---
license: apache-2.0
datasets:
- Minami-su/Amara-o2-dataset
---

<div align="left">
  <img src="/static-proxy?url=https%3A%2F%2Fcdn-uploads.huggingface.co%2Fproduction%2Fuploads%2F62d7f90b102d144db4b4245b%2FWIGeEeS5gBvATkSf2GSB-.png" 
       alt="Model Illustration" 
       style="width:50%; max-width:none;" />
</div>
 <blockquote> “何が綴られていたのか、私たちの文明では到底理解できない” <br/> （所阐述的内容超出了我们文明的理解范围） <br/> — sasakure.UK </blockquote>
                  
# How to use


迭代基于Amara-o1-7B-Qwen


```python

# Use a pipeline as a high-level helper

from transformers import pipeline

messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe = pipeline("text-generation", model="Minami-su/Amara-o2-7B-Qwen")
pipe(messages)


# Load model directly

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Minami-su/Amara-o2-7B-Qwen")
model = AutoModelForCausalLM.from_pretrained("Minami-su/Amara-o2-7B-Qwen")
```

####  Open Ended Generation Evaluation

<div align="left">



| Model | Arena-Hard | AlpacaEval 2.0 |
|-------|------------|----------------|
| DeepSeek-V2.5-0905 | 76.2 | 50.5 |
| Qwen2.5-72B-Instruct | 81.2 | 49.1 |
| LLaMA-3.1 405B | 69.3 | 40.5 |
| Amara-o1-7B-Qwen | ? | 42.12 |
| **Amara-o2-7B-Qwen** | ? | **51.33** |
| GPT-4o-0513 | 80.4 | 51.1 |
| Claude-Sonnet-3.5-1022 | 85.2 | 52.0 |
| DeepSeek-V3 | **85.5** | **70.0** |

Note: English open-ended conversation evaluations. For AlpacaEval 2.0, we use the length-controlled win rate as the metric.
</div>