xxx777xxxASD's picture
Adding Evaluation Results (#1)
feb79d4 verified
metadata
language:
  - en
license: llama3
tags:
  - moe
model-index:
  - name: L3-SnowStorm-v1.15-4x8B-B
    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: 60.67
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=xxx777xxxASD/L3-SnowStorm-v1.15-4x8B-B
          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: 81.6
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=xxx777xxxASD/L3-SnowStorm-v1.15-4x8B-B
          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: 68.12
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=xxx777xxxASD/L3-SnowStorm-v1.15-4x8B-B
          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: 51.69
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=xxx777xxxASD/L3-SnowStorm-v1.15-4x8B-B
          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: 76.56
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=xxx777xxxASD/L3-SnowStorm-v1.15-4x8B-B
          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: 69.45
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=xxx777xxxASD/L3-SnowStorm-v1.15-4x8B-B
          name: Open LLM Leaderboard

GGUF

Experimental RP-oriented MoE, the idea was to get a model that would be equal to or better than Mixtral 8x7B and it's finetunes in RP/ERP tasks.

There's:

Llama 3 SnowStorm v1.15B 4x8B

base_model: Sao10K_L3-8B-Stheno-v3.1
gate_mode: random
dtype: bfloat16
experts_per_token: 2
experts:
  - source_model: Nitral-AI_Poppy_Porpoise-1.0-L3-8B
  - source_model: NeverSleep_Llama-3-Lumimaid-8B-v0.1-OAS
  - source_model: openlynn_Llama-3-Soliloquy-8B-v2
  - source_model: Sao10K_L3-8B-Stheno-v3.1

Models used

Difference(from SnowStorm v1.0)

Vision

llama3_mmproj

image/png

Prompt format: Llama 3

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 68.01
AI2 Reasoning Challenge (25-Shot) 60.67
HellaSwag (10-Shot) 81.60
MMLU (5-Shot) 68.12
TruthfulQA (0-shot) 51.69
Winogrande (5-shot) 76.56
GSM8k (5-shot) 69.45