metadata

license: other
tags:
  - yi
  - moe
license_name: yi-license
license_link: https://huggingface.co/01-ai/Yi-34B-200K/blob/main/LICENSE
model-index:
  - name: Bagel-Hermes-2x34b
    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: 69.8
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Weyaxi/Bagel-Hermes-2x34b
          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: 85.26
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Weyaxi/Bagel-Hermes-2x34b
          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: 77.24
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Weyaxi/Bagel-Hermes-2x34b
          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: 64.82
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Weyaxi/Bagel-Hermes-2x34b
          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: 84.77
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Weyaxi/Bagel-Hermes-2x34b
          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: 68.69
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Weyaxi/Bagel-Hermes-2x34b
          name: Open LLM Leaderboard

Bagel-Hermes-2x34B

This is the model for Bagel-Hermes-2x34B. I used this repo to make this MOE model.

Prompt Template(s):

Since bagel-dpo-34b-v0.2 uses many prompt templates, and Nous-Hermes-2-Yi-34B uses ChatML, you can utilize ChatML and other prompt templates provided by bagel.

Note: I currently do not know which prompt template is best.

ChatML:

<|im_start|>system
{system}<|im_end|>
<|im_start|>user
{user}<|im_end|>
<|im_start|>assistant
{asistant}<|im_end|>

Alpaca (sort of)

Below is an instruction that describes a task.  Write a response that appropriately completes the request.

### Instruction:
{system}
{instruction}

### Response:

Vicuna

{system}
USER: {instruction}
ASSISTANT:

Visit bagel-dpo-34b-v0.2 to try more prompt templates.

Yaml Config to reproduce

base_model: nontoxic-bagel-34b-v0.2
gate_mode: hidden
dtype: bfloat16

experts:
  - source_model: bagel-dpo-34b-v0.2
    positive_prompts: ["question answering", "Q:", science", "biology", "chemistry", "physics"]

  - source_model: Nous-Hermes-2-Yi-34B
    positive_prompts: ["chat", "math", "reason", "mathematics", "solve", "count", "python", "javascript", "programming", "algorithm", "tell me", "assistant"]

Quantizationed versions

Quantizationed versions of this model is available thanks to TheBloke.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	75.10
AI2 Reasoning Challenge (25-Shot)	69.80
HellaSwag (10-Shot)	85.26
MMLU (5-Shot)	77.24
TruthfulQA (0-shot)	64.82
Winogrande (5-shot)	84.77
GSM8k (5-shot)	68.69

If you would like to support me:

☕ Buy Me a Coffee