README.md

metadata

language:
  - en
license: other
tags:
  - code
datasets:
  - reciprocate/dpo_ultra-capybara-code_filtered-best
license_name: tongyi-qianwen
license_link: https://huggingface.co/Qwen/Qwen1.5-7B-Chat/blob/main/LICENSE
pipeline_tag: text-generation
model-index:
  - name: Coder1.8-ORPO-TEST
    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: 38.82
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=raincandy-u/Coder1.8-ORPO-TEST
          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: 60.48
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=raincandy-u/Coder1.8-ORPO-TEST
          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: 46.7
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=raincandy-u/Coder1.8-ORPO-TEST
          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: 41.38
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=raincandy-u/Coder1.8-ORPO-TEST
          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: 59.75
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=raincandy-u/Coder1.8-ORPO-TEST
          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: 27.45
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=raincandy-u/Coder1.8-ORPO-TEST
          name: Open LLM Leaderboard

Coder1.8-ORPO-TEST

Model Description

Test model for ORPO finetune method, trained on ~20k code examples for 1 epoch on 2 x A40 cards with 4-bit QLora (lora rank=lora alpha=16).

Disclaimer

This is a test model and may generate incorrect responses. Use at your own risk.

Train Details

• Base: Qwen1.5-1.8B
• Training Data: ~20k code examples
• Epochs: 1
• Method: ORPO
• Hardware: 2 x A40
• Quantization: 4-bit QLora
• Lora Rank/Alpha: 16

Limitations

Limited training data and quantization may impact performance.

Join the Discussion

Have questions or feedback? Join our Discord server Here.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	45.76
AI2 Reasoning Challenge (25-Shot)	38.82
HellaSwag (10-Shot)	60.48
MMLU (5-Shot)	46.70
TruthfulQA (0-shot)	41.38
Winogrande (5-shot)	59.75
GSM8k (5-shot)	27.45