🚀 LaMini-Flan-T5-783M

This model is part of the LaMini-LM model series, fine - tuned for better text - to - text generation performance.

🚀 Quick Start

This model is one of our LaMini-LM model series presented in the paper "LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions". It is a fine - tuned version of google/flan-t5-large on the LaMini-instruction dataset, which contains 2.58M samples for instruction fine - tuning. For more information about our dataset, please refer to our project repository.

You can view other models of the LaMini-LM series as follows. Models marked with ✩ have the best overall performance given their size/architecture, so we recommend using them. More details can be found in our paper.

✨ Features

• Based on the google/flan-t5-large model.
• Fine - tuned on the LaMini-instruction dataset with 2.58M samples.
• Suitable for responding to human instructions in natural language.

💻 Usage Examples

Basic Usage

# pip install -q transformers
from transformers import pipeline

checkpoint = "{model_name}"

model = pipeline('text2text-generation', model = checkpoint)

input_prompt = 'Please let me know your thoughts on the given place and why you think it deserves to be visited: \n"Barcelona, Spain"'
generated_text = model(input_prompt, max_length=512, do_sample=True)[0]['generated_text']

print("Response", generated_text)

🔧 Technical Details

Training Procedure

We initialize with google/flan-t5-large and fine - tune it on our LaMini-instruction dataset. Its total number of parameters is 783M.

Training Hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0005
• train_batch_size: 128
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 512
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
• lr_scheduler_type: linear
• num_epochs: 5

📚 Documentation

We conducted two sets of evaluations: automatic evaluation on downstream NLP tasks and human evaluation on user - oriented instructions. For more detail, please refer to our paper.

📄 License

This model is licensed under CC By NC 4.0.

Citation

@article{lamini-lm,
  author       = {Minghao Wu and
                  Abdul Waheed and
                  Chiyu Zhang and
                  Muhammad Abdul-Mageed and
                  Alham Fikri Aji
                  },
  title        = {LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions},
  journal      = {CoRR},
  volume       = {abs/2304.14402},
  year         = {2023},
  url          = {https://arxiv.org/abs/2304.14402},
  eprinttype   = {arXiv},
  eprint       = {2304.14402}
}