dbcffa13576853302e4a73084abb3885

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

• Loss: 0.4149
• Data Size: 1.0
• Epoch Runtime: 9.2805
• Accuracy: 0.9167
• F1 Macro: 0.9197

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	1.8407	0	0.7917	0.1792	0.0511
No log	1	170	1.6520	0.0078	1.2602	0.2771	0.0726
No log	2	340	1.6721	0.0156	1.0915	0.3396	0.1578
No log	3	510	1.4644	0.0312	1.4953	0.4625	0.3499
No log	4	680	1.2879	0.0625	1.7882	0.55	0.4294
0.0791	5	850	0.9208	0.125	2.3879	0.6625	0.5383
0.0791	6	1020	0.5694	0.25	3.2836	0.7875	0.6602
0.6095	7	1190	0.4586	0.5	5.2947	0.8542	0.8443
0.3377	8.0	1360	0.2587	1.0	9.4271	0.9042	0.8955
0.1943	9.0	1530	0.2653	1.0	9.4969	0.9208	0.9075
0.1473	10.0	1700	0.3469	1.0	9.8595	0.9146	0.9218
0.1249	11.0	1870	0.3471	1.0	10.0873	0.9146	0.9163
0.0637	12.0	2040	0.4149	1.0	9.2805	0.9167	0.9197

Framework versions

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