Trelis/englishQ_germanA

---
dataset_info:
features:
- name: messages
list:
- name: content
dtype: string
- name: role
dtype: string
splits:
- name: train
num_bytes: 224640
num_examples: 1000
download_size: 119367
dataset_size: 224640
datasets:
- maxidl/LeoLM-HellaSwag_de-fixed
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---
# English Q, German A

Data set to fine-tune a model to respond in German.

Created from a fixed LeoLM Hellaswag [dataset](https://huggingface.co/datasets/maxidl/LeoLM-HellaSwag_de-fixed?row=0) using this code:
```
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("maxidl/LeoLM-HellaSwag_de-fixed", split='train')

# Function to keep only the conversation column renamed to 'messages' and remove all other columns
def format_to_messages(example):
return {'messages': [{"role": "user", "content": example['ctx']}, {"role": "assistant", "content": example['endings_de'][-1]}]}

# Apply the transformation
messages_dataset = dataset.map(format_to_messages, remove_columns=dataset.column_names)

print(messages_dataset)
```

原始信息汇总

数据集概述

数据集基本信息

• 名称: maxidl/LeoLM-HellaSwag_de-fixed
• 特征:
  • messages:
    • content (dtype: string)
    • role (dtype: string)

数据集分割

• 训练集:
  • 示例数量: 1000
  • 数据大小: 224640字节

数据集大小

• 下载大小: 119367字节
• 数据集总大小: 224640字节