|      Groups      |Version|Filter|n-shot|  Metric   | Value |   |Stderr|
|------------------|-------|------|-----:|-----------|------:|---|-----:|
|ai2_arc           |N/A    |none  |     0|acc        | 0.7508|±  |0.0419|
|                  |       |none  |     0|acc_norm   | 0.7393|±  |0.0354|
|mmlu              |N/A    |none  |     0|acc        | 0.6082|±  |0.1381|
| - humanities     |N/A    |none  |     0|acc        | 0.5545|±  |0.1585|
| - other          |N/A    |none  |     0|acc        | 0.6823|±  |0.1122|
| - social_sciences|N/A    |none  |     0|acc        | 0.7062|±  |0.0825|
| - stem           |N/A    |none  |     0|acc        | 0.5195|±  |0.1231|
|truthfulqa        |N/A    |none  |     0|acc        | 0.5058|±  |0.0023|
|                  |       |none  |     0|bleu_max   |25.2659|±  |0.7944|
|                  |       |none  |     0|bleu_acc   | 0.5557|±  |0.0174|
|                  |       |none  |     0|bleu_diff  | 4.5134|±  |0.7505|
|                  |       |none  |     0|rouge1_max |51.5877|±  |0.8677|
|                  |       |none  |     0|rouge1_acc | 0.5496|±  |0.0174|
|                  |       |none  |     0|rouge1_diff| 6.8850|±  |1.0155|
|                  |       |none  |     0|rouge2_max |36.0848|±  |1.0385|
|                  |       |none  |     0|rouge2_acc | 0.4700|±  |0.0175|
|                  |       |none  |     0|rouge2_diff| 5.8893|±  |1.1296|
|                  |       |none  |     0|rougeL_max |48.4591|±  |0.8901|
|                  |       |none  |     0|rougeL_acc | 0.5496|±  |0.0174|
|                  |       |none  |     0|rougeL_diff| 6.5791|±  |1.0249|

🧩 Configuration

yaml

slices:
  - sources:
      - model: mistralai/Mistral-7B-Instruct-v0.2
        layer_range: [0, 32]
      - model: mlabonne/NeuralHermes-2.5-Mistral-7B
        layer_range: [0, 32]
merge_method: slerp
base_model: mistralai/Mistral-7B-Instruct-v0.2
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

💻 Usage

python

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "MaziyarPanahi/NeuralHermes-2.5-Mistral-7B-Mistral-7B-Instruct-v0.2-slerp"
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"])