Aratako's picture
Adding Evaluation Results (#1)
180e3e3 verified
metadata
license: cc-by-nc-4.0
tags:
  - moe
  - merge
  - mergekit
base_model:
  - mlabonne/AlphaMonarch-7B
  - beowolx/CodeNinja-1.0-OpenChat-7B
  - SanjiWatsuki/Kunoichi-DPO-v2-7B
  - mlabonne/NeuralDaredevil-7B
model-index:
  - name: Beyonder-4x7B-random-lora
    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: 71.25
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Aratako/Beyonder-4x7B-random-lora
          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: 87.4
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Aratako/Beyonder-4x7B-random-lora
          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: 64.78
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Aratako/Beyonder-4x7B-random-lora
          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: 70.49
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Aratako/Beyonder-4x7B-random-lora
          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: 82.16
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Aratako/Beyonder-4x7B-random-lora
          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: 67.4
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Aratako/Beyonder-4x7B-random-lora
          name: Open LLM Leaderboard

Beyonder-4x7B-v3-random-lora

The idea was very simple. If heuristic methods for determining gate parameters in mergekit-based MoE models can work well, then perhaps we could obtain a better performing model by fine-tuning only the gate parameters.

This model is an attempt at testing that idea. Unfortunately, the performance degraded slightly, but I am sharing it as an experimental result.

Model Details

First, I created an MoE model using mergekit with gate_mode=random and the following four models (same as mlabonne/Beyonder-4x7B-v3):

Then, I used LoRA to fine-tune only the gate parameters by specifying "gate" in target_modules. The data used for fine-tuning is as follows. I used the Mistral prompt format.

The training was conducted on runpod using 4xA6000 GPUs. The main training parameters are as follows:

  • lora_r: 128
  • lora_alpha: 256
  • lora_dropout: 0.05
  • lora_target_modules: "gate"
  • learning_rate: 3e-4
  • num_train_epochs: 5
  • batch_size: 64
  • max_seq_length: 2048

Evaluation

The evaluation results show a slight degradation in performance. Apart from the possibility that this approach may not be effective, other potential causes could be issues with the dataset, training parameters, training setup (such as prompt formatting), and so on.

Nous (LLM AutoEval)

Model Average AGIEval GPT4All TruthfulQA Bigbench
mlabonne/AlphaMonarch-7B πŸ“„ 62.74 45.37 77.01 78.39 50.2
mlabonne/Beyonder-4x7B-v3 πŸ“„ 61.91 45.85 76.67 74.98 50.12
Aratako/Beyonder-4x7B-v3-random-lora πŸ“„ 60.29 45.82 76.69 69.94 48.72
mlabonne/NeuralDaredevil-7B πŸ“„ 59.39 45.23 76.2 67.61 48.52
SanjiWatsuki/Kunoichi-DPO-v2-7B πŸ“„ 58.29 44.79 75.05 65.68 47.65
mlabonne/Beyonder-4x7B-v2 πŸ“„ 57.13 45.29 75.95 60.86 46.4
beowolx/CodeNinja-1.0-OpenChat-7B πŸ“„ 50.35 39.98 71.77 48.73 40.92

MT-Bench

1-turn

Model Coding Extraction Humanities Math Reasoning Roleplay STEM Writing avg_score
mlabonne/Beyonder-4x7B-v3 6.7 8.3 9.7 6.7 6.3 9.3 9.7 10.0 8.33750
Aratako/Beyonder-4x7B-v3-random-lora 6.6 8.2 9.6 6.3 6.4 8.7 9.4 9.5 8.08750
mistralai/Mixtral-8x7B-Instruct-v0.1 5.3 8.5 9.9 6.8 6.0 9.1 9.55 8.9 8.00625

mt-bench-1turn

2-turn

Model Coding Extraction Humanities Math Reasoning Roleplay STEM Writing avg_score
mlabonne/Beyonder-4x7B-v3 5.4 7.6 10.0 3.5 5.5 9.0 9.6 9.1 7.46250
Aratako/Beyonder-4x7B-v3-random-lora 5.1 8.1 9.9 4.1 3.7 8.55 9.0 7.7 7.01875
mistralai/Mixtral-8x7B-Instruct-v0.1 4.1 8.4 9.8 4.7 5.6 9.0 9.2 9.5 7.53750

mt-bench-2turn

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 73.91
AI2 Reasoning Challenge (25-Shot) 71.25
HellaSwag (10-Shot) 87.40
MMLU (5-Shot) 64.78
TruthfulQA (0-shot) 70.49
Winogrande (5-shot) 82.16
GSM8k (5-shot) 67.40