README.md

metadata

license: cc-by-nc-4.0
tags:
  - dpo
datasets:
  - yleo/emerton_dpo_pairs
base_model: mlabonne/OmniBeagle14-7B
model-index:
  - name: ParrotOgno-7B
    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: 73.04
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=yleo/ParrotOgno-7B
          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: 89.03
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=yleo/ParrotOgno-7B
          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: 64.51
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=yleo/ParrotOgno-7B
          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: 76.53
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=yleo/ParrotOgno-7B
          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: 84.61
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=yleo/ParrotOgno-7B
          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: 69.6
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=yleo/ParrotOgno-7B
          name: Open LLM Leaderboard

---

🦜 ParrotOgno-7B

ParrotOgno-7B is a DPO fine-tune of paulml/OGNO-7B using the yleo/emerton_dpo_pairs_judge preference dataset created from Intel/orca_dpo_pairs by replacing gpt 3.5 answer by a gpt4 Turbo answer. Then, gpt4 Turbo is put as chosen whereas gpt4 is put as rejected.

🔍 Applications

This model uses a context window of 8k. It is compatible with different templates, like chatml and Llama's chat template.

🏆 Evaluation

Open LLM Leaderboard

To come...

💻 Usage

!pip install -qU transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "yleo/ParrotOgno-7B"
messages = [{"role": "user", "content": "How to improve LLM fine-tuning?"}]
tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
"text-generation",
model=model,
torch_dtype=torch.float16,
device_map="auto",
)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	76.22
AI2 Reasoning Challenge (25-Shot)	73.04
HellaSwag (10-Shot)	89.03
MMLU (5-Shot)	64.51
TruthfulQA (0-shot)	76.53
Winogrande (5-shot)	84.61
GSM8k (5-shot)	69.60