Update README.md
#1
by Doleeee - opened
README.md
CHANGED
|
@@ -17,38 +17,6 @@ Hunminai-1.0 is a Korean-aligned language model based on [Google's Gemma-3](http
|
|
| 17 |
- **Model Type**: Text Generation
|
| 18 |
- **Fine-Tuning Techniques**: Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO)
|
| 19 |
|
| 20 |
-
## Usage Application Form
|
| 21 |
-
To use this model, please complete the application form and submit it via email [davidkim205@gmail.com].
|
| 22 |
-
Access will be granted after your application is reviewed and approved.
|
| 23 |
-
We appreciate your cooperation and look forward to assisting you.
|
| 24 |
-
```
|
| 25 |
-
1. **Name:**
|
| 26 |
-
- (e.g., John Doe)
|
| 27 |
-
2. **Date of Birth:**
|
| 28 |
-
- (e.g., January 1, 1990)
|
| 29 |
-
3. **Affiliation:**
|
| 30 |
-
- Are you applying as a company or an individual? [ ] Company [ ] Individual
|
| 31 |
-
- Company Name (if applicable):
|
| 32 |
-
- Department (if applicable):
|
| 33 |
-
4. **Position/Role:**
|
| 34 |
-
- (e.g., Data Scientist, Researcher, etc.)
|
| 35 |
-
5. **Contact Information:**
|
| 36 |
-
- Email:
|
| 37 |
-
- Phone Number:
|
| 38 |
-
|
| 39 |
-
6. **Purpose of Use:**
|
| 40 |
-
- (e.g., Research and Development, Commercial use, Educational purposes, etc.)
|
| 41 |
-
|
| 42 |
-
7. **Detailed Reason for Use:**
|
| 43 |
-
- 1. Name and version of the model you wish to use:
|
| 44 |
-
- 2. Reason for selecting this model:
|
| 45 |
-
- 3. Objectives to achieve using this model:
|
| 46 |
-
- 4. Expected use cases (please describe in as much detail as possible):
|
| 47 |
-
|
| 48 |
-
8. **Data Security and Ethical Use Plan:**
|
| 49 |
-
- (Please describe your plans for data protection and ethical use.)
|
| 50 |
-
```
|
| 51 |
-
|
| 52 |
## Usage
|
| 53 |
|
| 54 |
Gemma 3 is supported starting from version 4.50.0 of the Transformers library.
|
|
@@ -66,7 +34,7 @@ Install the required package and run the example code below to load the Hunminai
|
|
| 66 |
from transformers import AutoProcessor, Gemma3ForConditionalGeneration
|
| 67 |
import torch
|
| 68 |
|
| 69 |
-
model_id = "davidkim205/Hunminai-
|
| 70 |
|
| 71 |
model = Gemma3ForConditionalGeneration.from_pretrained(
|
| 72 |
model_id, device_map="auto"
|
|
@@ -132,20 +100,20 @@ The table below contains a description of the Korean LLM evaluation benchmark da
|
|
| 132 |
|
| 133 |
### Benchmark Results
|
| 134 |
|
| 135 |
-
| |
|
| 136 |
-
|---------|----------------------------------:|-------------------------:|--------------------------:|-----------------------------------------:|----------------------------------------:|
|
| 137 |
-
| Avg. |
|
| 138 |
-
| bench |
|
| 139 |
-
| bench2 |
|
| 140 |
-
| ged |
|
| 141 |
-
| ged2 |
|
| 142 |
-
| tiny |
|
| 143 |
-
| ifeval |
|
| 144 |
-
| ged:E |
|
| 145 |
-
| ged:M |
|
| 146 |
-
| ged:H |
|
| 147 |
-
| ged2:E |
|
| 148 |
-
| ged2:M |
|
| 149 |
-
| ged2:H |
|
| 150 |
-
| gpqa |
|
| 151 |
-
| math500 |
|
|
|
|
| 17 |
- **Model Type**: Text Generation
|
| 18 |
- **Fine-Tuning Techniques**: Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO)
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
## Usage
|
| 21 |
|
| 22 |
Gemma 3 is supported starting from version 4.50.0 of the Transformers library.
|
|
|
|
| 34 |
from transformers import AutoProcessor, Gemma3ForConditionalGeneration
|
| 35 |
import torch
|
| 36 |
|
| 37 |
+
model_id = "davidkim205/Hunminai-1.0-12b"
|
| 38 |
|
| 39 |
model = Gemma3ForConditionalGeneration.from_pretrained(
|
| 40 |
model_id, device_map="auto"
|
|
|
|
| 100 |
|
| 101 |
### Benchmark Results
|
| 102 |
|
| 103 |
+
| | **davidkim205<br>ko-gemma<br>-3-12b** | google<br>gemma-3<br>-12b-it | unsloth<br>gemma-3<br>-12b-it | K-intelligence<br>Midm-2.0<br>-Base-Instruct | LGAI-EXAONE<br>EXAONE-3.5<br>-7.8B-Instruct |
|
| 104 |
+
|---------|---------------------------------------:|-----------------------------:|-------------------------------:|----------------------------------------------:|---------------------------------------------:|
|
| 105 |
+
| Avg. | **8.26** | 8.22 | 8.20 | 8.12 | 7.85 |
|
| 106 |
+
| bench | 7.96 | 8.00 | 7.83 | **8.01** | 7.70 |
|
| 107 |
+
| bench2 | 8.39 | 8.23 | **8.44** | 8.21 | 8.01 |
|
| 108 |
+
| ged | 8.65 | 8.61 | **8.73** | 8.10 | 8.25 |
|
| 109 |
+
| ged2 | 8.17 | 8.17 | 8.31 | **8.84** | 8.06 |
|
| 110 |
+
| tiny | **8.33** | **8.33** | 7.88 | 8.25 | 8.12 |
|
| 111 |
+
| ifeval | **8.37** | 8.30 | 8.33 | 8.24 | 6.76 |
|
| 112 |
+
| ged:E | **9.72** | **9.72** | 9.51 | **9.72** | 9.65 |
|
| 113 |
+
| ged:M | **9.63** | 9.55 | 9.39 | 9.31 | 9.10 |
|
| 114 |
+
| ged:H | 9.32 | 9.36 | 9.24 | **9.48** | 9.00 |
|
| 115 |
+
| ged2:E | 9.60 | 9.60 | **9.66** | 9.60 | 9.48 |
|
| 116 |
+
| ged2:M | 9.37 | **9.54** | **9.54** | 9.16 | 8.95 |
|
| 117 |
+
| ged2:H | **9.32** | 9.24 | 9.24 | 9.28 | 8.84 |
|
| 118 |
+
| gpqa | **3.18** | 2.88 | 2.98 | 2.68 | 3.13 |
|
| 119 |
+
| math500 | 5.60 | 5.58 | **5.70** | 4.80 | 4.88 |
|