---
license: apache-2.0
language: ja
library_name: transformers
tags:
  - continued-pretraining
  - language-model
model-index:
  - name: aokitools/japanese-laws-egov-instruct-202508182216
    results: []
---

# Experimental model in research stage


## Quickstart 

If you're using [Ollama](https://ollama.com/), run the following command first, then restart the Ollama app and select the newly added model.
```shell
ollama pull hf.co/aokitools/japanese-laws-egov-instruct-202508182216
```

If you want to remove it, run the following command:
```shell
ollama list
ollama rm hf.co/aokitools/japanese-laws-egov-instruct-202508182216:latest
ollama list
```


To use it from Python, use the following code.
```python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig

model_name = "aokitools/japanese-laws-egov-instruct-202508182216"

quant_config = BitsAndBytesConfig(
    load_in_8bit=True,
    llm_int8_threshold=6.0,
)

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto",
    quantization_config=quant_config,
)

# prepare the model input
prompt = "Give me a short introduction to large language model."
messages = [
	{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
	messages,
	tokenize=False,
	add_generation_prompt=True,
	enable_thinking=True # Switches between thinking and non-thinking modes. Default is True.
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# conduct text completion
generated_ids = model.generate(
	**model_inputs,
	max_new_tokens=256
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist() 

# parsing thinking content
try:
	# rindex finding 151668 (</think>)
	index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
	index = 0

thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")

print("thinking content:", thinking_content)
print("content:", content)
```


This model is a continual pretraining of [Qwen/Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B).

## Training details
- Base model: Qwen3-1.7B
- Tokenizer: QwenTokenizer

## License
- Apache 2.0 + Alibaba Qianwen License