Etheria-55b-v0.1 / README.md
Steelskull's picture
Adding Evaluation Results (#3)
81f6eb7 verified
metadata
license: apache-2.0
tags:
  - mergekit
  - Etheria
base_model: []
model-index:
  - name: Etheria-55b-v0.1
    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: 65.1
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Steelskull/Etheria-55b-v0.1
          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: 81.93
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Steelskull/Etheria-55b-v0.1
          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: 73.66
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Steelskull/Etheria-55b-v0.1
          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: 56.16
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Steelskull/Etheria-55b-v0.1
          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: 76.09
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Steelskull/Etheria-55b-v0.1
          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: 35.18
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Steelskull/Etheria-55b-v0.1
          name: Open LLM Leaderboard

Steelskull/Etheria-55b-v0.1

image/png

Merge Details

An attempt to make a functional goliath style merge to create a [Etheria] 55b-200k with two yi-34b-200k models.

due to the merge it 'theoretically' should have a context of 200k but I recommend starting at 32k and moveing up, as it is unknown (at this time) what the merge has done to the context length.

This is a merge of both VerA and VerB of Etheria-55b (There numbers were surprisingly good), I then created a sacrificial 55B out of the most performant yi-34b-200k Model and performed a Dare_ties merge and equalize the model into its current state.

recommended settings and Prompt Format:

Ive tested it up to 32k context using exl2 using these settings:

    "temp": 0.7,
    "temperature_last": true,
    "top_p": 1,
    "top_k": 0,
    "top_a": 0,
    "tfs": 1,
    "epsilon_cutoff": 0,
    "eta_cutoff": 0,
    "typical_p": 1,
    "min_p": 0.1,
    "rep_pen": 1.1,
    "rep_pen_range": 8192,
    "no_repeat_ngram_size": 0,
    "penalty_alpha": 0,
    "num_beams": 1,
    "length_penalty": 1,
    "min_length": 0,
    "encoder_rep_pen": 1,
    "freq_pen": 0,
    "presence_pen": 0,
    "do_sample": true,
    "early_stopping": false,
    "add_bos_token": false,
    "truncation_length": 2048,
    "ban_eos_token": true,
    "skip_special_tokens": true,
    "streaming": true,
    "mirostat_mode": 0,
    "mirostat_tau": 5,
    "mirostat_eta": 0.1,

Prompt format that work well

ChatML & Alpaca

Merge Method

This model was merged using the DARE TIES merge method using Merged-Etheria-55b as a base.

Configuration

The following YAML configuration was used to produce this model:


base_model: Merged-Etheria-55b
models:
  - model: Sacr-Etheria-55b
    parameters:
      weight: [0.22, 0.113, 0.113, 0.113, 0.113, 0.113]
      density: 0.61
  - model: Merged-Etheria-55b
    parameters:
      weight: [0.22, 0.113, 0.113, 0.113, 0.113, 0.113]
      density: 0.61
merge_method: dare_ties
tokenizer_source: union
parameters:
  int8_mask: true
dtype: bfloat16

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 64.69
AI2 Reasoning Challenge (25-Shot) 65.10
HellaSwag (10-Shot) 81.93
MMLU (5-Shot) 73.66
TruthfulQA (0-shot) 56.16
Winogrande (5-shot) 76.09
GSM8k (5-shot) 35.18