KangalKhan-HardRuby-7B

KangalKhan-HardRuby-7B is a merge of the following models using LazyMergekit:

• Yuma42/KangalKhan-Ruby-7B-Fixed
• Yuma42/KangalKhan-RawRuby-7B

🧩 Configuration

slices:
  - sources:
      - model: Yuma42/KangalKhan-Ruby-7B-Fixed
        layer_range: [0, 32]
      - model: Yuma42/KangalKhan-RawRuby-7B
        layer_range: [0, 32]
merge_method: slerp
base_model: Yuma42/KangalKhan-Ruby-7B-Fixed
parameters:
  t:
    - filter: self_attn
      value: [0.1, 0.55, 0.35, 0.75, 0.97]
    - filter: mlp
      value: [0.9, 0.45, 0.65, 0.25, 0.03]
    - value: 0.5
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Yuma42/KangalKhan-HardRuby-7B"
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"])

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	68.65
AI2 Reasoning Challenge (25-Shot)	66.55
HellaSwag (10-Shot)	85.41
MMLU (5-Shot)	63.46
TruthfulQA (0-shot)	56.94
Winogrande (5-shot)	78.30
GSM8k (5-shot)	61.26