metadata

license: apache-2.0
base_model: distilbert-base-uncased
tags:
  - emotions
  - sentiment-analysis
model-index:
  - name: Distilbert-base-uncased_dair-ai_emotion
    results: []
language:
  - en
metrics:
  - accuracy
  - f1
pipeline_tag: text-classification
datasets:
  - dair-ai/emotion

Load model directly

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("Arjun4707/Distilbert-base-uncased_dair-ai_emotion")

model = AutoModelForSequenceClassification.from_pretrained("Arjun4707/Distilbert-base-uncased_dair-ai_emotion", from_tf = True)

for more check out this notebook: https://github.com/BhammarArjun/NLP/blob/main/Model_validation_distilbert_emotions.ipynb

Model description

Model takes text as input and outputs an predictions for one of the 6 emotions.

  [label_0 :'anger', label_1 : 'fear', 
   label_2 : 'joy', label_3 : 'love', 
   label_4 : 'sadness', label_5 : 'surprise']

Distilbert-base-uncased_dair-ai_emotion

This model is a fine-tuned version of distilbert-base-uncased on an dair-ai/emotion dataset. It achieves the following results on the evaluation set:

Train Loss: 0.0896
Train Accuracy: 0.9582
Validation Loss: 0.1326
Validation Accuracy: 0.9375
Epoch: 4

Intended uses & limitations

Use to identify an emotion of a user from above mentioned emotions. The statements starts with 'I' in data. Need further training

Training and evaluation data

Training data size = 16000, validation data = 2000, and test data = 2000

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': {'module': 'keras.optimizers.schedules', 'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 2000, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}, 'registered_name': None}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
training_precision: float32

Training results

Train Loss	Train Accuracy	Validation Loss	Validation Accuracy	Epoch
0.5820	0.8014	0.2002	0.9305	0
0.1598	0.9366	0.1431	0.9355	1
0.1101	0.9515	0.1390	0.9355	2
0.0896	0.9582	0.1326	0.9375	3