 2e-5 cosine, batch 8–16, 3 epochs, max_seq_len 1024–2048, packing on
- QLoRA 4-bit (rank 32) for limited compute; export to GGUF for Ollama

## Rebuild / extend
The dataset is fully reproducible: `content.py` (hand-written seeds), `generators.py` (verified programmatic generators), `build.py` (dedup → filter → split → stats). Run `python build.py` to regenerate deterministically.