Distilbert is a variant of bert model(one of LLM models). This model with a classification head is used to classify the emotions of the input tweet. This model is a fine-tuned version of distilbert-base-uncased on the emotion dataset. It achieves the following results on the evaluation set:

Loss: 0.2195
Accuracy: 0.9235
F1: 0.9233

Emotion Labels

label_0: Sadness
label_1: Joy
label_2: Love
label_3: Anger
label_4: Fear
label_5: Surprise

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 64
eval_batch_size: 64
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.8537	1.0	250	0.3235	0.897	0.8958
0.2506	2.0	500	0.2195	0.9235	0.9233

Validation metrics

test_loss : 0.2194512039422989
test_accuracy : 0.9235
test_f1 : 0.923296474937779

Framework versions

Transformers 4.37.2
Pytorch 2.1.0+cu121
Datasets 2.16.1
Tokenizers 0.15.1

pt-sk/distilbert-finetuned-emotion

Distilbert-finetuned-emotion