llama3-8b-spaetzle-v20

llama3-8b-spaetzle-v20 is an int4-inc (intel auto-round) quantized merge of the following models:

Benchmarks

The GGUF q4_k_m version achieves on EQ-Bench v2_de 65.7 (171/171 parseable). From Intel's low bit open llm leaderboard:

Type	Model	Average ⬆️	ARC-c	ARC-e	Boolq	HellaSwag	Lambada	MMLU	Openbookqa	Piqa	Truthfulqa	Winogrande	#Params (B)	#Size (G)
🍒	cstr/llama3-8b-spaetzle-v20-int4-inc	66.43	61.77	85.4	82.75	62.79	71.73	64.17	37.4	80.41	43.21	74.66	7.04	5.74

🧩 Configuration

models:
  - model: cstr/llama3-8b-spaetzle-v13
    # no parameters necessary for base model
  - model: nbeerbower/llama-3-wissenschaft-8B-v2
    parameters:
      density: 0.65
      weight: 0.4        
merge_method: dare_ties
base_model: cstr/llama3-8b-spaetzle-v13
parameters:
  int8_mask: true
dtype: bfloat16
random_seed: 0
tokenizer_source: base

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "cstr/llama3-8b-spaetzle-v20"
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"])