README.md

metadata

license: apache-2.0
library_name: transformers
tags:
  - mergekit
  - merge
base_model:
  - mistralai/Mistral-7B-v0.1
  - cognitivecomputations/dolphin-2.2.1-mistral-7b
  - Open-Orca/Mistral-7B-OpenOrca
  - openchat/openchat-3.5-0106
  - mlabonne/NeuralHermes-2.5-Mistral-7B
  - GreenNode/GreenNode-mini-7B-multilingual-v1olet
  - berkeley-nest/Starling-LM-7B-alpha
  - viethq188/LeoScorpius-7B-Chat-DPO
  - meta-math/MetaMath-Mistral-7B
  - Intel/neural-chat-7b-v3-3
inference: false
model-index:
  - name: Moza-7B-v1.0
    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: 66.55
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=kidyu/Moza-7B-v1.0
          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: 83.45
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=kidyu/Moza-7B-v1.0
          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: 62.77
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=kidyu/Moza-7B-v1.0
          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: 65.16
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=kidyu/Moza-7B-v1.0
          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: 77.51
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=kidyu/Moza-7B-v1.0
          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: 62.55
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=kidyu/Moza-7B-v1.0
          name: Open LLM Leaderboard

Moza-7B-v1.0

This is a meme-merge of pre-trained language models, created using mergekit. Use at your own risk.

Details

Quantized Model

• GGUF

Merge Method

This model was merged using the DARE TIES merge method, using mistralai/Mistral-7B-v0.1 as a base.

The value for density are from this blogpost, and the weight was randomly generated and then assigned to the models, with priority (of using the bigger weight) to NeuralHermes, OpenOrca, and neural-chat. The models themselves are chosen by "vibes".

Models Merged

The following models were included in the merge:

• cognitivecomputations/dolphin-2.2.1-mistral-7b
• Open-Orca/Mistral-7B-OpenOrca
• openchat/openchat-3.5-0106
• mlabonne/NeuralHermes-2.5-Mistral-7B
• GreenNode/GreenNode-mini-7B-multilingual-v1olet
• berkeley-nest/Starling-LM-7B-alpha
• viethq188/LeoScorpius-7B-Chat-DPO
• meta-math/MetaMath-Mistral-7B
• Intel/neural-chat-7b-v3-3

Prompt Format

You can use Alpaca formatting for inference

### Instruction:

### Response:

Configuration

The following YAML configuration was used to produce this model:

base_model: mistralai/Mistral-7B-v0.1
models:
  - model: mlabonne/NeuralHermes-2.5-Mistral-7B
    parameters:
      density: 0.63
      weight: 0.83
  - model: Intel/neural-chat-7b-v3-3
    parameters:
      density: 0.63
      weight: 0.74
  - model: meta-math/MetaMath-Mistral-7B
    parameters:
      density: 0.63
      weight: 0.22
  - model: openchat/openchat-3.5-0106
    parameters:
      density: 0.63
      weight: 0.37
  - model: Open-Orca/Mistral-7B-OpenOrca
    parameters:
      density: 0.63
      weight: 0.76
  - model: cognitivecomputations/dolphin-2.2.1-mistral-7b
    parameters:
      density: 0.63
      weight: 0.69
  - model: viethq188/LeoScorpius-7B-Chat-DPO
    parameters:
      density: 0.63
      weight: 0.38
  - model: GreenNode/GreenNode-mini-7B-multilingual-v1olet
    parameters:
      density: 0.63
      weight: 0.13
  - model: berkeley-nest/Starling-LM-7B-alpha
    parameters:
      density: 0.63
      weight: 0.33
merge_method: dare_ties
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	69.66
AI2 Reasoning Challenge (25-Shot)	66.55
HellaSwag (10-Shot)	83.45
MMLU (5-Shot)	62.77
TruthfulQA (0-shot)	65.16
Winogrande (5-shot)	77.51
GSM8k (5-shot)	62.55