Dataset · beyoru
Kimi K3 distillation traces for SFT and tool use
beyoru/kimi-k3-distillation
A collection of responses and tool calls from the kimi-code/k3 model for SFT fine-tuning.
What's inside
A slice of another dataset filtered exclusively for the kimi-code/k3 teacher model in English and Chinese. The card lists 4347 rows (3918 train, 212 validation, 217 test) in text and tabular format, available in the default config (messages and tools) and the expanded canonical variant.
What you can do with it
- SFT fine-tuning of models for code generation and external tool calling (function calling)
- Distillation of reasoning and behavior patterns from the kimi-code/k3 model into more compact LLMs
- Tasks
- Text generation
- Languages
- English, Chinese
- Size
- 100K–1M rows
- Downloads this month
- 685
- Likes
- 20
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("beyoru/kimi-k3-distillation")