README.md

metadata

license: apache-2.0
tags:
  - merge
  - mergekit
  - lazymergekit
  - NousResearch/Hermes-2-Pro-Mistral-7B
  - instructlab/merlinite-7b-lab
base_model:
  - NousResearch/Hermes-2-Pro-Mistral-7B
  - instructlab/merlinite-7b-lab
model-index:
  - name: TriMistral-7B-TIES
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: AI2 Reasoning Challenge (25-Shot)
          type: ai2_arc
          config: ARC-Challenge
          split: test
          args:
            num_few_shot: 25
        metrics:
          - type: acc_norm
            value: 64.85
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Muhammad2003/TriMistral-7B-TIES
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: HellaSwag (10-Shot)
          type: hellaswag
          split: validation
          args:
            num_few_shot: 10
        metrics:
          - type: acc_norm
            value: 83.8
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Muhammad2003/TriMistral-7B-TIES
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU (5-Shot)
          type: cais/mmlu
          config: all
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 63.45
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Muhammad2003/TriMistral-7B-TIES
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: TruthfulQA (0-shot)
          type: truthful_qa
          config: multiple_choice
          split: validation
          args:
            num_few_shot: 0
        metrics:
          - type: mc2
            value: 56.47
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Muhammad2003/TriMistral-7B-TIES
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: Winogrande (5-shot)
          type: winogrande
          config: winogrande_xl
          split: validation
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 76.64
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Muhammad2003/TriMistral-7B-TIES
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GSM8k (5-shot)
          type: gsm8k
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 60.88
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Muhammad2003/TriMistral-7B-TIES
          name: Open LLM Leaderboard

TriMistral-7B-TIES

TriMistral-7B-TIES is a merge of the following models using LazyMergekit:

• NousResearch/Hermes-2-Pro-Mistral-7B
• instructlab/merlinite-7b-lab Special thanks to Charles Goddard for the quick implementation!

🧩 Configuration

models:
  - model: HuggingFaceH4/zephyr-7b-beta
    # no parameters necessary for base model
  - model: NousResearch/Hermes-2-Pro-Mistral-7B
    parameters:
      density: 0.5
      weight: 0.5
  - model: instructlab/merlinite-7b-lab
    parameters:
      density: 0.5
      weight: 0.3
merge_method: ties
base_model: HuggingFaceH4/zephyr-7b-beta
parameters:
  normalize: true
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Muhammad2003/TriMistral-7B-TIES"
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"])

🏆 Evaluation

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	67.68
AI2 Reasoning Challenge (25-Shot)	64.85
HellaSwag (10-Shot)	83.80
MMLU (5-Shot)	63.45
TruthfulQA (0-shot)	56.47
Winogrande (5-shot)	76.64
GSM8k (5-shot)	60.88