File size: 1,719 Bytes
3d47d78
 
 
 
 
 
 
 
 
15e2ab0
3d47d78
 
15e2ab0
3d47d78
 
15e2ab0
744067b
 
 
c47f3ad
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
---
license: mit
language:
- ja
base_model:
- unsloth/DeepSeek-R1-Distill-Qwen-14B-GGUF
---
このモデルは [unsloth/DeepSeek-R1-Distill-Qwen-14B-unsloth-bnb-4bit](https://huggingface.co./unsloth/DeepSeek-R1-Distill-Qwen-14B-unsloth-bnb-4bit) を日本語で微調整したモデルです.  
This is japanese finetuned version of [unsloth/DeepSeek-R1-Distill-Qwen-14B-unsloth-bnb-4bit](https://huggingface.co./unsloth/DeepSeek-R1-Distill-Qwen-14B-unsloth-bnb-4bit).  

システムプロンプトに「あなたはDeepSeek R1、親切で有能な日本人アシスタントです。」を追加し、日本語データで微調整を実行しています。  
We added a system prompt "あなたはDeepSeek R1、親切で有能な日本人アシスタントです。" and finetuned model using Japanese data.  

元のモデルの思考プロセスは中国語や英語で出力される事が多いですが、このモデルの思考プロセスは日本語で出力される事が多いです。  
The thought process of the original model is often output in Chinese or English, but the thought process of this model is often output in Japanese.  

DeepSeek-R1は推奨温度が0.5-0.7である事に注意してください  
Please note that the recommended temperature for DeepSeek-R1 is 0.5-0.7  

私が試したいくつかの推論ツールはまだ不具合があるように見えます。[llama.cpp](https://huggingface.co./dahara1/DeepSeek-R1-Distill-Qwen-14B-unsloth-gguf-japanese-imatrix)をおすすめしておきます  
Some of the inference tools I've tried still seem to have bugs. I recommend [llama.cpp](https://huggingface.co./dahara1/DeepSeek-R1-Distill-Qwen-14B-unsloth-gguf-japanese-imatrix).