0e5ddb5455f870ea142615637d9b0f13

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

• Loss: 1.2053
• Data Size: 1.0
• Epoch Runtime: 9.4127
• Accuracy: 0.7415
• F1 Macro: 0.7678
• Rouge1: 0.7422
• Rouge2: 0.0
• Rougel: 0.7415
• Rougelsum: 0.7411

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	Rouge1	Rouge2	Rougel	Rougelsum
No log	0	0	1.6862	0	1.2159	0.1392	0.0949	0.1385	0.0	0.1385	0.1392
No log	1	178	1.6112	0.0078	2.0753	0.2415	0.0800	0.2415	0.0	0.2422	0.2415
No log	2	356	1.4049	0.0156	1.6773	0.4339	0.2656	0.4339	0.0	0.4339	0.4332
No log	3	534	1.1444	0.0312	1.9260	0.4901	0.3529	0.4901	0.0	0.4911	0.4893
No log	4	712	0.8766	0.0625	2.2229	0.6683	0.5089	0.6687	0.0	0.6690	0.6690
No log	5	890	0.7523	0.125	2.7749	0.7159	0.5578	0.7166	0.0	0.7166	0.7159
0.0595	6	1068	0.6850	0.25	3.9285	0.7266	0.6059	0.7273	0.0	0.7273	0.7269
0.5914	7	1246	0.7254	0.5	5.7595	0.7273	0.7244	0.7280	0.0	0.7273	0.7273
0.4774	8.0	1424	0.5883	1.0	10.3765	0.7642	0.7649	0.7649	0.0	0.7649	0.7635
0.3485	9.0	1602	0.6933	1.0	10.2792	0.7557	0.7625	0.7564	0.0	0.7564	0.7557
0.252	10.0	1780	0.7956	1.0	9.5177	0.7528	0.7828	0.7536	0.0	0.7536	0.7528
0.1862	11.0	1958	0.9778	1.0	9.4852	0.7528	0.7808	0.7532	0.0	0.7536	0.7521
0.1215	12.0	2136	1.2053	1.0	9.4127	0.7415	0.7678	0.7422	0.0	0.7415	0.7411

Framework versions

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