d08a20d0d7310ba69cc96f2881b53f1a

This model is a fine-tuned version of google-bert/bert-base-german-dbmdz-uncased on the nyu-mll/glue [sst2] dataset. It achieves the following results on the evaluation set:

• Loss: 0.4709
• Data Size: 1.0
• Epoch Runtime: 98.8009
• Accuracy: 0.875
• F1 Macro: 0.8747
• Rouge1: 0.875
• Rouge2: 0.0
• Rougel: 0.875
• Rougelsum: 0.875

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	0.7911	0	0.9612	0.4954	0.4814	0.4954	0.0	0.4954	0.4954
No log	1	2104	0.6701	0.0078	2.7753	0.5521	0.4495	0.5521	0.0	0.5521	0.5521
No log	2	4208	0.5469	0.0156	2.7148	0.7350	0.7342	0.7350	0.0	0.7361	0.7338
0.0132	3	6312	0.4673	0.0312	4.3283	0.7859	0.7858	0.7859	0.0	0.7870	0.7859
0.4989	4	8416	0.4345	0.0625	7.5204	0.7928	0.7927	0.7928	0.0	0.7940	0.7917
0.3679	5	10520	0.5876	0.125	13.4831	0.7731	0.7640	0.7731	0.0	0.7743	0.7731
0.2723	6	12624	0.3147	0.25	26.6255	0.8600	0.8600	0.8611	0.0	0.8600	0.8600
0.2531	7	14728	0.3042	0.5	49.3163	0.8692	0.8692	0.8698	0.0	0.8692	0.8692
0.1801	8.0	16832	0.3356	1.0	99.7133	0.8773	0.8772	0.8773	0.0	0.8773	0.8773
0.1487	9.0	18936	0.3901	1.0	98.9114	0.8715	0.8715	0.8715	0.0	0.8715	0.8715
0.1431	10.0	21040	0.3970	1.0	98.0861	0.8542	0.8535	0.8542	0.0	0.8542	0.8542
0.0977	11.0	23144	0.4709	1.0	98.8009	0.875	0.8747	0.875	0.0	0.875	0.875

Framework versions

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