README.md

metadata

library_name: transformers
license: apache-2.0
base_model: google-bert/bert-base-multilingual-cased
tags:
  - generated_from_trainer
metrics:
  - accuracy
  - f1
model-index:
  - name: mbert-hatespeechdetection-tamil
    results: []

mbert-hatespeechdetection-tamil

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: 0.5521
• Accuracy: 0.8627
• F1: 0.8293

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
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.69	1.0	13	0.6642	0.6275	0.2963
0.6359	2.0	26	0.5405	0.7647	0.76
0.506	3.0	39	0.5011	0.7843	0.7843
0.3823	4.0	52	0.4513	0.8431	0.84
0.2723	5.0	65	0.3428	0.8627	0.8511
0.1862	6.0	78	0.3705	0.8431	0.8182
0.132	7.0	91	0.4117	0.8627	0.8511
0.0629	8.0	104	0.5471	0.8627	0.8293
0.0302	9.0	117	0.6216	0.8627	0.8293
0.0083	10.0	130	0.7940	0.8431	0.8000
0.038	11.0	143	0.7867	0.8627	0.8293
0.052	12.0	156	0.7889	0.8431	0.8182
0.0179	13.0	169	1.0893	0.8039	0.8
0.0032	14.0	182	0.9028	0.8627	0.8108
0.0021	15.0	195	0.8124	0.8431	0.7895
0.0018	16.0	208	0.6030	0.8627	0.8205
0.0014	17.0	221	0.9289	0.8235	0.8085
0.0015	18.0	234	0.6005	0.8824	0.8421
0.0013	19.0	247	0.5998	0.8824	0.8421
0.0012	20.0	260	0.5521	0.8627	0.8293

Framework versions

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