This dataset is basically the ultimate knowledge distillation, for when Pruning qwen-2.5 models. Includes 900k rows of three different qwen2.5 datasets including math code and general questions, all from magpie align. Huge thanks to them.
Mix of
Magpie-Align/Magpie-Qwen2.5-Coder-Pro-300K-v0.1, Magpie-Align/Magpie-Qwen2.5-Pro-300K-Filtered, Magpie-Align/Magpie-Qwen2.5-Math-Pro-300K-v0.1.
dataset_info:
features:
name: conversations
list:
name: from
dtype: string
name: value
dtype: string… See the full description on the dataset page:
https://huggingface.co/datasets/Pinkstackorg/Magpie-Align-qwen2.5-mixed.
