Qwen3-8B-mix

Qwen3-8B-mix is a merge of the following models using LazyMergekit:

• soob3123/GrayLine-Qwen3-8B
• mlxha/Qwen3-8B-grpo-medmcqa
• ValiantLabs/Qwen3-8B-Esper3

🧩 Configuration

slices:
  - sources:
      - model: soob3123/GrayLine-Qwen3-8B
        layer_range: [0, 36]
      - model: mlxha/Qwen3-8B-grpo-medmcqa
        layer_range: [0, 36]
      - model: ValiantLabs/Qwen3-8B-Esper3
        layer_range: [0, 36]
merge_method: multislerp
base_model: soob3123/GrayLine-Qwen3-8B
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]
    - filter: down_proj
      value: 0.5
    - value: 0.5
  weight:  # 添加一个名为 'weight' 的参数
    - filter: embed_tokens  # 使用 filter 来匹配 embed_tokens 层
      value: 0.5  # 为 embed_tokens.weight 指定一个合并权重
    - value: 0.5 # 可选：为其他没有匹配 filter 的层设置默认权重
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "leegang/Qwen3-8B-mix"
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"])