File size: 2,660 Bytes
ca2c22c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cf187b1
 
 
ca2c22c
 
 
 
 
 
 
 
 
cf187b1
ca2c22c
3168332
 
 
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
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
---
language:
- ja
tags:
- japanese-stablelm
- causal-lm
pipeline_tag: text-generation
base_model: stabilityai/japanese-stablelm-base-gamma-7b
license: apache-2.0
extra_gated_fields:
  Name: text
  Email: text
  Country: text
  Organization or Affiliation: text
  I allow Stability AI to contact me about information related to its models and research: checkbox
---


# Reproduced Japanese Stable LM Instruct Gamma 7B

## Model Description

This is a reproduction of 7B-parameter decoder-only Japanese language model fine-tuned on instruction-following datasets, built on top of the base model [Japanese Stable LM Base Gamma 7B](https://huggingface.co/stabilityai/japanese-stablelm-base-gamma-7b).

This model is trained with [notus](https://github.com/argilla-io/notus) code base.

*If you are in search of the official model, please check [Japanese Stable LM Instruct Gamma 7B](https://huggingface.co/stabilityai/japanese-stablelm-instruct-gamma-7b).*



## Model Details

### Training Datasets

- [Japanese translation of the Databricks Dolly-15k dataset](https://huggingface.co/datasets/kunishou/databricks-dolly-15k-ja)
- [Japanese translation of the subset of the Anthropic HH dataset](https://huggingface.co/datasets/fujiki/japanese_hh-rlhf-49k)
- [Wikinews](https://ja.wikinews.org/wi) [subset](https://huggingface.co/datasets/fujiki/llm-japanese-dataset_wikinews) of the [izumi-lab/llm-japanese-dataset](https://huggingface.co/datasets/izumi-lab/llm-japanese-dataset)


### Benchmarks

The result is evaluated by [Nejumi-leaderboard Neo](https://github.com/wandb/llm-leaderboard/tree/b2723944d4955768cb93c18ffe162a8ff4e88955).

- llm-jp-eval:

  |AVG   |EL |FA  |MC   |MR  |NLI   |QA     |RC     |chabsa|jamp |janli|jcommonsenseqa|jemhopqa|jnli |jsem |jsick|jsquad |mawps |niilc |wiki_coreference|wiki_dependency|wiki_ner|wiki_pas|wiki_reading|
  |------|---|----|-----|----|------|-------|-------|------|-----|-----|--------------|--------|-----|-----|-----|-------|------|------|----------------|---------------|--------|--------|------------|
  |0.26  |0  |0.14|0.27 |0.1 |0.302 |0.2619 |0.7464 |0.0   |0.15 |0.5  |0.27          |0.2528  |0.04 |0.67 |0.15 |0.7464 |0.1   |0.271 |0.0             |0.0            |0.0     |0.0     |0.7         |


- Japanese Mt-Bench:

  |coding|extraction|humanities|math|reasoning|roleplay|stem|writing|
  |------|----------|----------|----|---------|--------|----|-------|
  |1.3   |1.75      |2.35      |1.45|3.4      |5.8     |4.3 |3.1    |


- Overall Average: 0.283125


## Credits
The training was carried out by [Hwigeon Oh](https://huggingface.co/ohwi) and [Fujiki Nakamura](https://huggingface.co/fujiki).