README.md

metadata

tags:
  - merge
  - mergekit
  - lazymergekit
  - mlabonne/ChimeraLlama-3-8B-v2
  - nbeerbower/llama-3-stella-8B
  - uygarkurt/llama-3-merged-linear
base_model:
  - mlabonne/ChimeraLlama-3-8B-v2
  - nbeerbower/llama-3-stella-8B
  - uygarkurt/llama-3-merged-linear
license: other

NeuralLLaMa-3-8b-DT-v0.1

NeuralLLaMa-3-8b-DT-v0.1 is a merge of the following models using LazyMergekit:

• mlabonne/ChimeraLlama-3-8B-v2
• nbeerbower/llama-3-stella-8B
• uygarkurt/llama-3-merged-linear

🧩 Configuration

models:
  - model: NousResearch/Meta-Llama-3-8B
    # No parameters necessary for base model
  - model: mlabonne/ChimeraLlama-3-8B-v2
    parameters:
      density: 0.33
      weight: 0.2
  - model: nbeerbower/llama-3-stella-8B
    parameters:
      density: 0.44
      weight: 0.4
  - model: uygarkurt/llama-3-merged-linear
    parameters:
      density: 0.55
      weight: 0.4
merge_method: dare_ties
base_model: NousResearch/Meta-Llama-3-8B
parameters:
  int8_mask: true
dtype: float16

🗨️ Chats

💻 Usage

!pip install -qU transformers accelerate bitsandbytes

from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer, BitsAndBytesConfig
import torch

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16
)

MODEL_NAME = 'Kukedlc/NeuralLLaMa-3-8b-DT-v0.1'
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, device_map='cuda:0', quantization_config=bnb_config)

prompt_system = "You are an advanced language model that speaks Spanish fluently, clearly, and precisely.\
You are called Roberto the Robot and you are an aspiring post-modern artist."
prompt = "Create a piece of art that represents how you see yourself, Roberto, as an advanced LLm, with ASCII art, mixing diagrams, engineering and let yourself go."

chat = [
    {"role": "system", "content": f"{prompt_system}"},
    {"role": "user", "content": f"{prompt}"},
]

chat = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(chat, return_tensors="pt").to('cuda')
streamer = TextStreamer(tokenizer)
stop_token = "<|eot_id|>"
stop = tokenizer.encode(stop_token)[0]

_ = model.generate(**inputs, streamer=streamer, max_new_tokens=1024, do_sample=True, temperature=0.7, repetition_penalty=1.2, top_p=0.9, eos_token_id=stop)