robert-base-emotion

Model description:

roberta is Bert with better hyperparameter choices so they said it's Robustly optimized Bert during pretraining.

roberta-base finetuned on the emotion dataset using HuggingFace Trainer with below Hyperparameters

 learning rate 2e-5, 
 batch size 64,
 num_train_epochs=8,

Model Performance Comparision on Emotion Dataset from Twitter:

Model	Accuracy	F1 Score	Test Sample per Second
Distilbert-base-uncased-emotion	93.8	93.79	398.69
Bert-base-uncased-emotion	94.05	94.06	190.152
Roberta-base-emotion	93.95	93.97	195.639
Albert-base-v2-emotion	93.6	93.65	182.794

How to Use the model:

from transformers import pipeline
classifier = pipeline("text-classification",model='bhadresh-savani/roberta-base-emotion', return_all_scores=True)
prediction = classifier("I love using transformers. The best part is wide range of support and its easy to use", )
print(prediction)

"""
Output:
[[
{'label': 'sadness', 'score': 0.002281982684507966}, 
{'label': 'joy', 'score': 0.9726489186286926}, 
{'label': 'love', 'score': 0.021365027874708176}, 
{'label': 'anger', 'score': 0.0026395076420158148}, 
{'label': 'fear', 'score': 0.0007162453257478774}, 
{'label': 'surprise', 'score': 0.0003483477921690792}
]]
"""

Dataset:

Twitter-Sentiment-Analysis.

Training procedure

Colab Notebook follow the above notebook by changing the model name to roberta

Eval results

{
 'test_accuracy': 0.9395,
 'test_f1': 0.9397328860104454,
 'test_loss': 0.14367154240608215,
 'test_runtime': 10.2229,
 'test_samples_per_second': 195.639,
 'test_steps_per_second': 3.13
 }

Reference:

Natural Language Processing with Transformer By Lewis Tunstall, Leandro von Werra, Thomas Wolf

bhadresh-savani
/

roberta-base-emotion