File size: 1,876 Bytes
f9b07b4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
---

language:
- ru
- en
license: apache-2.0
task_categories:
- text-generation
tags:
- code
- python
- reasoning
- think
- deepseek-r1
- compiler-verified
- synthetic
- unsloth
size_categories:
- n<1K
---


# 🧠 Python Dataset (Think / DeepSeek-R1 Style)

[English](#english) | [Русский](#русский)

---

## English
**Python Dataset (Think)** is a high-purity synthetic dataset designed for teaching language models to reason step-by-step before producing Python code.

Each sample contains:
1. **Algorithmic task / Problem statement** (RU / EN).
2. **Deep step-by-step reasoning inside `<think>...</think>` tags** analyzing edge cases, algorithm logic, and time/space complexity $O(...)$.
3. **Verified Python code block** tested via AST parsing, bytecode compilation, and execution sandbox.

* Total Samples: 58
* Format: Standard ChatML (`messages`)
* Author: [Stray2024](https://huggingface.co/Stray2024)

---

## Русский
**Python Dataset (Think)** — очищенный синтетический датасет для обучения моделей глубокому пошаговому рассуждению перед генерацией кода на Python.

Каждый пример включает:
1. **Условие задачи** (русский / английский).
2. **Цепочку рассуждений внутри тегов `<think>...</think>`** (анализ алгоритма, проверка краевых случаев, сложность).
3. **Рабочий блок кода**, проверенный 3-ступенчатым фильтром компилятора (AST + байткод + тестовый запуск).

* Всего примеров: 58
* Формат: ChatML (`messages`)
* Автор: [Stray2024](https://huggingface.co/Stray2024)