README.md

metadata

license: mit
model-index:
  - name: openchat-mistral-7b-reproduce
    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: 26.62
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=clowman/openchat-mistral-7b-reproduce
          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: 27.96
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=clowman/openchat-mistral-7b-reproduce
          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: 24.17
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=clowman/openchat-mistral-7b-reproduce
          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: 48.36
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=clowman/openchat-mistral-7b-reproduce
          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: 50.67
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=clowman/openchat-mistral-7b-reproduce
          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: 0
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=clowman/openchat-mistral-7b-reproduce
          name: Open LLM Leaderboard

A reproduction of https://github.com/imoneoi/openchat.

Training command:

deepspeed --num_gpus=8 --module ochat.training_deepspeed.train \
          --model_path imone/Mistral_7B_with_EOT_token \
          --data_prefix ./data/ \
          --save_path ./checkpoints/mistral-7b/ \
          --batch_max_len 77824 \
          --epochs 10 \
          --save_every 1 \
          --deepspeed \
          --deepspeed_config deepspeed_config.json

deepspeed_config.json:

{
    "bf16": {
        "enabled": true
    },
    "zero_optimization": {
        "stage": 2
    },
    "gradient_clipping": 1.0,
    "gradient_accumulation_steps": 1,
    "train_micro_batch_size_per_gpu": 1,
    "steps_per_print": 100,
    "wall_clock_breakdown": false
}

Training data is https://huggingface.co/datasets/openchat/openchat_sharegpt4_dataset

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	29.63
AI2 Reasoning Challenge (25-Shot)	26.62
HellaSwag (10-Shot)	27.96
MMLU (5-Shot)	24.17
TruthfulQA (0-shot)	48.36
Winogrande (5-shot)	50.67
GSM8k (5-shot)	0.00