e148874ab2d85d6fb5e0f4e1f49672f3

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

• Loss: 1.2373
• Data Size: 1.0
• Epoch Runtime: 9.6657
• Accuracy: 0.7322
• F1 Macro: 0.7721
• Rouge1: 0.7330
• Rouge2: 0.0
• Rougel: 0.7330
• Rougelsum: 0.7322

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	1.6226	0	1.1950	0.2436	0.0910	0.2440	0.0	0.2436	0.2429
No log	1	178	1.4967	0.0078	2.1953	0.4034	0.2271	0.4041	0.0	0.4034	0.4034
No log	2	356	1.3842	0.0156	1.4818	0.3814	0.2244	0.3828	0.0	0.3821	0.3814
No log	3	534	1.2313	0.0312	1.8063	0.3622	0.2027	0.3629	0.0	0.3629	0.3622
No log	4	712	0.8986	0.0625	2.3543	0.6726	0.5200	0.6733	0.0	0.6733	0.6726
No log	5	890	0.8340	0.125	2.7493	0.6761	0.5244	0.6768	0.0	0.6768	0.6768
0.0605	6	1068	0.7215	0.25	3.7481	0.6989	0.6034	0.6996	0.0	0.7003	0.6996
0.644	7	1246	0.7018	0.5	5.7023	0.7259	0.7512	0.7266	0.0	0.7266	0.7259
0.5341	8.0	1424	0.6250	1.0	10.1786	0.7550	0.7778	0.7557	0.0	0.7557	0.7550
0.3954	9.0	1602	0.7780	1.0	10.2735	0.7472	0.7744	0.7479	0.0	0.7479	0.7468
0.2587	10.0	1780	0.9251	1.0	9.7202	0.7351	0.7650	0.7365	0.0	0.7358	0.7358
0.1816	11.0	1958	0.9816	1.0	9.7007	0.7379	0.7809	0.7393	0.0	0.7386	0.7386
0.1099	12.0	2136	1.2373	1.0	9.6657	0.7322	0.7721	0.7330	0.0	0.7330	0.7322

Framework versions

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