c39b2bea308e90c1a4281fa0baea94ac

This model is a fine-tuned version of google-bert/bert-base-german-dbmdz-uncased on the contemmcm/cls_20newsgroups dataset. It achieves the following results on the evaluation set:

• Loss: 0.4544
• Data Size: 1.0
• Epoch Runtime: 28.0910
• Accuracy: 0.8858
• F1 Macro: 0.8852

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	3.0495	0	2.2845	0.0496	0.0269
No log	1	499	3.0367	0.0078	2.6850	0.0554	0.0221
0.0302	2	998	2.9930	0.0156	2.8477	0.0607	0.0239
0.0543	3	1497	2.5765	0.0312	3.3265	0.2707	0.2108
0.0943	4	1996	1.5225	0.0625	4.1897	0.5817	0.5418
1.2924	5	2495	0.9832	0.125	5.8381	0.7122	0.6934
0.7966	6	2994	0.7101	0.25	9.0412	0.7797	0.7737
0.5687	7	3493	0.5397	0.5	15.3618	0.8382	0.8373
0.3657	8.0	3992	0.4601	1.0	29.1866	0.8576	0.8581
0.3011	9.0	4491	0.4187	1.0	27.7202	0.8687	0.8680
0.2244	10.0	4990	0.4308	1.0	28.1138	0.8747	0.8727
0.1735	11.0	5489	0.4215	1.0	28.7436	0.8773	0.8767
0.1248	12.0	5988	0.5024	1.0	27.9868	0.8778	0.8773
0.1357	13.0	6487	0.4544	1.0	28.0910	0.8858	0.8852

Framework versions

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