llama3-8b-spaetzle-v13

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

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

πŸ’» 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"])
Downloads last month
48
Safetensors
Model size
8.03B params
Tensor type
BF16
Β·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for cstr/llama3-8b-spaetzle-v13

Finetuned
(1)
this model

Spaces using cstr/llama3-8b-spaetzle-v13 6

Collection including cstr/llama3-8b-spaetzle-v13