rubert-base-cased-finetuned-panx-de

This model is a fine-tuned version of DeepPavlov/rubert-base-cased on the universal_dependencies dataset. It achieves the following results on the evaluation set:

• Loss: 1.3352
• F1: 0.6979

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: 24
• eval_batch_size: 24
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	F1
1.3039	1.0	131	1.0720	0.5871
0.7911	2.0	262	0.9581	0.6335
0.5448	3.0	393	0.9162	0.6718
0.3709	4.0	524	0.9722	0.6819
0.2565	5.0	655	1.0484	0.6847
0.1923	6.0	786	1.1410	0.6924
0.145	7.0	917	1.1692	0.6995
0.1104	8.0	1048	1.2592	0.6981
0.086	9.0	1179	1.3307	0.6972
0.0715	10.0	1310	1.3352	0.6979

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu117
• Datasets 2.7.1
• Tokenizers 0.13.2