heegyu/UltraInteract_pair_longest_multiturn

---
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
dataset_info:
features:
- name: task
dtype: string
- name: dataset
dtype: string
- name: trajectory
list:
- name: from
dtype: string
- name: value
dtype: string
- name: chosen
dtype: string
- name: rejected
dtype: string
- name: id
dtype: string
- name: parent_id
dtype: string
- name: __index_level_0__
dtype: string
- name: __index_level_1__
dtype: int64
splits:
- name: train
num_bytes: 196675986
num_examples: 26658
download_size: 78610248
dataset_size: 196675986
---
# Dataset Card for "UltraInteract_pair_longest_multiturn"
- Original Dataset: [openbmb/UltraInteract_pair](https://huggingface.co/datasets/openbmb/UltraInteract_pair)
- Filtered multiturn instances and longest item for each reasoning tree.

Data processing code:
```python
from datasets import load_dataset, Dataset

dataset = load_dataset("openbmb/UltraInteract_pair")

df = dataset['train'].to_pandas()
df["turns"] = df["trajectory"].apply(lambda x: len(x))
df = df[df.turns > 1]
df = df.groupby("parent_id").apply(lambda x: x[x["turns"] == x["turns"].max()])
print(df)
print(df.shape)

df = df.drop(columns=["turns"])
dataset['train'] = Dataset.from_pandas(df)
dataset.push_to_hub("heegyu/UltraInteract_pair_longest_multiturn")
```

原始信息汇总

数据集概述

数据集名称

UltraInteract_pair_longest_multiturn

数据集配置

• config_name: default
• data_files:
  • split: train
  • path: data/train-*

数据集特征

• task: string
• dataset: string
• trajectory:
  • from: string
  • value: string
• chosen: string
• rejected: string
• id: string
• parent_id: string
• index_level_0: string
• index_level_1: int64

数据集分割

• name: train
• num_bytes: 196675986
• num_examples: 26658

数据集大小

• download_size: 78610248
• dataset_size: 196675986