Update README.md
Browse files
README.md
CHANGED
|
@@ -5,7 +5,11 @@ language:
|
|
| 5 |
base_model:
|
| 6 |
- meta-llama/Llama-3.2-1B-Instruct
|
| 7 |
---
|
| 8 |
-
> @ 2024.10.07 Model [torchtorchkimtorch/Llama-3.2-Korean-GGACHI-1B-Instruct-v1](https://huggingface.co/torchtorchkimtorch/Llama-3.2-Korean-GGACHI-1B-Instruct-v1) Released!
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
|
| 11 |
# **GGACHI-1B-version1** #
|
|
@@ -14,4 +18,154 @@ base_model:
|
|
| 14 |
|
| 15 |
GGACHI-1B-Instruct-v1λ Llama-3.2-1B-Instruct λͺ¨λΈμ κΈ°λ°μΌλ‘ νλ νκ΅μ΄ νμ€ν¬ μνμ μ΅μ νλ instruction-tuned μΈμ΄ λͺ¨λΈμ
λλ€. 230,000κ° μ΄μμ κ³ νμ§ νκ΅μ΄ λ°μ΄ν°μ
μ μ¬μ©νμ¬ fine-tuningλμμ΅λλ€.
|
| 16 |
|
| 17 |
-
GGACHI-1B-Instruct-v1 is an instruction-tuned language model optimized for Korean language tasks, based on the Llama-3.2-1B-Instruct model. It has been fine-tuned using over 230,000 high-quality Korean language datasets.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
base_model:
|
| 6 |
- meta-llama/Llama-3.2-1B-Instruct
|
| 7 |
---
|
| 8 |
+
> @ 2024.10.07 Model [torchtorchkimtorch/Llama-3.2-Korean-GGACHI-1B-Instruct-v1](https://huggingface.co/torchtorchkimtorch/Llama-3.2-Korean-GGACHI-1B-Instruct-v1) Released!
|
| 9 |
+
|
| 10 |
+
> @ 2024.10.18 Performance for KOBEST of Llama-3.2-Korean-GGACHI-1B-Instruct-v1 has been updated!
|
| 11 |
+
|
| 12 |
+
> @ Announcements) Llama-3.2-Korean-GGACHI-1B-Instruct-v2 is set to be released soon.
|
| 13 |
|
| 14 |
|
| 15 |
# **GGACHI-1B-version1** #
|
|
|
|
| 18 |
|
| 19 |
GGACHI-1B-Instruct-v1λ Llama-3.2-1B-Instruct λͺ¨λΈμ κΈ°λ°μΌλ‘ νλ νκ΅μ΄ νμ€ν¬ μνμ μ΅μ νλ instruction-tuned μΈμ΄ λͺ¨λΈμ
λλ€. 230,000κ° μ΄μμ κ³ νμ§ νκ΅μ΄ λ°μ΄ν°μ
μ μ¬μ©νμ¬ fine-tuningλμμ΅λλ€.
|
| 20 |
|
| 21 |
+
GGACHI-1B-Instruct-v1 is an instruction-tuned language model optimized for Korean language tasks, based on the Llama-3.2-1B-Instruct model. It has been fine-tuned using over 230,000 high-quality Korean language datasets.
|
| 22 |
+
|
| 23 |
+
## λͺ¨λΈ μ±λ₯ (Model Performance)
|
| 24 |
+
|
| 25 |
+
|
| 26 |
+
#### - 0 shot ####
|
| 27 |
+
<table style="width:100%; text-align:center; border-collapse:collapse;">
|
| 28 |
+
<thead>
|
| 29 |
+
<tr>
|
| 30 |
+
<th style="border:1px solid black;">Task</th>
|
| 31 |
+
<th style="border:1px solid black;">Model</th>
|
| 32 |
+
<th style="border:1px solid black;">Accuracy</th>
|
| 33 |
+
</tr>
|
| 34 |
+
</thead>
|
| 35 |
+
<tbody>
|
| 36 |
+
<tr>
|
| 37 |
+
<td rowspan="2" style="border:1px solid black;">kobest_boolq</td>
|
| 38 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 39 |
+
<td style="border:1px solid black;"><strong>0.502</td>
|
| 40 |
+
</tr>
|
| 41 |
+
<tr>
|
| 42 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 43 |
+
<td style="border:1px solid black;"><strong>0.502</td>
|
| 44 |
+
</tr>
|
| 45 |
+
<tr>
|
| 46 |
+
<td rowspan="2" style="border:1px solid black;">kobest_copa</td>
|
| 47 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 48 |
+
<td style="border:1px solid black;">0.504</td>
|
| 49 |
+
</tr>
|
| 50 |
+
<tr>
|
| 51 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 52 |
+
<td style="border:1px solid black;"><strong>0.521</strong></td>
|
| 53 |
+
</tr>
|
| 54 |
+
<tr>
|
| 55 |
+
<td rowspan="2" style="border:1px solid black;">kobest_hellaswag</td>
|
| 56 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 57 |
+
<td style="border:1px solid black;">0.358</td>
|
| 58 |
+
</tr>
|
| 59 |
+
<tr>
|
| 60 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 61 |
+
<td style="border:1px solid black;"><strong>0.380</td>
|
| 62 |
+
</tr>
|
| 63 |
+
<tr>
|
| 64 |
+
<td rowspan="2" style="border:1px solid black;">kobest_sentineg</td>
|
| 65 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 66 |
+
<td style="border:1px solid black;">0.476</td>
|
| 67 |
+
</tr>
|
| 68 |
+
<tr>
|
| 69 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 70 |
+
<td style="border:1px solid black;"><strong>0.594</strong></td>
|
| 71 |
+
</tr>
|
| 72 |
+
</tbody>
|
| 73 |
+
</table>
|
| 74 |
+
|
| 75 |
+
