README.md

metadata

license: llama2
tags:
  - moe
  - merge
model-index:
  - name: Grafted-Llama2-2x70B
    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: 72.61
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lodrick-the-lafted/Grafted-Llama2-2x70B
          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: 89.57
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lodrick-the-lafted/Grafted-Llama2-2x70B
          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: 71.67
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lodrick-the-lafted/Grafted-Llama2-2x70B
          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: 66.49
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lodrick-the-lafted/Grafted-Llama2-2x70B
          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: 84.37
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lodrick-the-lafted/Grafted-Llama2-2x70B
          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: 57.92
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lodrick-the-lafted/Grafted-Llama2-2x70B
          name: Open LLM Leaderboard

The Llamas are WinterGoddess + AuroraNights.

This is yet another mergekit abomination.

This is probably more of a "dense" MoE than a sparse one.

Unfortunately, most of the testing I have tried with this model shows it works well for a couple sentences, then it starts spouting gibberish. Don't waste your bandwidth.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	73.77
AI2 Reasoning Challenge (25-Shot)	72.61
HellaSwag (10-Shot)	89.57
MMLU (5-Shot)	71.67
TruthfulQA (0-shot)	66.49
Winogrande (5-shot)	84.37
GSM8k (5-shot)	57.92

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	73.77
AI2 Reasoning Challenge (25-Shot)	72.61
HellaSwag (10-Shot)	89.57
MMLU (5-Shot)	71.67
TruthfulQA (0-shot)	66.49
Winogrande (5-shot)	84.37
GSM8k (5-shot)	57.92