NEBULA-XB-v1.0 / README.md
TeeZee's picture
Adding Evaluation Results (#1)
01702f3 verified
metadata
language:
  - en
license: apache-2.0
datasets:
  - Open-Orca/SlimOrca
model-index:
  - name: NEBULA-XB-v1.0
    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: 56.66
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/NEBULA-XB-v1.0
          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: 81.78
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/NEBULA-XB-v1.0
          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: 60.98
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/NEBULA-XB-v1.0
          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: 44.03
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/NEBULA-XB-v1.0
          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: 77.66
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=TeeZee/NEBULA-XB-v1.0
          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=TeeZee/NEBULA-XB-v1.0
          name: Open LLM Leaderboard

TeeZee/NEBULA-XB-v1.03

Experiment, can DUS be taken one or more steps further?

Technical notes:

  • pretrained model v03 finetuned on 50k entries from SlimOrca dataset
  • 18 layers removed from both models of finetuned GALAXY-XB-v03
  • model has 108 layers (((48-12)*2)-18)*2 = 108
  • second step in scaling DUS procedure

To evaluate

  • model performance after merge, should be a little lover that GALAXY finetuned on 50k of slimorca

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 53.52
AI2 Reasoning Challenge (25-Shot) 56.66
HellaSwag (10-Shot) 81.78
MMLU (5-Shot) 60.98
TruthfulQA (0-shot) 44.03
Winogrande (5-shot) 77.66
GSM8k (5-shot) 0.00