c455795be7703aa4e0d22c15ae827454

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

• Loss: 0.3005
• Data Size: 1.0
• Epoch Runtime: 32.6538
• Accuracy: 0.9071
• F1 Macro: 0.7635

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	1.0638	0	2.8418	0.3620	0.2228
No log	1	619	0.6880	0.0078	3.3688	0.7672	0.2894
No log	2	1238	0.5605	0.0156	3.4645	0.7819	0.3559
0.0141	3	1857	0.4067	0.0312	4.2954	0.8563	0.5462
0.0141	4	2476	0.4780	0.0625	5.0474	0.8415	0.5114
0.3486	5	3095	0.3515	0.125	6.9191	0.8744	0.6138
0.0297	6	3714	0.3119	0.25	10.5020	0.8949	0.5955
0.3176	7	4333	0.3700	0.5	17.8857	0.8851	0.6627
0.3165	8.0	4952	0.2777	1.0	35.1093	0.9038	0.6418
0.2417	9.0	5571	0.2891	1.0	32.3269	0.9018	0.7576
0.2585	10.0	6190	0.2857	1.0	33.2657	0.9032	0.7057
0.2506	11.0	6809	0.2890	1.0	31.8564	0.9040	0.7537
0.2192	12.0	7428	0.3005	1.0	32.6538	0.9071	0.7635

Framework versions

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