f7daaa25c652c1fb6d9c584259ec9fe6

This model is a fine-tuned version of FacebookAI/xlm-roberta-base on the contemmcm/hate-speech-and-offensive-language dataset. It achieves the following results on the evaluation set:

• Loss: 0.2751
• Data Size: 1.0
• Epoch Runtime: 54.6618
• Accuracy: 0.9121
• F1 Macro: 0.7217

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	1.2438	0	4.3611	0.0601	0.0378
No log	1	619	0.7204	0.0078	5.2062	0.7672	0.2894
No log	2	1238	0.7711	0.0156	5.4810	0.7672	0.2894
0.0183	3	1857	0.7181	0.0312	6.6856	0.7672	0.2894
0.0183	4	2476	0.6793	0.0625	8.1599	0.7672	0.2894
0.4038	5	3095	0.3618	0.125	11.5910	0.8819	0.5824
0.029	6	3714	0.3063	0.25	18.3590	0.9026	0.6434
0.3075	7	4333	0.2901	0.5	31.3736	0.8918	0.7520
0.2901	8.0	4952	0.2631	1.0	58.7930	0.9099	0.6133
0.2334	9.0	5571	0.3086	1.0	58.1872	0.8933	0.6997
0.2495	10.0	6190	0.2807	1.0	56.5037	0.9026	0.7346
0.241	11.0	6809	0.2950	1.0	57.0391	0.9022	0.7476
0.207	12.0	7428	0.2751	1.0	54.6618	0.9121	0.7217

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1