README.md

metadata

language:
  - en
license: apache-2.0
datasets:
  - Intel/orca_dpo_pairs
model-index:
  - name: TinyLlama-1.1B-orca-v1.0
    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: 36.35
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sreeramajay/TinyLlama-1.1B-orca-v1.0
          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: 61.23
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sreeramajay/TinyLlama-1.1B-orca-v1.0
          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: 25.18
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sreeramajay/TinyLlama-1.1B-orca-v1.0
          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: 36.58
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sreeramajay/TinyLlama-1.1B-orca-v1.0
          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: 61.4
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sreeramajay/TinyLlama-1.1B-orca-v1.0
          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: 2.27
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=sreeramajay/TinyLlama-1.1B-orca-v1.0
          name: Open LLM Leaderboard

Applied DPO to TinyLlama-1.1B-Chat-v1.0 using orca_dpo_pairs dataset

This is only experimental Model created by following instruction from the nice Blog Fine-tune a Mistral-7b model with Direct Preference Optimization

You can run this model using the following code:

# Format prompt
message = [
    {"role": "system", "content": "You are a helpful assistant chatbot."},
    {"role": "user", "content": "What is a Large Language Model?"}
]
tokenizer = AutoTokenizer.from_pretrained(new_model)
prompt = tokenizer.apply_chat_template(message, add_generation_prompt=True, tokenize=False)

# Create pipeline
pipeline = transformers.pipeline(
    "text-generation",
    model=new_model,
    tokenizer=tokenizer
)

# Generate text
sequences = pipeline(
    prompt,
    do_sample=True,
    temperature=0.7,
    top_p=0.9,
    num_return_sequences=1,
    max_length=200,
)
print(sequences[0]['generated_text'])
# <|system|>
# You are a helpful assistant chatbot.</s>
# <|user|>
# What is a Large Language Model?</s>
# <|assistant|>
# A Large Language Model (LLM) is a type of deep learning model that processes large amounts of text or data to improve the accuracy of natural language processing tasks such as sentiment analysis, machine translation, and question answering. LLMs are trained using large datasets, which allow them to generalize better and have better performance compared to traditional machine learning models. They are capable of handling vast amounts of text and can learn complex relationships between words, phrases, and sentences, making them an essential tool for natural language processing.

Results on GPT4ALL benchmark:

Tasks	Metric	Value		Stderr
arc_challenge	acc	0.3003	±	0.0134
	acc_norm	0.3276	±	0.0137
arc_easy	acc	0.6115	±	0.0100
	acc_norm	0.5354	±	0.0102
boolq	acc	0.6147	±	0.0085
hellaswag	acc	0.4633	±	0.0050
	acc_norm	0.6033	±	0.0049
openbookqa	acc	0.2480	±	0.0193
	acc_norm	0.3720	±	0.0216
piqa	acc	0.7470	±	0.0101
	acc_norm	0.7470	±	0.0101
winogrande	acc	0.6054	±	0.0137

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	37.17
AI2 Reasoning Challenge (25-Shot)	36.35
HellaSwag (10-Shot)	61.23
MMLU (5-Shot)	25.18
TruthfulQA (0-shot)	36.58
Winogrande (5-shot)	61.40
GSM8k (5-shot)	2.27