Modelpublic

Ghazi-nak/6_BiasDetctor_HTW_ANG.KI

sourceHugging Facemitupdated 1y agoView on Hugging Face

0likes2downloads

Model Card

6BiasDetctorHTW_ANG.KI

This model is a fine-tuned version of roberta-base on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.5508
—Accuracy: 0.9314
—F1: 0.9313
—Precision: 0.9313
—Recall: 0.9314

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
—trainbatchsize: 16
—evalbatchsize: 32
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 32
—optimizer: Use adamwtorch with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: cosine
—lrschedulerwarmup_steps: 300
—num_epochs: 6
—labelsmoothingfactor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
1.0529	0.4	250	0.7700	0.8142	0.8169	0.8336	0.8142
0.6938	0.8	500	0.6479	0.8762	0.8776	0.8854	0.8762
0.6216	1.2	750	0.6049	0.901	0.9020	0.9060	0.901
0.5676	1.6	1000	0.6214	0.8978	0.8979	0.9005	0.8978
0.5777	2.0	1250	0.5493	0.9236	0.9241	0.9252	0.9236
0.5118	2.4	1500	0.5507	0.9236	0.9236	0.9243	0.9236
0.4909	2.8	1750	0.5579	0.927	0.9269	0.9272	0.927
0.4595	3.2	2000	0.5508	0.9314	0.9313	0.9313	0.9314
0.4538	3.6	2250	0.5691	0.9244	0.9239	0.9248	0.9244
0.4445	4.0	2500	0.5581	0.9308	0.9308	0.9314	0.9308

Framework versions

—Transformers 4.51.3
—Pytorch 2.1.2
—Datasets 3.5.1
—Tokenizers 0.21.1

Training Information:

TrainOutput(globalstep=2500, trainingloss=0.6116815521240234, metrics={'trainruntime': 805.1081, 'trainsamplespersecond': 149.048, 'trainstepspersecond': 4.658, 'totalflos': 2.104945139712e+16, 'train_loss': 0.6116815521240234, 'epoch': 4.0})