Qwen2-4x1.5B-v2.5.1-A

Qwen2-4x1.5B-v2.5.1-A is a Mixture of Experts (MoE) made with the following models using LazyMergekit:

• cognitivecomputations/dolphin-2.9.3-qwen2-1.5b
• Replete-AI/Qwen2-1.5b-Instruct-Replete-Adapted
• M4-ai/Hercules-5.0-Qwen2-1.5B
• d-llm/Qwen2-1.5B-Instruct-orpo

🧩 Configuration

gate_mode: hidden
architecture: qwen
dtype: bfloat16
experts_per_token: 2
base_model: cognitivecomputations/dolphin-2.9.3-qwen2-1.5b
experts:
  - source_model: cognitivecomputations/dolphin-2.9.3-qwen2-1.5b
    positive_prompts:
      - "You are an educator, provide in-depth explanations on academic topics."
      - "You are a tutor, offer guidance and support on complex subjects."
    negative_prompts:
      - "code"
      - "algorithm"
      - "programming"
  - source_model: Replete-AI/Qwen2-1.5b-Instruct-Replete-Adapted
    positive_prompts:
      - "You are a software developer, write code in various programming languages to solve complex problems."
      - "You are a programmer, design and implement algorithms to optimize system performance."
    negative_prompts:
      - "explain"
      - "describe"
      - "define"
  - source_model: M4-ai/Hercules-5.0-Qwen2-1.5B
    positive_prompts:
      - "You are a content creator, rephrase and reorganize text to improve clarity and coherence."
      - "You are a writer, generate engaging and informative content on a wide range of topics."
      - "You are a knowledge expert, provide general knowledge on history, science, literature, and more."
    negative_prompts:
      - "code"
      - "algorithm"
      - "programming"
  - source_model: d-llm/Qwen2-1.5B-Instruct-orpo
    positive_prompts:
      - "You are a summarizer, condense complex information into concise summaries."
      - "You are a translator, translate text from one language to another while preserving meaning and context."
    negative_prompts:
      - "explain"
      - "describe"
      - "define"
shared_experts:
  - source_model: cognitivecomputations/dolphin-2.9.3-qwen2-1.5b
    positive_prompts:
      - "You are a conversationalist, engage in natural-sounding conversations on a wide range of topics."
    negative_prompts:
      - "code"
      - "algorithm"
      - "programming"
    residual_scale: 0.1

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer, pipeline
import torch

model = "djuna/Qwen2-4x1.5B-v2.5.1-A"

tokenizer = AutoTokenizer.from_pretrained(model)
generator = pipeline(
    "text-generation",
    model=model,
    model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)

messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = generator(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])