README.md

metadata

license: mit
tags:
  - mergekit
  - frankenstein
  - merge
base_model:
  - jondurbin/nontoxic-bagel-34b-v0.2
model-index:
  - name: shqiponja-59b-v1
    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: 70.05
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=nisten/shqiponja-59b-v1
          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: 84.06
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=nisten/shqiponja-59b-v1
          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: 75.54
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=nisten/shqiponja-59b-v1
          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: 70.43
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=nisten/shqiponja-59b-v1
          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: 80.27
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=nisten/shqiponja-59b-v1
          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: 15.47
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=nisten/shqiponja-59b-v1
          name: Open LLM Leaderboard

Shqiponja-59 V1

This is an untrained experimental 59B merged model.

Picked these two models specifically to compliment each others strengths.

Models Merged

• NousResearch/Nous-Hermes-2-Yi-34B
• jondurbin/nontoxic-bagel-34b-v0.2 Merged using the Undi95 style passthrough merge method.

The secret sauce

The following YAML configuration was used to produce this model:

dtype: bfloat16
merge_method: passthrough
slices:
- sources:
  - layer_range: [0, 52]
    model: /home/admin/nv1/nontoxic-bagel-34b-v0.2
- sources:
  - layer_range: [8, 60]
    model: /home/admin/nv1/Nous-Hermes-2-Yi-34B

License MIT - Enjoy

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	65.97
AI2 Reasoning Challenge (25-Shot)	70.05
HellaSwag (10-Shot)	84.06
MMLU (5-Shot)	75.54
TruthfulQA (0-shot)	70.43
Winogrande (5-shot)	80.27
GSM8k (5-shot)	15.47