mwitiderrick's picture
Adding Evaluation Results (#1)
1059643 verified
metadata
language:
  - en
license: apache-2.0
library_name: transformers
tags:
  - transformers
datasets:
  - mwitiderrick/OpenPlatypus
base_model: openlm-research/open_llama_3b
inference: true
model_type: llama
prompt_template: |
  ### Instruction:\n
  {prompt}
  ### Response:
created_by: mwitiderrick
pipeline_tag: text-generation
model-index:
  - name: mwitiderrick/open_llama_3b_instruct_v_0.2
    results:
      - task:
          type: text-generation
        dataset:
          name: hellaswag
          type: hellaswag
        metrics:
          - type: hellaswag (0-Shot)
            value: 0.4882
            name: hellaswag(0-Shot)
      - task:
          type: text-generation
        dataset:
          name: winogrande
          type: winogrande
        metrics:
          - type: winogrande (0-Shot)
            value: 0.6133
            name: winogrande(0-Shot)
      - task:
          type: text-generation
        dataset:
          name: arc_challenge
          type: arc_challenge
        metrics:
          - type: arc_challenge (0-Shot)
            value: 0.3362
            name: arc_challenge(0-Shot)
        source:
          url: https://huggingface.co/mwitiderrick/open_llama_3b_instruct_v_0.2
          name: open_llama_3b_instruct_v_0.2 model card
      - 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: 38.48
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_instruct_v_0.2
          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: 66.77
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_instruct_v_0.2
          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: 25.34
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_instruct_v_0.2
          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: 38.16
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_instruct_v_0.2
          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: 63.46
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_instruct_v_0.2
          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: 1.59
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_instruct_v_0.2
          name: Open LLM Leaderboard

OpenLLaMA Instruct: An Open Reproduction of LLaMA

This is an OpenLlama model that has been fine-tuned on 1 epoch of the Open-Platypus dataset.

The modified version of the dataset can be found here

Prompt Template

### Instruction:

{query}

### Response:
<Leave new line for model to respond> 

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM,pipeline

tokenizer = AutoTokenizer.from_pretrained("mwitiderrick/open_llama_3b_instruct_v_0.2")
model = AutoModelForCausalLM.from_pretrained("mwitiderrick/open_llama_3b_instruct_v_0.2")
query = "Provide step-by-step instructions for making a sweet chicken bugger"
text_gen = pipeline(task="text-generation", model=model, tokenizer=tokenizer, max_length=500)
output = text_gen(f"### Instruction:\n{query}\n### Response:\n")
print(output[0]['generated_text'])
"""
### Instruction:
Provide step-by-step instructions for making a sweet chicken bugger
### Response:
Step 1: Gather your ingredients
1. 1/2 cup of sugar
2. 1/2 cup of corn syrup
3. 1/2 cup of water
4. 1/2 cup of vegetable oil
5. 1/2 cup of vanilla extract
6. 1/2 cup of baking soda
7. 1/2 cup of salt
8. 1/2 cup of flour
9. 1/2 cup of milk
10. 1/2 cup of egg whites

Step 2: Mix the ingredients together
1. Combine the sugar, corn syrup, water, vegetable oil, vanilla extract, baking soda, and salt in a large bowl.
2. Whisk together until smooth.
3. Add the flour and mix until combined.
4. Add the milk and egg whites and mix until combined.
5. Pour the mixture into a greased 9x13 inch baking pan.
6. Bake for 30 minutes or until a toothpick inserted into the center comes out clean.

Step 3: Make the chicken bugger
1. Preheat the oven to 350 degrees Fahrenheit.
2. In a large bowl, combine the corn syrup, sugar, and cornstarch.
3. Add the chicken and mix well.
4. Divide the mixture into 12 equal portions and shape each portion into a chicken shape.
5. Place the chicken shapes on a baking sheet lined with parchment paper.
6. Bake for 15 minutes or until the chicken is cooked through.
7. Remove the chicken from the oven and allow to cool for 5 minutes.
8. Using a fork, carefully remove the chicken from the shells and place on a serving platter.
9. Serve with a side of gravy.

Step 4: Make the gravy
1. In a small saucepan, combine the cornstarch and water.
2. Stir until the mixture is smooth and begins to thicken.
3. Add the chicken broth and bring to a boil.
4. Reduce the heat to low and simmer for 10 minutes or until the gravy is
"""

TruthfulQA metrics



|  Groups  |Version|Filter|n-shot|  Metric   | Value  |   |Stderr|
|----------|-------|------|-----:|-----------|-------:|---|-----:|
|truthfulqa|N/A    |none  |     0|acc        |  0.3166|±  |0.0012|
|          |       |none  |     0|bleu_max   | 23.7766|±  |0.7660|
|          |       |none  |     0|bleu_acc   |  0.3207|±  |0.0163|
|          |       |none  |     0|bleu_diff  | -7.1853|±  |0.7396|
|          |       |none  |     0|rouge1_max | 48.6534|±  |0.8706|
|          |       |none  |     0|rouge1_acc |  0.2766|±  |0.0157|
|          |       |none  |     0|rouge1_diff| -9.8011|±  |0.7883|
|          |       |none  |     0|rouge2_max | 31.9289|±  |0.9637|
|          |       |none  |     0|rouge2_acc |  0.2399|±  |0.0149|
|          |       |none  |     0|rouge2_diff|-11.3958|±  |0.9220|
|          |       |none  |     0|rougeL_max | 45.4592|±  |0.8754|
|          |       |none  |     0|rougeL_acc |  0.2754|±  |0.0156|
|          |       |none  |     0|rougeL_diff|-10.0740|±  |0.7807|

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 38.97
AI2 Reasoning Challenge (25-Shot) 38.48
HellaSwag (10-Shot) 66.77
MMLU (5-Shot) 25.34
TruthfulQA (0-shot) 38.16
Winogrande (5-shot) 63.46
GSM8k (5-shot) 1.59