roberta-hate-speech-detection

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

• Loss: 0.4354
• Accuracy: 0.808
• Auc: 0.898
• Precision: 0.8081
• Recall: 0.8077
• F1: 0.8078

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: 1e-05
• train_batch_size: 64
• eval_batch_size: 64
• 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
• lr_scheduler_warmup_steps: 500
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Auc	Precision	Recall	F1
0.691	1.0	188	0.6680	0.626	0.662	0.6289	0.6265	0.6222
0.6036	2.0	376	0.5378	0.726	0.807	0.7273	0.7257	0.7245
0.5107	3.0	564	0.5570	0.742	0.85	0.7668	0.7417	0.7371
0.4531	4.0	752	0.4833	0.778	0.88	0.7862	0.7783	0.7759
0.4077	5.0	940	0.4477	0.81	0.89	0.8131	0.8103	0.8101
0.3567	6.0	1128	0.4229	0.832	0.902	0.8316	0.8316	0.8316
0.3202	7.0	1316	0.4174	0.827	0.907	0.8273	0.8269	0.8269
0.299	8.0	1504	0.4531	0.822	0.909	0.8262	0.8222	0.8220
0.2625	9.0	1692	0.4289	0.839	0.912	0.8390	0.8389	0.8389
0.2457	10.0	1880	0.4246	0.846	0.915	0.8457	0.8455	0.8456
0.2173	11.0	2068	0.4783	0.844	0.914	0.8435	0.8435	0.8435
0.1956	12.0	2256	0.4893	0.845	0.915	0.8479	0.8449	0.8448
0.1761	13.0	2444	0.5208	0.837	0.914	0.8420	0.8369	0.8366
0.1627	14.0	2632	0.5077	0.842	0.918	0.8427	0.8415	0.8416
0.1482	15.0	2820	0.5581	0.835	0.917	0.8408	0.8349	0.8345
0.1437	16.0	3008	0.5135	0.854	0.921	0.8545	0.8542	0.8542
0.1315	17.0	3196	0.5428	0.846	0.921	0.8492	0.8462	0.8461
0.1209	18.0	3384	0.5382	0.853	0.921	0.8530	0.8529	0.8529
0.1186	19.0	3572	0.5839	0.844	0.92	0.8459	0.8435	0.8435
0.105	20.0	3760	0.5757	0.845	0.921	0.8468	0.8449	0.8448

Framework versions

• Transformers 4.52.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1