README.md

metadata

language:
  - en
license: apache-2.0
tags:
  - openllama
  - 3b
datasets:
  - totally-not-an-llm/EverythingLM-data-V3
model-index:
  - name: open-llama-3b-v2-elmv3
    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: 42.06
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/open-llama-3b-v2-elmv3
          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: 73.28
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/open-llama-3b-v2-elmv3
          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: 27.61
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/open-llama-3b-v2-elmv3
          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: 35.54
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/open-llama-3b-v2-elmv3
          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: 64.96
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/open-llama-3b-v2-elmv3
          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: 3.41
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=aloobun/open-llama-3b-v2-elmv3
          name: Open LLM Leaderboard

Trained on 3 epoch of the EverythingLM data.

Eval Results :

I like to tweak smaller models than 3B and mix loras, but now I'm trying my hand at finetuning a 3B model. Lets see how it goes.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	41.14
AI2 Reasoning Challenge (25-Shot)	42.06
HellaSwag (10-Shot)	73.28
MMLU (5-Shot)	27.61
TruthfulQA (0-shot)	35.54
Winogrande (5-shot)	64.96
GSM8k (5-shot)	3.41