README.md

metadata

language:
  - en
license: apache-2.0
tags:
  - mistral
  - instruct
  - finetune
  - chatml
  - gpt4
  - synthetic data
  - distillation
  - dpo
  - rlhf
datasets:
  - unalignment/toxic-dpo-v0.1
base_model: teknium/OpenHermes-2.5-Mistral-7B
model-index:
  - name: ToxicHermes-2.5-Mistral-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: 64.59
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=joey00072/ToxicHermes-2.5-Mistral-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: 83.75
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=joey00072/ToxicHermes-2.5-Mistral-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: 63.67
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=joey00072/ToxicHermes-2.5-Mistral-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: 50.84
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=joey00072/ToxicHermes-2.5-Mistral-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: 77.9
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=joey00072/ToxicHermes-2.5-Mistral-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: 17.36
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=joey00072/ToxicHermes-2.5-Mistral-7B
          name: Open LLM Leaderboard

ToxicHermes

OpenHermes-2.5 model + toxic-dpo Dataset = ToxicHermes

fine-tuned with Direct Preference Optimization (DPO)

• Base Model: teknium/OpenHermes-2.5-Mistral-7B
• Dataset: unalignment/toxic-dpo-v0.1

Usage

You can also run this model using the following code:

import transformers
from transformers import AutoTokenizer


model = "joey00072/ToxicHermes-2.5-Mistral-7B"
# 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(model)
prompt = tokenizer.apply_chat_template(message, add_generation_prompt=True, tokenize=False)

# Create pipeline
pipeline = transformers.pipeline(
    "text-generation",
    model=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'])

Training hyperparameters

LoRA:

• r=16
• lora_alpha=16
• lora_dropout=0.05
• bias="none"
• task_type="CAUSAL_LM"
• target_modules=['k_proj', 'gate_proj', 'v_proj', 'up_proj', 'q_proj', 'o_proj', 'down_proj']

Training arguments:

• per_device_train_batch_size=4
• gradient_accumulation_steps=4
• gradient_checkpointing=True
• learning_rate=5e-5
• lr_scheduler_type="cosine"
• max_steps=200
• optim="paged_adamw_32bit"
• warmup_steps=100

DPOTrainer:

• beta=0.1
• max_prompt_length=1024
• max_length=1536

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	59.69
AI2 Reasoning Challenge (25-Shot)	64.59
HellaSwag (10-Shot)	83.75
MMLU (5-Shot)	63.67
TruthfulQA (0-shot)	50.84
Winogrande (5-shot)	77.90
GSM8k (5-shot)	17.36