7eafca75795c71c726d8d69f8fe499f8

This model is a fine-tuned version of google-bert/bert-base-german-dbmdz-uncased on the dair-ai/emotion [split] dataset. It achieves the following results on the evaluation set:

• Loss: 0.2671
• Data Size: 1.0
• Epoch Runtime: 25.6516
• Accuracy: 0.9239
• F1 Macro: 0.8823

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	1.8843	0	1.5484	0.1845	0.1069
No log	1	500	1.6028	0.0078	1.9859	0.3488	0.0862
No log	2	1000	1.5998	0.0156	2.1769	0.3488	0.0862
No log	3	1500	1.4984	0.0312	2.7616	0.3805	0.1266
No log	4	2000	1.2313	0.0625	3.6272	0.5509	0.2242
0.075	5	2500	1.0773	0.125	5.2694	0.5912	0.2500
0.7431	6	3000	0.6117	0.25	8.3518	0.7651	0.7026
0.0497	7	3500	0.2717	0.5	14.5855	0.9017	0.8581
0.2089	8.0	4000	0.2319	1.0	27.7481	0.9168	0.8691
0.1607	9.0	4500	0.2823	1.0	27.5021	0.9103	0.8733
0.1584	10.0	5000	0.1857	1.0	26.1921	0.9274	0.8818
0.1133	11.0	5500	0.2186	1.0	27.4286	0.9234	0.8799
0.1276	12.0	6000	0.2563	1.0	25.7610	0.9189	0.8685
0.1153	13.0	6500	0.2867	1.0	24.9254	0.9173	0.8698
0.1071	14.0	7000	0.2671	1.0	25.6516	0.9239	0.8823

Framework versions

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