StrangeMerges_10-7B-slerp

StrangeMerges_10-7B-slerp is a merge of the following models using LazyMergekit:

• flemmingmiguel/MBX-7B-v3
• Gille/StrangeMerges_9-7B-dare_ties

🧩 Configuration

slices:
  - sources:
      - model: flemmingmiguel/MBX-7B-v3
        layer_range: [0, 32]
      - model: Gille/StrangeMerges_9-7B-dare_ties
        layer_range: [0, 32]
merge_method: slerp
base_model: Gille/StrangeMerges_9-7B-dare_ties
parameters:
  t:
    - filter: self_attn
      value: [0.1, 0.5, 0.3, 0.7, 0.9]
    - filter: mlp
      value: [0.9, 0.5, 0.7, 0.3, 0.1]
    - value: 0.5
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Gille/StrangeMerges_10-7B-slerp"
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"])

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	74.77
AI2 Reasoning Challenge (25-Shot)	72.35
HellaSwag (10-Shot)	88.30
MMLU (5-Shot)	64.87
TruthfulQA (0-shot)	69.49
Winogrande (5-shot)	83.50
GSM8k (5-shot)	70.13