NeuralSynthesis-7B-v0.1

NeuralSynthesis-7B-v0.1 is a merge of the following models using LazyMergekit:

🧩 Configuration

models:
  - model: MiniMoog/Mergerix-7b-v0.3
  - model: automerger/Ognoexperiment27Multi_verse_model-7B
  - model: AurelPx/Percival_01-7b-slerp
  - model: automerger/YamshadowExperiment28-7B
merge_method: model_stock
base_model: automerger/YamshadowExperiment28-7B
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Kukedlc/NeuralSynthesis-7B-v0.1"
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.	76.80
AI2 Reasoning Challenge (25-Shot)	73.04
HellaSwag (10-Shot)	89.18
MMLU (5-Shot)	64.37
TruthfulQA (0-shot)	78.15
Winogrande (5-shot)	85.24
GSM8k (5-shot)	70.81