ramonda-7b-dpo-ties

ramonda-7b-dpo-ties is a merge of the following models using LazyMergekit:

• paulml/OGNO-7B
• bardsai/jaskier-7b-dpo-v4.3

Benchmark

Open LLM Leaderboard

Model	Average	ARC	HellaSwag	MMLU	TruthfulQA	Winogrande	GSM8K
mayacinka/ramonda-7b-dpo-ties	76.19	72.7	89.69	64.5	77.17	84.77	68.92

LLM AutoEval

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
ramonda-7b-dpo-ties	44.67	77.16	77.6	49.06	62.12

🧩 Configuration

models:
  - model: bardsai/jaskier-7b-dpo-v5.6
    # no parameters necessary for base model
  - model: paulml/OGNO-7B
    parameters:
      density: 0.9
      weight: 0.5
  - model: bardsai/jaskier-7b-dpo-v4.3
    parameters:
      density: 0.5
      weight: 0.3
merge_method: ties
base_model: bardsai/jaskier-7b-dpo-v5.6
parameters:
  normalize: true
dtype: float16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "mayacinka/ramonda-7b-dpo-ties"
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.	76.19
AI2 Reasoning Challenge (25-Shot)	72.70
HellaSwag (10-Shot)	89.09
MMLU (5-Shot)	64.50
TruthfulQA (0-shot)	77.17
Winogrande (5-shot)	84.77
GSM8k (5-shot)	68.92