File size: 1,203 Bytes
be68c8b
 
 
 
 
 
ef78c6a
 
be68c8b
 
e9fe295
e30b698
 
 
 
 
 
 
21b9e3b
e9fe295
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
---
license: llama2
datasets:
- hotchpotch/jaqket_v1_qa_wikija_context
language:
- ja
pipeline_tag: text-generation
inference: false
---


# お知らせ

より回答が適切になるように学習させたモデル、https://huggingface.co/hotchpotch/youri-7b-stf-qa-context-jaqket-jsquad-gptq もあります。

---


[youri-7b-instruction](https://huggingface.co/rinna/youri-7b-instruction) を元に SFT で Q&A + コンテキスト形式を学習し、AutoGTPQで量子化したモデルです。詳しくは以下を参照ください。

- [Q&A + RAG に特化したLLMをSFTで学習させ4bit量子化モデルを作り、GPT3.5以上の性能を7Bモデルで達成する](https://secon.dev/entry/2023/12/15/080000-qa-rag-llm-sft/)
- [学習コードや評価コード、サンプルコードなど](https://github.com/hotchpotch/youri-7b-sft-qa-context-jaqket/)
- [hotchpotch/youri-7b-sft-qa-context-jaqket-gptq](https://huggingface.co/hotchpotch/youri-7b-sft-qa-context-jaqket-gptq)
  - AutoGPTQ で量子化したモデル
- [hotchpotch/youri-7b-sft-qa-context-jaqket-awq](https://huggingface.co/hotchpotch/youri-7b-sft-qa-context-jaqket-awq)
  - AutoAWQ で量子化したモデル