distilbert-base-multilingual-cased-finetuned

This model is a fine-tuned version of distilbert/distilbert-base-multilingual-cased on Arabic tweets for Emotion detection dataset. It achieves the following results on the evaluation set:

• Loss: 1.6740
• Accuracy: 0.6643
• F1: 0.6611

Model description

The model has been trained to classify text inputs into distinct emotional categories based on the fine-tuned understanding of the emotions dataset. The fine-tuned model has demonstrated high accuracy and F1 scores on the evaluation set.

Intended uses & limitations

Intended Uses

• Sentiment analysis
• Emotional classification in text
• Emotion-based recommendation systems

Limitations

• May show biases based on the training dataset
• Optimized for emotional classification and may not cover nuanced emotional subtleties

Training and evaluation data

Emotions dataset with labeled emotional categories here.

The emotional categories are as follows:

• LABEL_0 : none
• LABEL_1 : anger
• LABEL_2 : joy
• LABEL_3 : sadness
• LABEL_4 : love
• LABEL_5 : sympathy
• LABEL_6 : surprise
• LABEL_7 : fear

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.4725	1.0	252	1.0892	0.6604	0.6625
0.3392	2.0	504	1.2096	0.6594	0.6649
0.2575	3.0	756	1.2745	0.6723	0.6706
0.1979	4.0	1008	1.3719	0.6713	0.6666
0.1757	5.0	1260	1.4239	0.6723	0.6652
0.1414	6.0	1512	1.5074	0.6663	0.6666
0.1073	7.0	1764	1.5703	0.6783	0.6722
0.0812	8.0	2016	1.6218	0.6673	0.6638
0.0615	9.0	2268	1.6676	0.6693	0.6642
0.0531	10.0	2520	1.6740	0.6643	0.6611

Framework versions

• Transformers 4.42.4
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1