LaMini-T5-223M / README.md
afaji's picture
auto-rename
b56df50
|
raw
history blame
6.2 kB
metadata
license: cc-by-nc-4.0
tags:
  - generated_from_trainer
  - instruction fine-tuning
model-index:
  - name: flan-t5-small-distil-v2
    results: []
language:
  - en
pipeline_tag: text2text-generation
widget:
  - text: how can I become more healthy?
    example_title: example

Title

LaMini-T5-223M

Model License

This model is one of our LaMini-LM model series in paper "LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions". This model is a fine-tuned version of t5-base on LaMini-instruction dataset that contains 2.58M samples for instruction fine-tuning. For more information about our dataset, please refer to our project repository.
You can view other LaMini-LM model series as follow. Note that not all models are performing as well. Models with ✩ are those with the best overall performance given their size/architecture. More details can be seen in our paper.

Base model LaMini-LM series (#parameters)
T5 LaMini-T5-61M LaMini-T5-223M LaMini-T5-738M
Flan-T5 LaMini-Flan-T5-77M LaMini-Flan-T5-248M LaMini-Flan-T5-783M
Cerebras-GPT LaMini-Cerebras-111M LaMini-Cerebras-256M LaMini-Cerebras-590M LaMini-Cerebras-1.3B
GPT-2 LaMini-GPT-124M LaMini-GPT-774M LaMini-GPT-1.5B
GPT-Neo LaMini-Neo-125M LaMini-Neo-1.3B
GPT-J coming soon
LLaMA coming soon

Use

Intended use

We recommend using the model to response to human instructions written in natural language.

We now show you how to load and use our model using HuggingFace pipline().

# pip install -q transformers
from transformers import pipeline

checkpoint = "{model_name}"

model = pipeline('text2text-generation', model=checkpoint, use_auth_token=True, device=0)

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 = generator(input_prompt, max_length=512, do_sample=True)[0]['generated_text']

print("Response": generated_text)

Training Procedure

Title

We initialize with t5-base and fine-tune it on our LaMini-instruction dataset. Its total number of parameters is 223M.

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

Evaluation

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.

Limitations

More information needed

Citation

@misc{lamini,
      title={LaMini-LM: A Diverse Herd of Distilled Models from Large-Scale Instructions}, 
      author={Minghao Wu and Abdul Waheed and Chiyu Zhang and Muhammad Abdul-Mageed and Alham Fikri Aji},
      year={2023},
      publisher = {GitHub},
      journal = {GitHub repository},
}