Modelpublic

ramesh070/mbert-hatespeechdetection-malayalam

sourceHugging Faceapache-2.0updated 10mo agoView on Hugging Face

0likes88downloads

Model Card

mbert-hatespeechdetection-malayalam

This model is a fine-tuned version of google-bert/bert-base-multilingual-cased on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.4429
—Accuracy: 0.9318
—F1: 0.9375

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: 32
—evalbatchsize: 32
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.5749	1.0	23	0.4325	0.7955	0.7805
0.2836	2.0	46	0.2142	0.9091	0.9111
0.1191	3.0	69	0.2161	0.9432	0.9474
0.0512	4.0	92	0.5017	0.8864	0.8864
0.0434	5.0	115	0.4366	0.8977	0.8989
0.0417	6.0	138	0.3117	0.9091	0.9149
0.0096	7.0	161	0.8158	0.8636	0.8605
0.0167	8.0	184	0.5991	0.8977	0.8989
0.0222	9.0	207	0.5119	0.9091	0.9111
0.0006	10.0	230	0.3724	0.9318	0.9375
0.0062	11.0	253	0.3441	0.9545	0.9592
0.0034	12.0	276	0.3466	0.9318	0.9375
0.0004	13.0	299	0.4061	0.9432	0.9474
0.0037	14.0	322	0.3898	0.9432	0.9474
0.0003	15.0	345	0.4712	0.9318	0.9362
0.0003	16.0	368	0.4017	0.9318	0.9388
0.0003	17.0	391	0.4050	0.9318	0.9388
0.0003	18.0	414	0.4401	0.9432	0.9474
0.0002	19.0	437	0.4425	0.9318	0.9375
0.0002	20.0	460	0.4429	0.9318	0.9375

Framework versions

—Transformers 4.53.3
—Pytorch 2.9.1+cu128
—Datasets 4.4.1
—Tokenizers 0.21.2