README.md

---
base_model:
- karakuri-ai/karakuri-lm-32b-thinking-2501-exp
- Qwen/Qwen2.5-Coder-32B-Instruct
tags:
- merge
- mergekit
- lazymergekit
- karakuri-ai/karakuri-lm-32b-thinking-2501-exp
- Qwen/Qwen2.5-Coder-32B-Instruct
license: apache-2.0
language:
- ja
- en
---

# Qwen2.5-Coder-32B-Instruct-karakuri-thinking-slerp

Qwen2.5-Coder-32B-Instruct-karakuri-thinking-slerpは、 [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing) を使った以下のモデルのマージです:
* [karakuri-ai/karakuri-lm-32b-thinking-2501-exp](https://huggingface.co/karakuri-ai/karakuri-lm-32b-thinking-2501-exp)
* [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct)

## 作成意図
日本語のReasoningモデルにコーディング能力を付与する目的で作成しました。

## 🧩 マージ設定

```yaml
slices:
  - sources:
      - model: karakuri-ai/karakuri-lm-32b-thinking-2501-exp
        layer_range: [0, 64]
      - model: Qwen/Qwen2.5-Coder-32B-Instruct
        layer_range: [0, 64]
merge_method: slerp
base_model: karakuri-ai/karakuri-lm-32b-thinking-2501-exp
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16
```

## 💻 使い方

```python
!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "smorce/Qwen2.5-Coder-32B-Instruct-karakuri-thinking-slerp"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
```