README.md

metadata

license: llama3.1
library_name: transformers
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
base_model:
  - Joseph717171/Llama-3.1-SuperNova-8B-Lite_TIES_with_Base
  - ArliAI/Llama-3.1-8B-ArliAI-RPMax-v1.2
model-index:
  - name: L3.1-Moe-2x8B-v0.2
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: IFEval (0-Shot)
          type: HuggingFaceH4/ifeval
          args:
            num_few_shot: 0
        metrics:
          - type: inst_level_strict_acc and prompt_level_strict_acc
            value: 73.48
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=moeru-ai/L3.1-Moe-2x8B-v0.2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: BBH (3-Shot)
          type: BBH
          args:
            num_few_shot: 3
        metrics:
          - type: acc_norm
            value: 32.95
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=moeru-ai/L3.1-Moe-2x8B-v0.2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MATH Lvl 5 (4-Shot)
          type: hendrycks/competition_math
          args:
            num_few_shot: 4
        metrics:
          - type: exact_match
            value: 15.26
            name: exact match
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=moeru-ai/L3.1-Moe-2x8B-v0.2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GPQA (0-shot)
          type: Idavidrein/gpqa
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 6.71
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=moeru-ai/L3.1-Moe-2x8B-v0.2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MuSR (0-shot)
          type: TAUR-Lab/MuSR
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 11.38
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=moeru-ai/L3.1-Moe-2x8B-v0.2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU-PRO (5-shot)
          type: TIGER-Lab/MMLU-Pro
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 31.76
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=moeru-ai/L3.1-Moe-2x8B-v0.2
          name: Open LLM Leaderboard

L3.1-Moe-2x8B-v0.2

This model is a Mixture of Experts (MoE) made with mergekit-moe. It uses the following base models:

• Joseph717171/Llama-3.1-SuperNova-8B-Lite_TIES_with_Base
• ArliAI/Llama-3.1-8B-ArliAI-RPMax-v1.2

Heavily inspired by mlabonne/Beyonder-4x7B-v3.

Quantized models

GGUF by mradermacher

• mradermacher/L3.1-Moe-2x8B-v0.2-i1-GGUF
• mradermacher/L3.1-Moe-2x8B-v0.2-GGUF

Mergekit config

mergekit_moe_config.yml

base_model: Joseph717171/Llama-3.1-SuperNova-8B-Lite_TIES_with_Base
gate_mode: hidden
dtype: bfloat16
experts:
  - source_model: Joseph717171/Llama-3.1-SuperNova-8B-Lite_TIES_with_Base
    positive_prompts: &common_prompts
      - "chat"
      - "assistant"
      - "tell me"
      - "explain"
      - "I want"
      - "code"
      - "python"
      - "javascript"
      - "programming"
      - "algorithm"
      - "reason"
      - "math"
      - "mathematics"
      - "solve"
      - "count"
    negative_prompts: &rp_prompts
      - "storywriting"
      - "write"
      - "scene"
      - "story"
      - "character"
  - source_model: ArliAI/Llama-3.1-8B-ArliAI-RPMax-v1.2
    positive_prompts: *rp_prompts
    negative_prompts: *common_prompts

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	28.59
IFEval (0-Shot)	73.48
BBH (3-Shot)	32.95
MATH Lvl 5 (4-Shot)	15.26
GPQA (0-shot)	6.71
MuSR (0-shot)	11.38
MMLU-PRO (5-shot)	31.76