qingy2024/Dolci-Think-SFT-ctx8k
收藏Hugging Face2026-02-26 更新2026-03-29 收录
下载链接：
https://hf-mirror.com/datasets/qingy2024/Dolci-Think-SFT-ctx8k
下载链接
链接失效反馈官方服务：
资源简介：
---
dataset_info:
features:
- name: messages
list:
- name: content
dtype: string
- name: role
dtype: string
- name: id
dtype: string
- name: source
dtype: string
- name: token_count
dtype: int64
splits:
- name: train
num_bytes: 8889709856
num_examples: 1003315
- name: val
num_bytes: 987750461
num_examples: 111480
- name: train_balanced
num_bytes: 886024923
num_examples: 99999
- name: val_balanced
num_bytes: 98456074
num_examples: 11112
download_size: 10926499351
dataset_size: 10861941314
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: val
path: data/val-*
- split: train_balanced
path: data/train_balanced-*
- split: val_balanced
path: data/val_balanced-*
---
# Dolci-Think-SFT-ctx8k
This is a processed version of [allenai/Dolci-Think-SFT-32B](https://huggingface.co/datasets/allenai/Dolci-Think-SFT-32B) filtered so that only rows with context length < 8,192 tokens are kept. An additional `token_count` column is added for optional additional filtering.
**Splits:**
- `train`: 1.11M rows of high-quality SFT data.
- `train_balanced`: filter `train` to only rows with between 1000 and 7000 tokens, then randomly sampled 100k rows (seed 42).
- `val`: everything in `train` that's not in `train_balanced`.
提供机构：
qingy2024
