Simple Transformers

This library is based on the Transformers library by HuggingFace. Simple Transformers lets you quickly train and evaluate Transformer models. Only 3 lines of code are needed to initialize, train, and evaluate a model.

Supported Tasks:

• Information Retrieval (Dense Retrieval)
• (Large) Language Models (Training, Fine-tuning, and Generation)
• Encoder Model Training and Fine-tuning
• Sequence Classification
• Token Classification (NER)
• Question Answering
• Language Generation
• T5 Model
• Seq2Seq Tasks
• Multi-Modal Classification
• Conversational AI.

Table of contents

• Simple Transformers
• Table of contents
  • Setup
    • With Conda
      • Optional
  • Usage
    • A quick example
    • Experiment Tracking with Weights and Biases
  • Current Pretrained Models
  • Contributors ✨
  • How to Contribute
    • How to Update Docs
  • Acknowledgements

Setup

With Conda

1. Install Anaconda or Miniconda Package Manager from here
2. Create a new virtual environment and install packages.

$ conda create -n st python pandas tqdm
$ conda activate st

Using Cuda:

$ conda install pytorch>=1.6 cudatoolkit=11.0 -c pytorch

Without using Cuda

$ conda install pytorch cpuonly -c pytorch

3. Install simpletransformers.

$ pip install simpletransformers

Optional

1. Install Weights and Biases (wandb) for tracking and visualizing training in a web browser.

$ pip install wandb

Usage

All documentation is now live at simpletransformers.ai

Simple Transformer models are built with a particular Natural Language Processing (NLP) task in mind. Each such model comes equipped with features and functionality designed to best fit the task that they are intended to perform. The high-level process of using Simple Transformers models follows the same pattern.

1. Initialize a task-specific model
2. Train the model with train_model()
3. Evaluate the model with eval_model()
4. Make predictions on (unlabelled) data with predict()

However, there are necessary differences between the different models to ensure that they are well suited for their intended task. The key differences will typically be the differences in input/output data formats and any task specific features/configuration options. These can all be found in the documentation section for each task.

The currently implemented task-specific Simple Transformer models, along with their task, are given below.

• Please refer to the relevant section in the docs for more information on how to use these models.
• Example scripts can be found in the examples directory.
• See the Changelog for up-to-date changes to the project.

A quick example

from simpletransformers.classification import ClassificationModel, ClassificationArgs
import pandas as pd
import logging


logging.basicConfig(level=logging.INFO)
transformers_logger = logging.getLogger("transformers")
transformers_logger.setLevel(logging.WARNING)

# Preparing train data
train_data = [
    ["Aragorn was the heir of Isildur", 1],
    ["Frodo was the heir of Isildur", 0],
]
train_df = pd.DataFrame(train_data)
train_df.columns = ["text", "labels"]

# Preparing eval data
eval_data = [
    ["Theoden was the king of Rohan", 1],
    ["Merry was the king of Rohan", 0],
]
eval_df = pd.DataFrame(eval_data)
eval_df.columns = ["text", "labels"]

# Optional model configuration
model_args = ClassificationArgs(num_train_epochs=1)

# Create a ClassificationModel
model = ClassificationModel(
    "roberta", "roberta-base", args=model_args
)

# Train the model
model.train_model(train_df)

# Evaluate the model
result, model_outputs, wrong_predictions = model.eval_model(eval_df)

# Make predictions with the model
predictions, raw_outputs = model.predict(["Sam was a Wizard"])

Experiment Tracking with Weights and Biases

• Weights and Biases makes it incredibly easy to keep track of all your experiments. Check it out on Colab here:

Current Pretrained Models

For a list of pretrained models, see Hugging Face docs.

The model_types available for each task can be found under their respective section. Any pretrained model of that type found in the Hugging Face docs should work. To use any of them set the correct model_type and model_name in the args dictionary.

Contributors ✨

Thanks goes to these wonderful people (emoji key):

hawktang 💻	Mabu Manaileng 💻	Ali Hamdi Ali Fadel 💻	Tovly Deutsch 💻	hlo-world 💻	huntertl 💻	Yann Defretin 💻 📖 💬 🤔
Manuel 📖 💻	Gilles Jacobs 📖	shasha79 💻	Mercedes Garcia 💻	Hammad Hassan Tarar 💻 📖	Todd Cook 💻	Knut O. Hellan 💻 📖
nagenshukla 💻	flaviussn 💻 📖	Marc Torrellas 🚧	Adrien Renaud 💻	jacky18008 💻	Matteo Senese 💻