MT-Gen4_gemma-3-12B_flatten

---

EN

Merge to try to align the model results on the UGI leaderboard, to create the source for the modifications.

In general, just a tentative description.

---

RU

Объединение, чтобы попытаться выровнять результаты модели в таблице лидеров UGI, для создания исходника для модификаций.

В целом просто пробное описание.

---

MT-Gen4_gemma-3-12B_flatten

MT-Gen4_gemma-3-12B_flatten is a merge of the following models using LazyMergekit:

• zelk12/26_05_2025_Test_LazyMergekit_gemma-3-12B
• zelk12/MT4-gemma-3-12B

🧩 Configuration

models:
  - model: zelk12/26_05_2025_Test_LazyMergekit_gemma-3-12B
    #no parameters necessary for base model
  - model: zelk12/26_05_2025_Test_LazyMergekit_gemma-3-12B
    parameters:
      density: 0.438
      weight: 0.561

  - model: zelk12/MT4-gemma-3-12B
    parameters:
      density: 0.561
      weight: 0.438

merge_method: dare_ties
base_model: zelk12/26_05_2025_Test_LazyMergekit_gemma-3-12B
parameters:
  normalize: true
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "zelk12/MT-Gen4_gemma-3-12B_flatten"
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"])