File size: 5,213 Bytes
f5a88af
 
234ce04
f5a88af
 
 
 
 
 
 
 
 
 
 
1584156
 
 
 
 
1f8050c
 
1584156
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1f8050c
 
1584156
 
 
 
 
 
 
 
 
 
 
 
 
f5a88af
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
---
base_model:
- rtzr/ko-gemma-2-9b-it
- NeuroWhAI/ko-gemma-2-9b-it-fn
license: gemma
tags:
- merge
- mergekit
- lazymergekit
---

# ko-gemma-2-9b-it-fn-merged

ko-gemma-2-9b-it-fn-merged is a merge of the following models using [mergekit](https://github.com/arcee-ai/mergekit):
* [rtzr/ko-gemma-2-9b-it](https://huggingface.co/rtzr/ko-gemma-2-9b-it) : korean base.
* [NeuroWhAI/ko-gemma-2-9b-it-fn](https://huggingface.co/NeuroWhAI/ko-gemma-2-9b-it-fn) : fine-tuned with a function call dataset.

## Chat template

도ꡬ가 μžˆλŠ” 경우:

```
<start_of_turn>user
You are a helpful assistant with access to the following functions. Use them if required -
{
    "name": "search_books",
    "description": "Search for books based on title, author, or genre",
    "parameters": {
        "type": "object",
        "properties": {
            "title": {
                "type": "string",
                "description": "The title of the book"
            },
            "author": {
                "type": "string",
                "description": "The author of the book"
            },
            "genre": {
                "type": "string",
                "description": "The genre of the book"
            }
        },
        "required": []
    }
}

To use these functions respond with:
<functioncall> {"name": "function_name", "arguments": {"arg_1": "value_1", "arg_1": "value_1", ...}} </functioncall>

Edge cases you must handle:
 - If there are no functions that match the user request, you will respond politely that you cannot help.

User Question:
책을 μ°Ύκ³  μžˆλŠ”λ° 제λͺ©μ„ κΈ°μ–΅ν•˜μ§€ λͺ»ν•˜λ„€μš”. μ €μžμ˜ 이름은 쑰지 μ˜€μ›°μž…λ‹ˆλ‹€.<end_of_turn>
<start_of_turn>model
<functioncall> {"name": "search_books", "arguments": {"author": "George Orwell"}} </functioncall> <end_of_turn>
<start_of_turn>user
{"books": [{"title": "1984", "author": "George Orwell", "genre": "λ””μŠ€ν† ν”Όμ•„"}, {"title": "동물농μž₯", "author": "George Orwell", "genre": "μ •μΉ˜ ν’μž"}]} <end_of_turn>
<start_of_turn>model
쑰지 μ˜€μ›°μ˜ 책을 두 ꢌ μ°Ύμ•˜μ–΄μš”. 첫 λ²ˆμ§ΈλŠ” "1984"λΌλŠ” λ””μŠ€ν† ν”Όμ•„ μ†Œμ„€μ΄μ—μš”. 두 λ²ˆμ§ΈλŠ” μ •μΉ˜ ν’μžμΈ "동물농μž₯"μž…λ‹ˆλ‹€. <end_of_turn>
<start_of_turn>user
였, λ§žμ•„μš”, "동물농μž₯"을 μ°Ύκ³  μžˆμ—ˆμ–΄μš”. 같은 μž₯λ₯΄μ˜ 책도 찾아쀄 수 μžˆμ„κΉŒμš”?<end_of_turn>
<start_of_turn>model
<functioncall> {"name": "search_books", "arguments": {"genre": "μ •μΉ˜ ν’μž"}} </functioncall> <end_of_turn>
<start_of_turn>user
{"books": [{"title": "Catch-22", "author": "Joseph Heller", "genre": "μ •μΉ˜ ν’μž"}, {"title": "Slaughterhouse-Five", "author": "Kurt Vonnegut", "genre": "μ •μΉ˜ ν’μž"}]} <end_of_turn>
<start_of_turn>model
λ‹Ήμ—°νžˆμš”, μ •μΉ˜ ν’μž μž₯λ₯΄μ— ν•΄λ‹Ήν•˜λŠ” μ±… 두 ꢌ 더 μ•Œλ €λ“œλ¦΄κ²Œμš”. Joseph Heller의 "Catch-22"와 Kurt Vonnegut의 "Slaughterhouse-Five"μž…λ‹ˆλ‹€. <end_of_turn>
```

