MT7-Gen3_gemma-3-12B

MT7-Gen3_gemma-3-12B

MT7-Gen3_gemma-3-12B is a merge of the following models using LazyMergekit:

• IlyaGusev/saiga_gemma3_12b
• zelk12/MT1-gemma-3-12B
• soob3123/amoral-gemma3-12B-v2
• zelk12/MT-Gen1-gemma-3-12B
• zelk12/MT-gemma-3-12B

🧩 Configuration

models:
  - model: TheDrummer/Fallen-Gemma3-12B-v1
    #no parameters necessary for base model
  - model: IlyaGusev/saiga_gemma3_12b
    parameters:
      density: 0.8
      weight: 0.8

  - model: zelk12/MT1-gemma-3-12B
    parameters:
      density: 0.792
      weight: 0.792

  - model: soob3123/amoral-gemma3-12B-v2
    parameters:
      density: 0.684
      weight: 0.684

  - model: zelk12/MT-Gen1-gemma-3-12B
    parameters:
      density: 0.518
      weight: 0.518

  - model: zelk12/MT-gemma-3-12B
    parameters:
      density: 0.5
      weight: 0.5

merge_method: dare_ties
base_model: TheDrummer/Fallen-Gemma3-12B-v1
parameters:
  normalize: true
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "zelk12/MT7-Gen3_gemma-3-12B"
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"])