Datasets:
Download README.md from openenv/sudoku-tool-calls-sft: direct link, hf CLI and curl.
- Browser
- Download file 2.76 kB
-
https://huggingface.co/datasets/openenv/sudoku-tool-calls-sft/resolve/main/README.md
- Command line
-
hf download hf://datasets/openenv/sudoku-tool-calls-sft/README.md
-
curl -L -o README.md https://huggingface.co/datasets/openenv/sudoku-tool-calls-sft/resolve/main/README.md
license: apache-2.0
task_categories:
- text-generation
tags:
- sudoku
- tool-calling
- openenv
- trl
- sft
size_categories:
- 1K<n<10K
dataset_info:
features:
- name: messages
list:
- name: role
dtype: string
- name: content
dtype: string
- name: reasoning_content
dtype: string
- name: tool_calls
list:
- name: type
dtype: string
- name: function
struct:
- name: name
dtype: string
- name: arguments
struct:
- name: row
dtype: int64
- name: col
dtype: int64
- name: number
dtype: int64
- name: name
dtype: string
- name: tools
dtype: string
splits:
- name: train
num_bytes: 17749144
num_examples: 1000
download_size: 1793721
dataset_size: 17749144
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Sudoku tool-call SFT
1,000 solved Sudoku games written as tool-calling conversations, for a short SFT warmup before training with GRPO on the OpenEnv Sudoku environment. Used in the OpenEnv Sudoku tutorial to train openenv/Qwen3-1.7B-Sudoku-SFT.
A solver played every game against the TextArena Sudoku-v0 environment from OpenEnv 0.10.0, run in-process, on boards from TextArena's own generator with 50, 55, 60 or 65 clues (15 to 31 empty cells). It only uses single candidates: the one digit a cell can take, or the one cell a digit can take in a row, column or box. Boards that need more than that are skipped, so every reasoning step can be checked on the grid.
messages: the user prompt (the rules plus the starting grid), then one assistant turn per move, with a shortreasoning_contentand a call to theplacetool, each followed by atoolmessage with the updated grid. The last assistant turn says the puzzle is solved.tools: theplacetool schema, as a JSON string.
This is the format TRL's GRPOTrainer renders with environment_factory, so a model trained on it plays the same way during RL. Train with SFTTrainer and assistant_only_loss=True. Games are 8k to 15k tokens long, since every tool result carries the whole grid.
Games have 24 moves on average, and the reasoning is about 45 words per move. None of the boards appears in the evaluation puzzles of the tutorial.
Regenerate
uv run generate_sft_data.py --games 1000 --push-to-hub <user>/sudoku-tool-calls-sft
The script is in OpenEnv examples/sudoku.