llama3-8b-spaetzle-v13

llama3-8b-spaetzle-v13 is a merge of the following models:

• Azure99/blossom-v5-llama3-8b
• VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct

No change in llama3 prompt format template.

Benchmarks

It should work not too bad for German or English, e.g. it achieves 64.14(.10) on EQ Bench v2_de with 170(171)/171 per q4km GGUF (bpe fixed). And for English EQ-Bench Score (v2): 75.59, Parseable: 171.

Average	ARC	HellaSwag	MMLU	TruthfulQA	Winogrande	GSM8K
71.26	68.69	85.05	68.06	59.43	79.24	67.1

Sample output

>>> What is 7777 + 3333?
To find the sum, I'll add the two numbers together:

7777 + 3333 = 11110

So the answer is 11110.

>>> I have 10 apples. I find 3 gold coins in the bottom of a river. The river runs near a big city that has something to do with what I can spend the coins on
... . I then lose 4 apples but gain a gold coin. Three birds run into my path and drop 6 apples each. I play an online game and win 6 gold coins but I have to
...  share them equally with my 2 teammates. I buy apples for all the coins I have. The price of an apple is 0.5 coins. How many apples do I have? And where i
... s the river?
Let's break down what you've gained and lost:

* You found 3 gold coins in the river.
* You lost 4 apples, so let's subtract those from your initial 10: 10 - 4 = 6 apples left.
* Then you gained a gold coin, but that doesn't affect the number of apples.
* The birds drop 6 apples each, and there are three birds, so that's an extra 18 apples: 6 (per bird) * 3 (birds) = 18.
* You now have a total of 6 (initial apples) + 18 (from birds) = 24 apples.
* In the online game, you won 6 gold coins, but you had to share them equally with your teammates. That means each of you got 2 coins: 6 / 3 = 2 coins 
per person.
* You now have a total of 3 (initial coins) + 1 (new coin) + 2 (game winnings, shared) = 6 gold coins.
* With the 6 gold coins, you can buy apples at a rate of 0.5 coins per apple. To find out how many apples you can buy, divide your total coins by the 
cost per apple: 6 / 0.5 = 12 apples.
* Since you already have 24 apples, let's add the new ones to that: 24 + 12 = 36 apples.

So, you now have a total of 36 apples.

As for the river, it's not specified which city it runs near, so we can't pinpoint its exact location. But at least you found those coins!

>>> Two years ago, I was three times as old as my brother was. In three years, I will be twice as old as my brother. How old is each of us now? 
Let's define the current age of you and your brother with variables.

You are Y years old.
Your brother is X years old.

Two years ago:
You were Y - 2 years old (because we're looking back in time).
Your brother was X - 2 years old.

Since two years ago, you were three times as old as your brother was, the equation would be:

Y - 2 = 3(X - 2)

Simplify and solve for X:

Y - 2 = 3X - 6
Y = 3X - 4

Now we know that Y (your current age) is three times your brother's age minus four. Let's use this to find your current age.

In three years:
You will be Y + 3 years old.
Your brother will be X + 3 years old.

Since in three years, you'll be twice as old as your brother:

Y + 3 = 2(X + 3)

Simplify and solve for X (your brother's current age):

Y + 3 = 2X + 6
Y - 2X = 3
(3X - 4) - 2X = 3
X - 4 = 3
X = 7

Now that we have your brother's current age, we can find yours:

Y = 3X - 4
= 3(7) - 4
= 21 - 4
= 17

So you are currently 17 years old, and your brother is 7 years old.

🧩 Configuration

models:
  - model: VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
    # no parameters necessary for base model
  - model: Azure99/blossom-v5-llama3-8b
    parameters:
      density: 0.65
      weight: 0.4        
merge_method: dare_ties
base_model: VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct
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-v13"
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"])