도ꡬ가 μ—†λŠ” 경우:

```
<start_of_turn>user
You are a helpful assistant, with no access to external functions.

User Question:
주어진 숫자 μ‹œν€€μŠ€λŠ” λ‹€μŒκ³Ό κ°™μŠ΅λ‹ˆλ‹€: 2, 4, 6, 8, 10. 각 μˆ«μžλŠ” λ°”λ‹€μ˜ νŒŒλ™κ³Ό λ§ˆμ°¬κ°€μ§€λ‘œ μ‹œκ°„μ˜ μˆœκ°„μ„ λ‚˜νƒ€λƒ…λ‹ˆλ‹€. μ‹œν€€μŠ€λ₯Ό κ³„μ†ν•˜λ©΄ 각 숫자 μ‚¬μ΄μ˜ 차이가 λ™μΌν•˜λ‹€λŠ” 것을 μ•Œ 수 μžˆμŠ΅λ‹ˆλ‹€. 마치 λ°”λžŒκ³Ό ν•΄λ₯˜κ°€ λ°”λ‹€μ˜ νŒŒλ™μ„ ν˜•μ„±ν•˜λŠ” κ²ƒμ²˜λŸΌ, 무언가가 숫자 μ‹œν€€μŠ€λ₯Ό ν˜•μ„±ν•˜κ³  μžˆμŠ΅λ‹ˆλ‹€. μ‹œν€€μŠ€μ—μ„œ λ‹€μŒ μˆ«μžλŠ” 무엇이며, μ‹œκ°„κ³Όμ˜ κ΄€κ³„μ—μ„œ μ–΄λ–€ 의미λ₯Ό κ°€μ§€λŠ”μ§€ μ•Œλ €μ£Όμ„Έμš”.<end_of_turn>
<start_of_turn>model
기술된 νŒ¨ν„΄μ— 따라, μ‹œν€€μŠ€μ—μ„œ λ‹€μŒ μˆ«μžλŠ” 12μž…λ‹ˆλ‹€. μ΄λŠ” 숫자 10으둜 ν‘œν˜„λœ 이전 μ‹œκ°„μ΄ 2개의 μ‹œκ°„ λ‹¨μœ„ λ’€λ₯Ό λ‚˜νƒ€λƒ…λ‹ˆλ‹€. νŒ¨ν„΄μ€ μ‹œν€€μŠ€μ—μ„œ 각 후속 μˆ«μžκ°€ 이전 숫자 λ‹€μŒμ— 2개의 μ‹œκ°„ λ‹¨μœ„μΈ 것을 μ‹œμ‚¬ν•©λ‹ˆλ‹€. λ”°λΌμ„œ 12 λ‹€μŒ μˆ«μžλŠ” 14이며, κ·Έ λ‹€μŒμ€ 16이 λ©λ‹ˆλ‹€. μ΄λŸ¬ν•œ νŒ¨ν„΄μ€ μΌμ •ν•œ 차이가 μžˆλŠ” λ“±μ°¨ μˆ˜μ—΄λ‘œ μ•Œλ €μ Έ μžˆμŠ΅λ‹ˆλ‹€. <end_of_turn>
<start_of_turn>user
νŒ¨ν„΄μ— κ΄€ν•΄ μ΄μ•ΌκΈ°ν•˜μžλ©΄, ν”Όλ³΄λ‚˜μΉ˜ μˆ˜μ—΄μ— λŒ€ν•΄ κΆκΈˆν•©λ‹ˆλ‹€. 이것은 무엇이고 μ–΄λ–»κ²Œ μž‘λ™ν•˜λ‚˜μš”?<end_of_turn>
<start_of_turn>model
ν”Όλ³΄λ‚˜μΉ˜ μˆ˜μ—΄μ€ 각 μˆ«μžκ°€ 이전 두 숫자의 ν•©μœΌλ‘œ 이루어진 μˆ˜μ—΄μž…λ‹ˆλ‹€. μ‹œμž‘μ€ 0κ³Ό 1λ‘œλΆ€ν„° μ‹œμž‘ν•˜μ—¬ 이전 두 숫자λ₯Ό 더해 λ‹€μŒ 숫자λ₯Ό μ–»λŠ” λ°©μ‹μœΌλ‘œ κ³„μ†λ©λ‹ˆλ‹€. λ”°λΌμ„œ μˆ˜μ—΄μ€ 0, 1, 1, 2, 3, 5, 8, 13, 21 λ“±μœΌλ‘œ μ΄μ–΄μ§‘λ‹ˆλ‹€. 이전 두 숫자λ₯Ό λ”ν•˜λ©΄μ„œ λ‹€μŒ 숫자λ₯Ό κ΅¬ν•˜λŠ” νŒ¨ν„΄μ€ λ¬΄ν•œνžˆ μ΄μ–΄μ§€κ²Œ λ©λ‹ˆλ‹€. ν”Όλ³΄λ‚˜μΉ˜ μˆ˜μ—΄μ€ μ΄νƒˆλ¦¬μ•„ μˆ˜ν•™μžμΈ λ ˆμ˜€λ‚˜λ₯΄λ„ ν”Όλ³΄λ‚˜μΉ˜κ°€ μ„œμ–‘μ— μ†Œκ°œν•œ κ²ƒμ—μ„œ 이름을 λ”°μ™”μŠ΅λ‹ˆλ‹€. ν”Όλ³΄λ‚˜μΉ˜ μˆ˜μ—΄μ€ λ‚˜λ¬΄μ˜ 가지 치기, 쀄기에 잎이 λ°°μ—΄λ˜λŠ” ν˜•νƒœ, λ‹¬νŒ½μ΄ 껍질의 λͺ¨μ–‘κ³Ό 같이 μžμ—°μ—μ„œ 많이 λ‚˜νƒ€λ‚©λ‹ˆλ‹€. <end_of_turn>
```

## 🧩 Configuration

```yaml
models:
  - model: rtzr/ko-gemma-2-9b-it
  - model: NeuroWhAI/ko-gemma-2-9b-it-fn
    parameters:
      density: 0.3
      weight: 0.2
merge_method: dare_ties
base_model: rtzr/ko-gemma-2-9b-it
parameters:
  int8_mask: true
dtype: bfloat16
```