README.md

metadata

license: apache-2.0
tags:
  - merge
  - mergekit
  - lazymergekit
  - abideen/DareVox-7B
  - udkai/Garrulus
language:
  - en

NexoNimbus-7B

NexoNimbus-7B is a merge of the following models:

• abideen/DareVox-7B
• udkai/Garrulus

🏆 Evaluation NexoNimbus-7B is the 5th best-performing 7B LLM on the Open LLM Leaderboard:

Task	Version	Metric	Value		Stderr
arc_challenge	0	acc	68.25	±	1.36
		acc_norm	70.81	±	1.38
hellaswag	0	acc	70.86	±	0.45
		acc_norm	87.86	±	0.32
gsm8k	0	acc	70.35	±	1.25
winogrande	0	acc	84.84	±	1.00
mmlu	0	acc	64.69	±	1.00

Average: 73.5%

TruthfulQA

Task	Version	Metric	Value		Stderr
truthfulqa_mc	1	mc1	46.26	±	1.74
		mc2	62.42	±	1.54

🧩 Configuration

slices:
  - sources:
      - model: abideen/DareVox-7B 
        layer_range: [0, 32]
      - model: udkai/Garrulus
        layer_range: [0, 32]
merge_method: slerp
base_model: abideen/DareVox-7B 
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "abideen/NexoNimbus-7B"
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"])