6ded5be0b5bd3b983c2c5cb6e6ee578e

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

• Loss: 0.6007
• Data Size: 1.0
• Epoch Runtime: 16.3879
• Accuracy: 0.8692
• F1 Macro: 0.8660

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	3.0066	0	1.6857	0.0507	0.0052
No log	1	499	3.0064	0.0078	2.1331	0.0527	0.0091
0.0301	2	998	3.0007	0.0156	1.9626	0.0718	0.0260
0.0542	3	1497	2.9875	0.0312	2.2697	0.0680	0.0251
0.1019	4	1996	2.6253	0.0625	2.7208	0.1895	0.1237
2.2206	5	2495	1.7800	0.125	3.6829	0.4390	0.3664
1.363	6	2994	1.1989	0.25	5.5516	0.6182	0.5902
0.9345	7	3493	0.7742	0.5	9.1053	0.7510	0.7496
0.5742	8.0	3992	0.5342	1.0	16.1643	0.8311	0.8299
0.3926	9.0	4491	0.5097	1.0	16.2041	0.8427	0.8412
0.2894	10.0	4990	0.5172	1.0	16.5160	0.8523	0.8521
0.1945	11.0	5489	0.4764	1.0	16.2864	0.8614	0.8595
0.1595	12.0	5988	0.5472	1.0	16.2194	0.8634	0.8633
0.1502	13.0	6487	0.5436	1.0	15.9684	0.8679	0.8683
0.1328	14.0	6986	0.5681	1.0	16.3111	0.8662	0.8647
0.133	15.0	7485	0.6007	1.0	16.3879	0.8692	0.8660

Framework versions

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