#### - 5 shot ####
|
| 76 |
+
<table style="width:100%; text-align:center; border-collapse:collapse;">
|
| 77 |
+
<thead>
|
| 78 |
+
<tr>
|
| 79 |
+
<th style="border:1px solid black;">Task</th>
|
| 80 |
+
<th style="border:1px solid black;">Model</th>
|
| 81 |
+
<th style="border:1px solid black;">Accuracy</th>
|
| 82 |
+
</tr>
|
| 83 |
+
</thead>
|
| 84 |
+
<tbody>
|
| 85 |
+
<tr>
|
| 86 |
+
<td rowspan="2" style="border:1px solid black;">kobest_boolq</td>
|
| 87 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 88 |
+
<td style="border:1px solid black;"><strong>0.571</td>
|
| 89 |
+
</tr>
|
| 90 |
+
<tr>
|
| 91 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 92 |
+
<td style="border:1px solid black;">0.565</td>
|
| 93 |
+
</tr>
|
| 94 |
+
<tr>
|
| 95 |
+
<td rowspan="2" style="border:1px solid black;">kobest_copa</td>
|
| 96 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 97 |
+
<td style="border:1px solid black;">0.526</td>
|
| 98 |
+
</tr>
|
| 99 |
+
<tr>
|
| 100 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 101 |
+
<td style="border:1px solid black;"><strong>0.549</strong></td>
|
| 102 |
+
</tr>
|
| 103 |
+
<tr>
|
| 104 |
+
<td rowspan="2" style="border:1px solid black;">kobest_hellaswag</td>
|
| 105 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 106 |
+
<td style="border:1px solid black;">0.364</td>
|
| 107 |
+
</tr>
|
| 108 |
+
<tr>
|
| 109 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 110 |
+
<td style="border:1px solid black;"><strong>0.398</td>
|
| 111 |
+
</tr>
|
| 112 |
+
<tr>
|
| 113 |
+
<td rowspan="2" style="border:1px solid black;">kobest_sentineg</td>
|
| 114 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 115 |
+
<td style="border:1px solid black;">0.725</td>
|
| 116 |
+
</tr>
|
| 117 |
+
<tr>
|
| 118 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 119 |
+
<td style="border:1px solid black;"><strong>0.795</strong></td>
|
| 120 |
+
</tr>
|
| 121 |
+
</tbody>
|
| 122 |
+
</table>
|
| 123 |
+
|
| 124 |
+
#### - 10 shot ####
|
| 125 |
+
<table style="width:100%; text-align:center; border-collapse:collapse;">
|
| 126 |
+
<thead>
|
| 127 |
+
<tr>
|
| 128 |
+
<th style="border:1px solid black;">Task</th>
|
| 129 |
+
<th style="border:1px solid black;">Model</th>
|
| 130 |
+
<th style="border:1px solid black;">Accuracy</th>
|
| 131 |
+
</tr>
|
| 132 |
+
</thead>
|
| 133 |
+
<tbody>
|
| 134 |
+
<tr>
|
| 135 |
+
<td rowspan="2" style="border:1px solid black;">kobest_boolq</td>
|
| 136 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 137 |
+
<td style="border:1px solid black;"><strong>0.593</td>
|
| 138 |
+
</tr>
|
| 139 |
+
<tr>
|
| 140 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 141 |
+
<td style="border:1px solid black;">0.571</td>
|
| 142 |
+
</tr>
|
| 143 |
+
<tr>
|
| 144 |
+
<td rowspan="2" style="border:1px solid black;">kobest_copa</td>
|
| 145 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 146 |
+
<td style="border:1px solid black;">0.525</td>
|
| 147 |
+
</tr>
|
| 148 |
+
<tr>
|
| 149 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 150 |
+
<td style="border:1px solid black;"><strong>0.549</strong></td>
|
| 151 |
+
</tr>
|
| 152 |
+
<tr>
|
| 153 |
+
<td rowspan="2" style="border:1px solid black;">kobest_hellaswag</td>
|
| 154 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 155 |
+
<td style="border:1px solid black;">0.356</td>
|
| 156 |
+
</tr>
|
| 157 |
+
<tr>
|
| 158 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 159 |
+
<td style="border:1px solid black;"><strong>0.394</td>
|
| 160 |
+
</tr>
|
| 161 |
+
<tr>
|
| 162 |
+
<td rowspan="2" style="border:1px solid black;">kobest_sentineg</td>
|
| 163 |
+
<td style="border:1px solid black;">Llama-3.2-1B-Instruct</td>
|
| 164 |
+
<td style="border:1px solid black;">0.768</td>
|
| 165 |
+
</tr>
|
| 166 |
+
<tr>
|
| 167 |
+
<td style="border:1px solid black;"><strong>GGACHI</strong></td>
|
| 168 |
+
<td style="border:1px solid black;"><strong>0.821</strong></td>
|
| 169 |
+
</tr>
|
| 170 |
+
</tbody>
|
| 171 |
+
</table>
|