qwen3-4b-marged-lightning

qwen3-4b-marged-lightning

qwen3-4b-marged-lightning is a merge of the following models using LazyMergekit:

• fireworks1231/agentic-4b-sft
• fireworks1231/qwen3-4b-2-sfted-grpo-sft-h200-1200step

🧩 Configuration

models:
  - model: Qwen/Qwen3-4B
    # no parameters necessary for base model
  - model: fireworks1231/agentic-4b-sft
    parameters:
      density: 0.8
      weight: 0.3
  - model: fireworks1231/qwen3-4b-2-sfted-grpo-sft-h200-1200step
    parameters:
      density: 0.9
      weight: 0.7
merge_method: ties
base_model: Qwen/Qwen3-4B
parameters:
  normalize: true
dtype: float16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "fireworks1231/qwen3-4b-marged-lightning"
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"])