Qwerus-7B
Qwerus-7B is a merge of the following models using LazyMergekit:
Benchmark on reasoning tasks using lighteval:
| Task | Version | Metric | Value | Stderr | |
|---|---|---|---|---|---|
| aime24 | 1 | extractive_match | 0.1333 | ± | 0.0631 |
| math_500 | 1 | extractive_match | 0.7420 | ± | 0.0196 |
In comparison, Qwen2.5-7B-Instruct:
| Task | Version | Metric | Value | Stderr | |
|---|---|---|---|---|---|
| aime24 | 1 | extractive_match | 0.1667 | ± | 0.0692 |
| math_500 | 1 | extractive_match | 0.8220 | ± | 0.0171 |
🧩 Configuration
models:
- model: Qwen/Qwen2.5-7B
# No parameters necessary for base model
- model: PRIME-RL/Eurus-2-7B-PRIME
parameters:
density: 0.56
weight: 0.5
- model: Qwen/Qwen2.5-7B-Instruct
parameters:
density: 0.56
weight: 0.5
merge_method: dare_ties
base_model: Qwen/Qwen2.5-7B
dtype: bfloat16
💻 Usage
!pip install -qU transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "mlabonne/Qwerus-7B"
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"])