README.md

---
base_model:
- tokyotech-llm/Swallow-MS-7b-v0.1
- mistralai/Mistral-7B-v0.1
- nitky/Flavor-7b
- stabilityai/japanese-stablelm-base-gamma-7b
library_name: transformers
tags:
- mergekit
- merge
language:
  - ja
  - en
pipeline_tag: text-generation
license: apache-2.0
---
# Oumuamua-7b-base

This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).

## Output example

### Input text
```
日本で最も高い山の名前は
```

### Output text
```
日本で最も高い山の名前は、富士山。
その標高は3776メートル。
世界でも20位以内に入る高さを誇る。
その富士山の麓にあるのが、静岡県富士市。
富士市は、富士山の麓にあるため、観光地としても有名である。
富士山の麓にあることから、富士市は観光地としても有名である。
富士山を眺めることができるスポットが多く、特に富士市の中心部から見る富士山は、その美しさから「日本一の眺望」と言われている。
```

## Test environment

This model was tested using [text-generation-webui](https://github.com/oobabooga/text-generation-webui/tree/main). I use preset `min_p` and `Null preset` with temperature=0.3 for Generation.


## Usage
### Use the base model

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "nitky/Oumuamua-7b-base"
tokenizer = AutoTokenizer.from_pretrained(model_name)

model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")
prompt = "日本で最も高い山の名前は"
input_ids = tokenizer.encode(
    prompt,
    add_special_tokens=False,
    return_tensors="pt"
)
tokens = model.generate(
    input_ids.to(device=model.device),
    max_new_tokens=256,
    do_sample=True,
    temperature=0.3
)

out = tokenizer.decode(tokens[0], skip_special_tokens=True)
print(out)
```

## Merge Details
### Merge Method

This model was merged using the [Model Stock](https://arxiv.org/abs/2403.19522) merge method using [tokyotech-llm/Swallow-MS-7b-v0.1](https://huggingface.co/tokyotech-llm/Swallow-MS-7b-v0.1) as a base.

### Models Merged

The following models were included in the merge:
* [tokyotech-llm/Swallow-MS-7b-v0.1](https://huggingface.co/tokyotech-llm/Swallow-MS-7b-v0.1)
* [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1)
* [nitky/Flavor-7b](https://huggingface.co/nitky/Flavor-7b)
* [stabilityai/japanese-stablelm-base-gamma-7b](https://huggingface.co/stabilityai/japanese-stablelm-base-gamma-7b)

### Configuration

The following YAML configuration was used to produce this model:

```yaml
merge_method: task_arithmetic
base_model: mistralai/Mistral-7B-v0.1
models:
  - model: tokyotech-llm/Swallow-MS-7b-v0.1
    parameters:
      weight:
      - filter: embed_tokens
        value: 1.0
      - value: 0
dtype: bfloat16
tokenizer_source: model:tokyotech-llm/Swallow-MS-7b-v0.1
name: Mistral-7B-v0.1-VE-Swallow-MS
---
merge_method: task_arithmetic
base_model: nitky/Flavor-7b # private model
models:
  - model: tokyotech-llm/Swallow-MS-7b-v0.1
    parameters:
      weight:
      - filter: embed_tokens
        value: 1.0
      - value: 0
dtype: bfloat16
tokenizer_source: model:tokyotech-llm/Swallow-MS-7b-v0.1
name: Flavor-7b-VE-Swallow-MS
---
merge_method: task_arithmetic
base_model: stabilityai/japanese-stablelm-base-gamma-7b
models:
  - model: tokyotech-llm/Swallow-MS-7b-v0.1
    parameters:
      weight:
      - filter: embed_tokens
        value: 1.0
      - value: 0
dtype: bfloat16
tokenizer_source: model:tokyotech-llm/Swallow-MS-7b-v0.1
name: japanese-stablelm-base-gamma-7b-VE-Swallow-MS
---
merge_method: task_arithmetic
base_model: Mistral-7B-v0.1-VE-Swallow-MS
models:
  - model: tokyotech-llm/Swallow-MS-7b-v0.1
    parameters:
      weight: 1.0
  - model: Flavor-7b-VE-Swallow-MS
    parameters:
      weight: 0.5
  - model: japanese-stablelm-base-gamma-7b-VE-Swallow-MS
    parameters:
      weight: -0.5
dtype: bfloat16
name: Oumuamua-7b-base-preset
---
merge_method: model_stock
base_model: Mistral-7B-v0.1-VE-Swallow-MS
models:
  - model: tokyotech-llm/Swallow-MS-7b-v0.1
  - model: Oumuamua-7b-base-preset
dtype: bfloat16
name: Oumuamua-7b-base
```