lucasmccabe-lmi/oig_small_chip2_noncode

---
dataset_info:
features:
- name: instruction
dtype: string
- name: output
dtype: string
- name: input
dtype: string
splits:
- name: train
num_bytes: 54969398.0
num_examples: 150453
download_size: 34598598
dataset_size: 54969398.0
---
# Dataset Card for "oig_small_chip2_noncode"

From LAION's Open Instruction Generalist (OIG) dataset, we provide a subset whose samples are not code-related. OIG text elements are formatted as dialogue exerpts between a "human" and "bot" agent. The code generation prompt is parsed from the initial "human" agent's statement and the resultant response from the "bot" agent's statement. We then reformat the text/response pairs according to the format of the original Alpaca dataset; that is, instruction/input/output triplets.

The OIG dataset was prepared by LAION, and released under the Apache 2.0 license.

Numbers:

Prompts: 150453

Tokens: 11522004 using the EleutherAI/gpt-neox-20b tokenizer (counting instruction+input+output)

原始信息汇总

数据集概述

数据集名称

"oig_small_chip2_noncode"

数据集来源

源自LAION的Open Instruction Generalist (OIG)数据集的一个子集，该子集不包含与代码相关的样本。

数据集内容

数据集包含对话片段，格式为“人类”和“机器人”代理之间的对话。数据集中的文本/响应对被重新格式化为指令/输入/输出三元组，与原始Alpaca数据集的格式一致。

数据集特征

• instruction: 数据类型为字符串
• output: 数据类型为字符串
• input: 数据类型为字符串

数据集分割

• train: 包含150453个样本，总字节数为54969398.0

数据集大小

• 下载大小: 34598598字节
• 数据集大小: 54969398.0字节

数据集许可证

Apache 2.0许可证

数据集统计

• 提示数量: 150453
• 令牌数量: 11522004（使用EleutherAI/gpt-neox-20b分词器计数指令+输入+输出）