metadata

license: apache-2.0
library_name: transformers
tags:
  - mergekit
  - merge
base_model:
  - localfultonextractor/Erosumika-7B
  - Nitral-AI/Infinitely-Laydiculous-7B
  - Kunocchini-7b-128k-test
  - Endevor/EndlessRP-v3-7B
  - ChaoticNeutrals/BuRP_7B
  - daybreak-kunoichi-2dpo-7b
model-index:
  - name: Confluence-Renegade-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: 31.91
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Nekochu/Confluence-Renegade-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: 45.38
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Nekochu/Confluence-Renegade-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: 31.48
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Nekochu/Confluence-Renegade-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: 51.47
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Nekochu/Confluence-Renegade-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: 63.14
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Nekochu/Confluence-Renegade-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: 0
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Nekochu/Confluence-Renegade-7B
          name: Open LLM Leaderboard

My first merge of RP models 7B using mergekit, They are just r/ trend RP, half is BuRP_7B. not used any, Dumb merge but hopfully lucky merge! ^^'

Update 03/2024:

Original model Card Confluence-Renegade-7B [8.0bpw-exl]
Added Model and merge recipe branch: Confluence-Renegade-7B-v2
Added Model and merge recipe branch: RoleBeagle-Moistral-11B-v2 [7B truncated] and Quants RoleBeagle-Moistral-11B-v2-2.4bpw-h6-exl2, 4.25bpw-h6, 8.0bpw-h8
Added Branch: Confluence-Shortcake-20B Model recipes and Quants here Confluence-Shortcake-20B-2.4bpw-h6-exl2, 4.25bpw-h6, 8.0bpw-h8

Name symbolize by Confluence for many unique RP model with Renegade mostly come from no-guardrail.

Download branch instructions

git clone --single-branch --branch Confluence-Shortcake-20B-2.4bpw-h6-exl2 https://huggingface.co/Nekochu/Confluence-Renegade-7B

Configuration Confluence-Renegade-7B

The following YAML configuration was used to produce this model:

models:
  - model: ./modela/Erosumika-7B
    parameters:
      density: [1, 0.8, 0.6]
      weight: 0.2
  - model: ./modela/Infinitely-Laydiculous-7B
    parameters:
      density: [0.9, 0.7, 0.5]
      weight: 0.2
  - model: ./modela/Kunocchini-7b-128k-test
    parameters:
      density: [0.8, 0.6, 0.4]
      weight: 0.2
  - model: ./modela/EndlessRP-v3-7B
    parameters:
      density: [0.7, 0.5, 0.3]
      weight: 0.2
  - model: ./modela/daybreak-kunoichi-2dpo-7b
    parameters:
      density: [0.5, 0.3, 0.1]
      weight: 0.2
merge_method: dare_linear
base_model: ./modela/Mistral-7B-v0.1
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16
name: intermediate-model
---
slices:
  - sources:
      - model: intermediate-model
        layer_range: [0, 32]
      - model: ./modela/BuRP_7B
        layer_range: [0, 32]
merge_method: slerp
base_model: intermediate-model
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5 # fallback for rest of tensors
dtype: bfloat16
name: gradient-slerp

mergekit-mega config.yml ./output-model-directory --cuda --allow-crimes --lazy-unpickle

Models Merged Confluence-Renegade-7B

The following models were included in the merge:

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	37.23
AI2 Reasoning Challenge (25-Shot)	31.91
HellaSwag (10-Shot)	45.38
MMLU (5-Shot)	31.48
TruthfulQA (0-shot)	51.47
Winogrande (5-shot)	63.14
GSM8k (5-shot)	0.00