metadata

license: apache-2.0
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

LaMini-FLAN-T5-Small

This model is one of our LaMini model series in paper "LaMini: Distilling Knowledge from Large Language Models". This model is a fine-tuned version of google/flan-t5-small on LaMini dataset that contains 2.58M samples for instruction fine-tuning. For more information about our dataset, please refer to our project repository.

Training Procedure

We initialize with google/flan-t5-small and fine-tune it on our LaMini dataset. Its total number of parameters is 61M.

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.

More Models

You can download LaMini model series as follow. Note that not all models are performing as well. More details can be seen in our paper.

Click to expand

LaMini Language Models collection.
Name	Architecture	Initialization
LaMini-T5-61M	encoder-decoder	T5-small
LaMini-T5-223M	encoder-decoder	T5-base
LaMini-T5-738M	encoder-decoder	T5-large
LaMini-Flan-T5-77M	encoder-decoder	Flan-T5-small
LaMini-Flan-T5-248M	encoder-decoder	Flan-T5-base
LaMini-Flan-T5-783M	encoder-decoder	Flan-T5-large
LaMini-Cb-111M	decoder-only	Cerebras-GPT-111M
LaMini-Cb-256M	decoder-only	Cerebras-GPT-256M
LaMini-Cb-590M	decoder-only	Cerebras-GPT-590M
LaMini-Cb-1.3B	decoder-only	Cerebras-GPT-1.3B
LaMini-GPT-124M	decoder-only	GPT-2
LaMini-GPT-774M	decoder-only	GPT-2 large
LaMini-GPT-1.5B	decoder-only	GPT-2 xl

Use

Intended use

We recommend to use model to reponse to human instructions wrote in natural language.

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

CPU

Click to expand

# pip install -q transformers
from transformers import pipeline

checkpoint = "{model_name}"

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

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, repetition_penalty=1.5)[0]['generated_text']

print("Response": generated_text)

GPU

Click to expand

# 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, repetition_penalty=1.5)[0]['generated_text']

print("Response": generated_text)

Limitations

More information needed

Citation

@misc{,
      title={LaMini: Distilling Knowledge from Large Language Models}, 
      author={},
      year={2023},
      eprint={},
      archivePrefix={},
      primaryClass={}
}