Neural 4 QA 7b

🚀 Neural-4-QA-7b

Neural-4-QA-7b is a merged model that combines the strengths of multiple base models. It uses LazyMergekit to integrate various models, aiming to provide better performance in question - answering tasks.

🚀 Quick Start

Neural-4-QA-7b is a merge of the following models using LazyMergekit:

• yam-peleg/Experiment21-7B
• CultriX/NeuralTrix-bf16
• louisgrc/Montebello_7B_SLERP
• CorticalStack/pastiche-crown-clown-7b-dare-dpo
• chihoonlee10/T3Q-Mistral-Orca-Math-DPO

✨ Features

• Model Merging: Combines multiple high - quality base models to enhance performance.
• Flexible Configuration: Allows users to adjust parameters such as density and weight during the merging process.

📦 Installation

To use this model, you need to install the necessary libraries. You can use the following command:

!pip install -qU transformers accelerate

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer
import transformers
import torch

model = "Kukedlc/Neural-4-QA-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"])

📚 Documentation

🧩 Configuration

models:
  - model: chihoonlee10/T3Q-Mistral-Orca-Math-DPO
    # No parameters necessary for base model
  - model: yam-peleg/Experiment21-7B
    parameters:
      density: 0.66
      weight: 0.2
  - model: CultriX/NeuralTrix-bf16
    parameters:
      density: 0.55
      weight: 0.2
  - model: louisgrc/Montebello_7B_SLERP
    parameters:
      density: 0.55
      weight: 0.2
  - model: CorticalStack/pastiche-crown-clown-7b-dare-dpo
    parameters:
      density: 0.44
      weight: 0.2
  - model: chihoonlee10/T3Q-Mistral-Orca-Math-DPO
    parameters:
      density: 0.66
      weight: 0.2
merge_method: dare_ties
base_model: chihoonlee10/T3Q-Mistral-Orca-Math-DPO
parameters:
  int8_mask: true
dtype: bfloat16

📄 License

This model is released under the apache - 2.0 license.