8d035128984629b7ca05dee113e1128a

This model is a fine-tuned version of google-bert/bert-large-cased-whole-word-masking on the contemmcm/hate-speech-and-offensive-language dataset. It achieves the following results on the evaluation set:

• Loss: 0.4093
• Data Size: 1.0
• Epoch Runtime: 66.1596
• Accuracy: 0.8805
• F1 Macro: 0.5727

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	1.2504	0	4.4669	0.0895	0.0644
No log	1	619	0.6800	0.0078	5.1260	0.7672	0.2894
No log	2	1238	0.5725	0.0156	5.9639	0.7672	0.2894
0.0139	3	1857	0.3453	0.0312	8.2028	0.8914	0.5985
0.0139	4	2476	0.3144	0.0625	9.5851	0.9038	0.6046
0.3224	5	3095	0.3014	0.125	13.0323	0.8809	0.7375
0.0278	6	3714	0.2894	0.25	22.3813	0.9048	0.6080
0.2968	7	4333	0.3151	0.5	35.3351	0.8994	0.7471
0.3014	8.0	4952	0.2659	1.0	66.1613	0.9065	0.6087
0.2223	9.0	5571	0.3205	1.0	65.0570	0.8925	0.7378
0.236	10.0	6190	0.2975	1.0	68.0724	0.9083	0.7646
0.3257	11.0	6809	0.4293	1.0	66.8356	0.8803	0.5731
0.2928	12.0	7428	0.4093	1.0	66.1596	0.8805	0.5727

Framework versions

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