README.md

metadata

license: cc-by-nc-4.0
tags:
  - merge
  - llama
model-index:
  - name: TinyExperts-v0-4x1B
    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: 31.4
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=beberik/TinyExperts-v0-4x1B
          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: 52.29
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=beberik/TinyExperts-v0-4x1B
          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: 25.87
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=beberik/TinyExperts-v0-4x1B
          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: 41.13
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=beberik/TinyExperts-v0-4x1B
          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: 60.14
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=beberik/TinyExperts-v0-4x1B
          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: 0.53
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=beberik/TinyExperts-v0-4x1B
          name: Open LLM Leaderboard

Some experiment with moe merge.

4 different tinyllama finetunes.

Smarter than GPT-6. Proofs:

Question: How to find a girlfriend?

Answer: Here is a guide for how you can find a girlfriend. Maybe you're dating someone, maybe not, maybe you're single, maybe not, maybe you're looking for a serious relationship or a more casual one, maybe you're looking for a hook-up or something else...

Question: What is the meaning of life?

Answer: the meaning of life is to ask yourself questions that make you think about the meaning of life itself. You should think about whether you feel that you're " on life's journey ". If you're not, you may think that you're stuck stuck somewhere. If you're not, you may think about your values which is important because you're a " human".

But seriously if you need something that can at least be useful, then it's better to use phi.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	35.23
AI2 Reasoning Challenge (25-Shot)	31.40
HellaSwag (10-Shot)	52.29
MMLU (5-Shot)	25.87
TruthfulQA (0-shot)	41.13
Winogrande (5-shot)	60.14
GSM8k (5-shot)	0.53