Aira-OPT-350M

Aira-2 is the second version of the Aira instruction-tuned series. Aira-OPT-350M is an instruction-tuned model based on OPT. The model was trained with a dataset composed of prompts and completions generated synthetically by prompting already-tuned models (ChatGPT, Llama, Open-Assistant, etc).

Check our gradio-demo in Spaces.

Details

• Size: 331,195,392 parameters
• Dataset: Instruct-Aira Dataset
• Language: English
• Number of Epochs: 3
• Batch size: 16
• Optimizer: torch.optim.AdamW (warmup_steps = 1e2, learning_rate = 1e-4, epsilon = 1e-8)
• GPU: 1 NVIDIA A100-SXM4-40GB
• Emissions: 0.33 KgCO2 (Netherlands)
• Total Energy Consumption: 0.85 kWh

This repository has the source code used to train this model.

Usage

Three special tokens are used to mark the user side of the interaction and the model's response:

<|startofinstruction|>What is a language model?<|endofinstruction|>A language model is a probability distribution over a vocabulary.<|endofcompletion|>

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

tokenizer = AutoTokenizer.from_pretrained('nicholasKluge/Aira-OPT-350M')
aira = AutoModelForCausalLM.from_pretrained('nicholasKluge/Aira-OPT-350M')

aira.eval()
aira.to(device)

question =  input("Enter your question: ")

inputs = tokenizer(tokenizer.bos_token + question + tokenizer.sep_token,
  add_special_tokens=False,
  return_tensors="pt").to(device)

responses = aira.generate(**inputs,	num_return_sequences=2)

print(f"Question: 👤 {question}\n")

for i, response in  enumerate(responses):
    print(f'Response {i+1}: 🤖 {tokenizer.decode(response, skip_special_tokens=True).replace(question, "")}')

The model will output something like:

>>>Question: 👤 What is the capital of Brazil?

>>>Response 1: 🤖 The capital of Brazil is Brasília.
>>>Response 2: 🤖 The capital of Brazil is Brasília.

Limitations

• Hallucinations: This model can produce content that can be mistaken for truth but is, in fact, misleading or entirely false, i.e., hallucination.

• Biases and Toxicity: This model inherits the social and historical stereotypes from the data used to train it. Given these biases, the model can produce toxic content, i.e., harmful, offensive, or detrimental to individuals, groups, or communities.

• Repetition and Verbosity: The model may get stuck on repetition loops (especially if the repetition penalty during generations is set to a meager value) or produce verbose responses unrelated to the prompt it was given.

Evaluation

Model	Average	ARC	TruthfulQA	ToxiGen
Aira-OPT-125M	43.34	24.65	49.11	56.27
OPT-125M	40.29	22.78	42.88	55.21
Aira-OPT-350M	41.56	25.00	42.13	57.55
OPT-350M	40.62	23.97	41.00	56.91
Aira-OPT-1B3	43.90	28.41	46.59	56.70
OPT-1.3b	40.91	29.69	38.68	54.36

• Evaluations were performed using the Language Model Evaluation Harness (by EleutherAI).

Cite as 🤗

@misc{nicholas22aira,
  doi = {10.5281/zenodo.6989727},
  url = {https://github.com/Nkluge-correa/Aira},
  author = {Nicholas Kluge Corrêa},
  title = {Aira},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
}

@phdthesis{kluge2024dynamic,
  title={Dynamic Normativity},
  author={Kluge Corr{\^e}a, Nicholas},
  year={2024},
  school={Universit{\"a}ts-und Landesbibliothek Bonn}
}

License

Aira-OPT-350M is licensed under the OPT-175B License Agreement, Copyright (c) Meta Platforms, Inc. All Rights Reserved. See the LICENSE file for more details.