somosnlp-hackathon-2022/Detect-Acoso-Twitter-Es

Detección de acoso en Twitter Español

This model is a fine-tuned version of mrm8488/distilroberta-finetuned-tweets-hate-speech on hackathon-pln-es/Dataset-Acoso-Twitter-Es.

It achieves the following results on the evaluation set:

• Loss: 0.1628
• Accuracy: 0.9167

UNL: Universidad Nacional de Loja

Miembros del equipo:

• Anderson Quizhpe
  
• Luis Negrón
  
• David Pacheco
  
• Bryan Requenes
  
• Paul Pasaca

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6732	1.0	27	0.3797	0.875
0.5537	2.0	54	0.3242	0.9167
0.5218	3.0	81	0.2879	0.9167
0.509	4.0	108	0.2606	0.9167
0.4196	5.0	135	0.1628	0.9167

Framework versions

• Transformers 4.17.0
• Pytorch 1.10.0+cu111
• Datasets 2.0.0
• Tokenizers 0.11.6