MathAndMagic/function-calling

---
language:
- en
license: apache-2.0
size_categories:
- 10K<n<100K
task_categories:
- conversational
dataset_info:
features:
- name: conversations
list:
- name: from
dtype: string
- name: value
dtype: string
- name: functions_enabled
dtype: bool
- name: function_called
dtype: bool
splits:
- name: train
num_bytes: 130998694
num_examples: 86864
download_size: 45215003
dataset_size: 130998694
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---

# function-calling

The purpose of this dataset is to give function calling abilities to your LLM.

Derived from [hypervariance/function-calling-sharegpt](https://huggingface.co/datasets/hypervariance/function-calling-sharegpt) with a system prompt cleanup and other minor fixes.

## Prompt format

### With functions

```
## Configuration

Functions: enabled

## Available Functions

{ ... }
{ ... }

---

You are a helpful assistant.
```

### Without functions

```
## Configuration

Functions: disabled

---

You are a helpful assistant.
```

## Function calling

Dataset uses the following format for the function call request:

```
<|fn_start|>{"name": "generate_password", "arguments": "{\"length\": 42}"}<|fn_end|>
```

Note, that `arguments` is a JSON string in order to comply with the OpenAI format.