Downloads · 30 days
161
5% of all-time downloads
RichardErkhov/tesser-ai_-_Tesser-Llama-3-Ko-8B-gguf
tesser-ai_-_Tesser-Llama-3-Ko-8B-gguf is a machine learning model from RichardErkhov. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Downloads · 30 days
161
5% of all-time downloads
All-time downloads
3.2K
Public
Repo size
92.9 GB
Likes
0
Public
Click a slice to open those files.
.gguf108 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Tesser-Llama-3-Ko-8B - GGUF
| Name | Quant method | Size |
|---|---|---|
| Tesser-Llama-3-Ko-8B.Q2_K.gguf | Q2_K | 2.96GB |
| Tesser-Llama-3-Ko-8B.IQ3_XS.gguf | IQ3_XS | 3.28GB |
| Tesser-Llama-3-Ko-8B.IQ3_S.gguf | IQ3_S | 3.43GB |
| Tesser-Llama-3-Ko-8B.Q3_K_S.gguf | Q3_K_S | 3.41GB |
| Tesser-Llama-3-Ko-8B.IQ3_M.gguf | IQ3_M | 3.52GB |
| Tesser-Llama-3-Ko-8B.Q3_K.gguf | Q3_K | 3.74GB |
| Tesser-Llama-3-Ko-8B.Q3_K_M.gguf | Q3_K_M | 3.74GB |
| Tesser-Llama-3-Ko-8B.Q3_K_L.gguf | Q3_K_L | 4.03GB |
| Tesser-Llama-3-Ko-8B.IQ4_XS.gguf | IQ4_XS | 4.18GB |
| Tesser-Llama-3-Ko-8B.Q4_0.gguf | Q4_0 | 4.34GB |
| Tesser-Llama-3-Ko-8B.IQ4_NL.gguf | IQ4_NL | 4.38GB |
| Tesser-Llama-3-Ko-8B.Q4_K_S.gguf | Q4_K_S | 4.37GB |
| Tesser-Llama-3-Ko-8B.Q4_K.gguf | Q4_K | 4.58GB |
| Tesser-Llama-3-Ko-8B.Q4_K_M.gguf | Q4_K_M | 4.58GB |
| Tesser-Llama-3-Ko-8B.Q4_1.gguf | Q4_1 | 4.78GB |
| Tesser-Llama-3-Ko-8B.Q5_0.gguf | Q5_0 | 5.21GB |
| Tesser-Llama-3-Ko-8B.Q5_K_S.gguf | Q5_K_S | 5.21GB |
| Tesser-Llama-3-Ko-8B.Q5_K.gguf | Q5_K | 5.34GB |
| Tesser-Llama-3-Ko-8B.Q5_K_M.gguf | Q5_K_M | 5.34GB |
| Tesser-Llama-3-Ko-8B.Q5_1.gguf | Q5_1 | 5.65GB |
| Tesser-Llama-3-Ko-8B.Q6_K.gguf | Q6_K | 6.14GB |
| Tesser-Llama-3-Ko-8B.Q8_0.gguf | Q8_0 | 7.95GB |
language:
This model was continually pretrained from the Meta-Llama-3-8B, using English and Korean datasets. The goal is to enhance its proficiency in Korean while maintaining its English language capabilities from the original model.
We sampled 16B tokens from the following datasets for training:
<table> <tr> <td><strong>Sources</strong> </td> <td><strong>Tokens (Llama-3-8B)</strong> </td> </tr> <tr> <td>AI-Hub </td> <td>9.2B </td> </tr> <tr> <td>Modu Corpus </td> <td>5.8B </td> </tr> <tr> <td>Wikipedia </td> <td>5.4B </td> </tr> </table>This model has not been fine-tuned, so you will need to train it on your own dataset before using it.
We evaluated this model using both English and Korean benchmarks, and compared it with similar models that were continually pretrained from the Meta-Llama-3-8B.
<table> <tr> <td></td> <td colspan="4"><strong>English</strong></td> <td colspan="3"><strong>Korean</strong></td> </tr> <tr> <td><strong>Model</strong></td> <td><strong>MMLU (5 shots)</strong></td> <td><strong>HellaSwag (10 shots)</strong></td> <td><strong>GSM8K (8 shots, CoT)</strong></td> <td><strong>BBH (3 shots, CoT)</strong></td> <td><strong>KMMLU (5 shots)</strong></td> <td><strong>HAE-RAE (5 shots)</strong></td> <td><strong>KoBEST (5 shots)</strong></td> </tr> <tr> <td>meta-llama/Meta-Llama-3-8B</td> <td><strong>65.1</strong></td> <td><strong>82.1</strong></td> <td><strong>52.0</strong></td> <td><strong>61.9</strong></td> <td>40.2</td> <td>61.1</td> <td>69.2</td> </tr> <tr> <td>saltlux/Ko-Llama3-Luxia-8B</td> <td>57.1</td> <td>77.1</td> <td>32.3</td> <td>51.8</td> <td>39.4</td> <td>69.2</td> <td>71.9</td> </tr> <tr> <td>beomi/Llama-3-Open-Ko-8B</td> <td>56.2</td> <td>77.4</td> <td>31.5</td> <td>46.8</td> <td>40.3</td> <td>68.1</td> <td><u>72.1</u></td> </tr> <tr> <td>beomi/Llama-3-KoEn-8B</td> <td>52.5</td> <td>77.7</td> <td>21.2</td> <td>43.2</td> <td><u>40.8</u></td> <td><u>71.3</u></td> <td><strong>73.8</strong></td> </tr> <tr> <td><strong>tesser-ai/Tesser-Llama-3-Ko-8B</strong></td> <td><u>60.5</u></td> <td><u>79.8</u></td> <td><u>40.3</u></td> <td><u>56.3</u></td> <td><strong>42.5</strong></td> <td><strong>72.1</strong></td> <td><strong>73.8</strong></td> </tr> </table>We trained this model using a context length of 4k due to resource limitations and to maximize training speed. However, the original model was trained with a context length of 8k, so an 8k context length could work well in downstream tasks.
This model follows the original Llama-3 license.