metadata

license: mit
tags:
  - generated_from_trainer
model-index:
  - name: mlcovid19-classifier
    results: []

mlcovid19-classifier

Mulit-lingual COVID-19 Fake News Detection and Intervention

This model is a fine-tuned version of xlm-roberta-base on Multi-lingual COVID19 Fake News dataset. Please visite our project website for more info. It achieves the following results on the evaluation set:

Loss: 0.4116
F1 Macro: 0.6750
F1 Misinformation: 0.9407
F1 Factual: 0.8529
F1 Other: 0.2315
Prec Macro: 0.7057
Prec Misinformation: 0.9229
Prec Factual: 0.8958
Prec Other: 0.2983

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

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
gradient_accumulation_steps: 8
total_train_batch_size: 256
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 4367
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Macro	F1 Misinformation	F1 Factual	F1 Other	Prec Macro	Prec Misinformation	Prec Factual	Prec Other
0.8111	3.67	500	0.4101	0.5506	0.9162	0.7356	0.0	0.5421	0.8969	0.7295	0.0
0.3688	7.35	1000	0.3397	0.5770	0.9321	0.7988	0.0	0.5694	0.9111	0.7972	0.0
0.3012	11.03	1500	0.3011	0.5912	0.9415	0.8322	0.0	0.5955	0.9104	0.8761	0.0
0.249	14.7	2000	0.3020	0.5931	0.9404	0.8388	0.0	0.5841	0.9206	0.8317	0.0
0.1957	18.38	2500	0.3308	0.6402	0.9406	0.8433	0.1365	0.7126	0.9234	0.8445	0.3699
0.1438	22.06	3000	0.3502	0.6615	0.9406	0.8529	0.1911	0.6952	0.9283	0.8543	0.3030
0.0996	25.73	3500	0.4116	0.6750	0.9407	0.8529	0.2315	0.7057	0.9229	0.8958	0.2983
0.0657	29.41	4000	0.4413	0.6422	0.9428	0.8497	0.1342	0.7126	0.9269	0.8453	0.3655

Framework versions

Transformers 4.23.0
Pytorch 1.12.1+cu113
Datasets 2.5.2
Tokenizers 0.13.1