09e4c16f1c4a7be86385c86cbfa0afc1

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

• Loss: 0.3375
• Data Size: 1.0
• Epoch Runtime: 58.2490
• Accuracy: 0.8973
• F1 Macro: 0.7428

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	1.1475	0	4.7808	0.1727	0.0982
No log	1	619	0.6676	0.0078	5.5226	0.7672	0.2894
No log	2	1238	0.6740	0.0156	5.6928	0.7672	0.2894
0.0158	3	1857	0.4742	0.0312	7.0787	0.7672	0.2894
0.0158	4	2476	0.3289	0.0625	8.9260	0.8925	0.5926
0.343	5	3095	0.2989	0.125	12.7327	0.8971	0.7067
0.0272	6	3714	0.2851	0.25	19.5891	0.9040	0.6067
0.2908	7	4333	0.2718	0.5	33.1045	0.9030	0.7598
0.258	8.0	4952	0.2572	1.0	60.1549	0.9077	0.6479
0.2172	9.0	5571	0.3295	1.0	57.9570	0.8941	0.7553
0.2085	10.0	6190	0.2959	1.0	59.0081	0.9097	0.7734
0.1947	11.0	6809	0.3498	1.0	58.8880	0.9020	0.7627
0.1638	12.0	7428	0.3375	1.0	58.2490	0.8973	0.7428

Framework versions

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