NeuralMonarchCoderPearlBeagle

NeuralMonarchCoderPearlBeagle is a merge of the following models using LazyMergekit:

• abideen/MonarchCoder-7B
• eldogbbhed/NeuralPearlBeagle

Goals

This is a TIES merge, formed from MonarchCoder-7b (A merge of Alpha Monarch and TessCoder) and NeuralPearlBeagle(which is a merge of mlabonne's NeuralBeagle14-7b and Pearl-7B-Slerp). It is a somewhat haphazard experiment to see if we can merge more math and coding capabilities into the already outstanding NeuralBeagle14-7b and still maintain the same positive chat abilities.

If you find this or my other merges useful, please consider sending a bit of BTC so I don't have to use Google Colab :D

BTC: bc1q8lc4mzdtdyz7fx44vaw3jn8qg6w4c3ypfxpdrv

ETH/POLYGON: 0x102a6fd187db8441d2cbead33ac70e87f382f114

🧩 Configuration

models:
  - model: abideen/MonarchCoder-7B
    parameters:
      density: 0.6
      weight: 0.5
  - model: eldogbbhed/NeuralPearlBeagle
    parameters:
      density: 0.8
      weight: 0.8
merge_method: ties
base_model: eldogbbhed/NeuralPearlBeagle
parameters:
  normalize: true
  int8_mask: true
dtype: float16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "eldogbbhed/NeuralMonarchCoderPearlBeagle"
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.	71.50
AI2 Reasoning Challenge (25-Shot)	68.52
HellaSwag (10-Shot)	87.22
MMLU (5-Shot)	64.53
TruthfulQA (0-shot)	61.19
Winogrande (5-shot)	80.51
GSM8k (5-shot)	67.02