|
--- |
|
license: apache-2.0 |
|
language: |
|
- ja |
|
pipeline_tag: text-generation |
|
tags: |
|
- Mistral |
|
--- |
|
|
|
# Japanese-WizardLM2-ChatV-7B-GGUF |
|
GGUF conversion of "Japanese-WizardLM2-ChatV-7B" |
|
|
|
This model, Japanese-WizardLM2-ChatV-7B, is based on "chatntq-ja-7b-v1.0 ", and was created by subtracting "Mistral-7B-v0.1" from "WizardLM-2-7b" ChatVector was added by a factor of 1.0. |
|
We aimed to add the high performance of WizardLM-2 to the Japanese language capability of ChatNTQ. |
|
|
|
このモデル、Japanese-WizardLM2-ChatV-7Bは、”chatntq-ja-7b-v1.0”をベースに、"WizardLM-2-7b"から"Mistral-7B-v0.1"を差し引いて作ったChatVectorを1.0倍で足しました。 |
|
ChatNTQの日本語能力にWizardLM-2の性能の高さが加わる事を狙いました。 |
|
|
|
### Performance |
|
|
|
<table> |
|
<tr> |
|
<th>Model</th> |
|
<th><a href="https://huggingface.co/CohereForAI/c4ai-command-r-plus">c4ai-command-r-plus(Cohere API)</a></th> |
|
<th><a href="https://huggingface.co/TFMC/Japanese-Starling-ChatV-7B-GGUF">JA-Starling-ChatV-7B-GGUF(fp16)</th> |
|
<th>JA-WizardLM2-ChatV-7B-GGUF (This model)(fp16)</th> |
|
<th><a href="https://huggingface.co/TFMC/ChatNTQ-JA-7b-v1.0-GGUF">ChatNTQ-JA-7b-v1.0-GGUF(Q8_0)</a></th> |
|
</tr> |
|
<tr> |
|
<td>Parameters</td> |
|
<td>104B</td> |
|
<td>7B(Mistral)</td> |
|
<td>7B(Mistral)</td> |
|
<td>7B(Mistral)</td> |
|
</tr> |
|
<tr> |
|
<td>ELYZAtasks100<br>average score</td> |
|
<td>4.04</td> |
|
<td>3.77</td> |
|
<td>3.40</td> |
|
<td>2.74</td> |
|
</tr> |
|
</table> |
|
|
|
This score was evaluated by Cohere API command-r-plus using the "<a href="https://huggingface.co/datasets/elyza/ELYZA-tasks-100">ELYZA-tasks-100</a>", a Japanese model with instruction-tuning. |
|
|
|
このスコアはinstruction-tuningを行った日本語モデルのベンチマーク「ELYZA-tasks-100」を使い、Cohere APIのcommand-r-plusにより評価させたものです。 |
|
|
|
### Prompt Template |
|
|
|
- Llama-2-Chat |
|
<pre><code>[INST] <<SYS>>\nあなたは役に立つアシスタントです。\n<</SYS>>\n\n{prompt} [/INST]</code></pre> |
|
|
