Datasetpublic

michaelowusuntim6/linux-kernel-ioctl-qwen35

Linux ioctl Census Description Teaches domain-specific instruction following and code generation for this expert. Source mjbommar/linux-ioctl-census - CC-BY-4.0 Formatted for the MoE-orchestrator project (https://github.com/michaelowusuntim6/MoE-orchestrator). Expert target: linux_kernel. Format Each record is a JSON object with a messages field formatted for Qwen3.5's native chat template: {"messages": [ {"role": "system"… See the full description on the dataset page: https://huggingface.co/datasets/michaelowusuntim6/linux-kernel-ioctl-qwen35.

sourceHugging Facecc-by-4.0updated 5d agoView on Hugging Face

0likes43downloads

Dataset Card

Linux ioctl Census

Description

Teaches domain-specific instruction following and code generation for this expert.

Source

—mjbommar/linux-ioctl-census - CC-BY-4.0

Formatted for the MoE-orchestrator project (https://github.com/michaelowusuntim6/MoE-orchestrator). Expert target: linux_kernel.

Format

Each record is a JSON object with a messages field formatted for Qwen3.5's native chat template:

json

{"messages": [
  {"role": "system", "content": "..."},
  {"role": "user", "content": "..."},
  {"role": "assistant", "content": "..."}
]}

The records are consumed via tokenizer.apply_chat_template(). Special tokens (<|im_start|>, <|im_end|>) are added by the template, never embedded in content.

Splits

—train: 1,261 records
—val: 28 records

Usage

python

from datasets import load_dataset
ds = load_dataset("michaelowusuntim6/linux-kernel-ioctl-qwen35", split="train")
print(ds[0]["messages"])

License

cc-by-4.0. Upstream sources keep their own licences - see the source list above and docs/DATASET_SOURCES.md in the MoE-orchestrator repository for per-source detail.