File size: 2,156 Bytes
db7354e 0ec8116 db7354e b5509c0 f3cbe79 b5509c0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 |
---
language:
- en
datasets:
- glaiveai/glaive-function-calling-v2
---
The model was finetuned using the glaive dataset using qlora and full finetuning using FSDP.
Dataset link : [Link here](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2)
For training , inference and evaluation kindly check this repository:
https://github.com/Srini-98/Function-Calling-Using-Mistral
Use the following prompt format
```
SYSTEM: You are a helpful assistant with access to the following functions. Use them if required -
{
"name": "function_name",
"description": "description",
"parameters": {
"type": "object",
"properties": {
"param_name1": {
"type": "string",
"description": "description of param"
},
"param_name2": {
"type": "string",
"description": "description of param"
},
"param_name3":{
"type: "string",
"description" : "description of param"
}
},
"required": [
"param_name1",
]
}
}
USER: {question here}
ASSISTANT: {model answer} <|endoftext|>
```
Example:
```
SYSTEM: You are a helpful assistant with access to the following functions. Use them if required -
{
"name": "calculate_tax",
"description": "Calculate the tax amount",
"parameters": {
"type": "object",
"properties": {
"income": {
"type": "number",
"description": "The income amount"
}
},
"required": [
"income"
]
}
}
USER: Hi, I need to calculate my tax for this year. My income is $70,000.
ASSISTANT: <functioncall> {"name": "calculate_tax", "arguments": '{"income": 70000}'} <|endoftext|>
FUNCTION RESPONSE: {"tax_amount": 17500}
ASSISTANT: Based on your income, your tax for this year is $17,500. <|endoftext|>
```
The answer generation can be stopped with the <|endoftext|> token. You can add multiple functions as well and set param names. "Required" field forces model to always call that param. |