Diana-7B / README.md
sethuiyer's picture
Update README.md
339ae47 verified
metadata
inference: false
base_model:
  - SanjiWatsuki/Silicon-Maid-7B
  - sethuiyer/Aika-7B
  - sethuiyer/Nandine-7b
  - mlabonne/AlphaMonarch-7B
library_name: transformers
tags:
  - mergekit
  - merge
  - not-for-all-audiences
license: cc
model-index:
  - name: sethuiyer/Diana-7B
    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: 68.34
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Diana-7B
          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: 86.73
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Diana-7B
          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: 64.58
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Diana-7B
          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: 60.55
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Diana-7B
          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: 80.19
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Diana-7B
          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: 63.23
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sethuiyer/Diana-7B
          name: Open LLM Leaderboard
language:
  - en
pipeline_tag: text-generation

Diana-7B

Diana

This is Diana-7b, rated 93.56/100 by GPT-4 on a collection of 30 synthetic prompts generated by GPT-4.

Diana stands for Deep Insight and Analytical Narrative Assistant and is a merge of the following models using mergekit:

  1. mlabonne/AlphaMonarch-7B: This model has impressive conversational abilities, formal and sophisticated style, and strong reasoning skills.
  2. sethuiyer/Aika-7b: A merge of SanjiWatsuki/Silicon-Maid-7B, Guilherme34/Samantha-v2, jan-hq/stealth-v1.3, and senseable/WestLake-7B-v2, Aika-7b is designed for natural and human-like interactions, accurate information delivery, comprehensive analysis, emotional intelligence, clarity, and structure.
  3. SanjiWatsuki/Silicon-Maid-7B: This model is known for its excellent multi-turn conversational skills and logical coherence.
  4. sethuiyer/Nandine-7b: A merge of senseable/Westlake-7B, Guilherme34/Samantha-v2, and uukuguy/speechless-mistral-six-in-one-7b, Nandine-7b excels in narrative skill, empathetic interaction, intellectual depth, and eloquent communication.

By combining these models, Diana-7B offers a balanced blend of capabilities, making it suitable for various tasks and providing a comprehensive AI companion for creative writing, thoughtful discussions, problem-solving, and general assistance.

OpenLLM Benchmark

Model Average ⬆️ ARC HellaSwag MMLU TruthfulQA Winogrande GSM8K
sethuiyer/Diana-7B 📑 70.6 68.34 86.73 64.58 60.55 80.19 63.23

Nous Benchmark

Model AGIEval GPT4All TruthfulQA Bigbench Average
Diana-7B 44.38 75.1 60.55 44.58 56.09

Configuration

The following YAML configuration was used to produce this model:


base_model: mlabonne/AlphaMonarch-7B
dtype: bfloat16
merge_method: dare_ties
models:
- model: mlabonne/AlphaMonarch-7B
- model: sethuiyer/Aika-7B
  parameters:
    density: 0.85
    weight: 0.30
- model: SanjiWatsuki/Silicon-Maid-7B
  parameters:
    density: 0.85
    weight: 0.50
- model: sethuiyer/Nandine-7b
  parameters:
    density: 0.85
    weight: 0.30
parameters:
  int8_mask: true

Prompt Template

{bos}user
{ .Prompt }{eos}
{bos}assistant

GGUF

GGUF files are available at Diana-7B-GGUF

Ollama

Diana is now available on Ollama. You can use it by running the command ollama run stuehieyr/diana in your terminal. If you have limited computing resources, check out this video to learn how to run it on a Google Colab backend.