hafidhsoekma/test-g8b-qwen3-ties-0

test-g8b-qwen3-linear

test-g8b-qwen3-linear is a merge of the following models using LazyMergekit:

• itdel/Gasing-8B-v0.1
• Qwen/Qwen3-8B

🧩 Configuration

merge_method: ties
base_model: itdel/Gasing-8B-v0.1
models:
  - model: itdel/Gasing-8B-v0.1
    parameters:
      density: 0.6  # Mempertahankan 60% parameter untuk kemampuan bahasa Indonesia
      weight: 0.7   # Bobot lebih tinggi untuk bahasa Indonesia
  - model: Qwen/Qwen3-8B
    parameters:
      density: 0.8  # Mempertahankan 80% parameter untuk kemampuan math
      weight: 0.8   # Bobot tinggi untuk kemampuan math
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16
tokenizer_source: union

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "hafidhsoekma/test-g8b-qwen3-linear"
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"])