README.md

metadata

license: apache-2.0
tags:
  - mistral
  - dpo
  - una
  - finetune
  - chatml
  - instruct
datasets:
  - Intel/orca_dpo_pairs
model-index:
  - name: Neural-una-cybertron-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: 69.03
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PulsarAI/Neural-una-cybertron-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: 84.51
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PulsarAI/Neural-una-cybertron-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: 62.79
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PulsarAI/Neural-una-cybertron-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: 64.99
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PulsarAI/Neural-una-cybertron-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: 80.66
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PulsarAI/Neural-una-cybertron-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: 52.31
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=PulsarAI/Neural-una-cybertron-7b
          name: Open LLM Leaderboard

Neural-una-cybertron-7b

Neural-una-cybertron-7b is an fblgit/una-cybertron-7b-v2-bf16 model that has been further fine-tuned with Direct Preference Optimization (DPO) using the Intel/orca_dpo_pairs dataset.

This model was created after examining the procedure of mlabonne/NeuralHermes-2.5-Mistral-7B model. Special thanks to @mlabonne.

Addionatal Information

This model was fine-tuned on Nvidia A100-SXM4-40GB GPU.

The total training time was 1 hour and 10 minutes.

Prompt Template(s)

ChatML

<|im_start|>system
{system}<|im_end|>
<|im_start|>user
{user}<|im_end|>
<|im_start|>assistant
{asistant}<|im_end|>

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.	69.05
AI2 Reasoning Challenge (25-Shot)	69.03
HellaSwag (10-Shot)	84.51
MMLU (5-Shot)	62.79
TruthfulQA (0-shot)	64.99
Winogrande (5-shot)	80.66
GSM8k (5-shot)	52.31