You can deploy this model and start earning money today!

RichardErkhov/
cstr_-_llama3-8b-spaetzle-v20-gguf

Name: cstr_-_llama3-8b-spaetzle-v20-gguf
Availability: InStock

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

llama3-8b-spaetzle-v20 - GGUF

Model creator: https://huggingface.co/cstr/
Original model: https://huggingface.co/cstr/llama3-8b-spaetzle-v20/

Name	Quant method	Size
llama3-8b-spaetzle-v20.Q2_K.gguf	Q2_K	2.96GB
llama3-8b-spaetzle-v20.IQ3_XS.gguf	IQ3_XS	3.28GB
llama3-8b-spaetzle-v20.IQ3_S.gguf	IQ3_S	3.43GB
llama3-8b-spaetzle-v20.Q3_K_S.gguf	Q3_K_S	3.41GB
llama3-8b-spaetzle-v20.IQ3_M.gguf	IQ3_M	3.52GB
llama3-8b-spaetzle-v20.Q3_K.gguf	Q3_K	3.74GB
llama3-8b-spaetzle-v20.Q3_K_M.gguf	Q3_K_M	3.74GB
llama3-8b-spaetzle-v20.Q3_K_L.gguf	Q3_K_L	4.03GB
llama3-8b-spaetzle-v20.IQ4_XS.gguf	IQ4_XS	4.18GB
llama3-8b-spaetzle-v20.Q4_0.gguf	Q4_0	4.34GB
llama3-8b-spaetzle-v20.IQ4_NL.gguf	IQ4_NL	4.38GB
llama3-8b-spaetzle-v20.Q4_K_S.gguf	Q4_K_S	4.37GB
llama3-8b-spaetzle-v20.Q4_K.gguf	Q4_K	4.58GB
llama3-8b-spaetzle-v20.Q4_K_M.gguf	Q4_K_M	4.58GB
llama3-8b-spaetzle-v20.Q4_1.gguf	Q4_1	4.78GB
llama3-8b-spaetzle-v20.Q5_0.gguf	Q5_0	5.21GB
llama3-8b-spaetzle-v20.Q5_K_S.gguf	Q5_K_S	5.21GB
llama3-8b-spaetzle-v20.Q5_K.gguf	Q5_K	5.34GB
llama3-8b-spaetzle-v20.Q5_K_M.gguf	Q5_K_M	5.34GB
llama3-8b-spaetzle-v20.Q5_1.gguf	Q5_1	5.65GB
llama3-8b-spaetzle-v20.Q6_K.gguf	Q6_K	6.14GB
llama3-8b-spaetzle-v20.Q8_0.gguf	Q8_0	7.95GB

Original model description:

tags:

merge
mergekit
lazymergekit
nbeerbower/llama-3-wissenschaft-8B-v2 base_model:
nbeerbower/llama-3-wissenschaft-8B-v2 license: llama3 language:
en
de

llama3-8b-spaetzle-v20

llama3-8b-spaetzle-v20 is a merge of the following models:

Benchmarks

On EQ-Bench v2_de it achieves 65.7 (171/171 parseable). From Open LLM Leaderboard (details):

Model	Average	ARC	HellaSwag	MMLU	TruthfulQA	Winogrande	GSM8K
cstr/llama3-8b-spaetzle-v20	71.83	70.39	85.69	68.52	60.98	78.37	67.02

🧩 Configuration

yaml

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

python

!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"])

RichardErkhov/cstr_-_llama3-8b-spaetzle-v20-gguf