README.md

metadata

license: mit
library_name: transformers
tags:
  - mergekit
  - merge
base_model:
  - rhysjones/phi-2-orange-v2
  - Weyaxi/Einstein-v4-phi2
model-index:
  - name: L-MChat-Small
    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: 61.6
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-Small
          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: 75.9
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-Small
          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: 57.41
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-Small
          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: 49.94
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-Small
          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: 74.98
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-Small
          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: 58.98
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Artples/L-MChat-Small
          name: Open LLM Leaderboard

L-MChat-Small

This was a test of mine how small merges perform, because there are a lot of 7b merges and higher but not a lot of 2b merges.

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

• rhysjones/phi-2-orange-v2
• Weyaxi/Einstein-v4-phi2

Configuration

The following YAML configuration was used to produce this model:

slices:
- sources:
  - model: Weyaxi/Einstein-v4-phi2
    layer_range:
    - 0
    - 32
  - model: rhysjones/phi-2-orange-v2
    layer_range:
    - 0
    - 32
merge_method: slerp
base_model: rhysjones/phi-2-orange-v2
parameters:
  t:
  - filter: self_attn
    value:
    - 0
    - 0.5
    - 0.3
    - 0.7
    - 1
  - filter: mlp
    value:
    - 1
    - 0.5
    - 0.7
    - 0.3
    - 0
  - value: 0.5
dtype: bfloat16

Usage

Use it with the ChatML format, you can also use the Inference-API for this Model.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	63.14
AI2 Reasoning Challenge (25-Shot)	61.60
HellaSwag (10-Shot)	75.90
MMLU (5-Shot)	57.41
TruthfulQA (0-shot)	49.94
Winogrande (5-shot)	74.98
GSM8k (5-shot)	58.98