294134d60f960e3ddb861591fa506a0a

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

• Loss: 1.5579
• Data Size: 1.0
• Epoch Runtime: 26.9562
• Accuracy: 0.3488
• F1 Macro: 0.0862

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	1.8648	0	1.7515	0.0791	0.0245
No log	1	500	1.6179	0.0078	1.8615	0.2908	0.0993
No log	2	1000	1.5864	0.0156	2.0138	0.3468	0.0911
No log	3	1500	1.5779	0.0312	2.5964	0.3488	0.0862
No log	4	2000	1.5885	0.0625	3.3856	0.2908	0.0751
0.0861	5	2500	1.5811	0.125	4.9078	0.2908	0.0751
1.6012	6	3000	1.5865	0.25	7.8747	0.2908	0.0751
0.2596	7	3500	1.5666	0.5	13.7990	0.3488	0.0862
1.5902	8.0	4000	1.5631	1.0	27.1187	0.3488	0.0862
1.5901	9.0	4500	1.5609	1.0	26.6384	0.3488	0.0862
1.5968	10.0	5000	1.5628	1.0	25.9409	0.3488	0.0862
1.5911	11.0	5500	1.5604	1.0	26.4676	0.3488	0.0862
1.5628	12.0	6000	1.5632	1.0	25.3616	0.3488	0.0862
1.5652	13.0	6500	1.5595	1.0	25.7595	0.3488	0.0862
1.5853	14.0	7000	1.5622	1.0	27.5021	0.3488	0.0862
1.574	15.0	7500	1.5668	1.0	27.0713	0.2908	0.0751
1.5773	16.0	8000	1.5570	1.0	28.2524	0.3488	0.0862
1.57	17.0	8500	1.5597	1.0	27.0000	0.3488	0.0862
1.5923	18.0	9000	1.5590	1.0	27.5061	0.3488	0.0862
1.5957	19.0	9500	1.5580	1.0	27.9192	0.3488	0.0862
1.5857	20.0	10000	1.5579	1.0	26.9562	0.3488	0.0862

Framework versions

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