herbert-large-cased-upos

This model is a fine-tuned version of allegro/herbert-large-cased on the universal_dependencies dataset. It achieves the following results on the evaluation set:

• Loss: 0.0611
• Precision: 0.9166
• Recall: 0.8826
• F1: 0.8928
• Accuracy: 0.9828

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: 16
• eval_batch_size: 8
• 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	Precision	Recall	F1	Accuracy
No log	1.0	438	0.2798	0.8362	0.8222	0.8271	0.8779
No log	2.0	876	0.1613	0.9287	0.8511	0.8677	0.9240
No log	3.0	1314	0.0967	0.8845	0.8530	0.8562	0.9539
No log	4.0	1752	0.0917	0.9103	0.8461	0.8657	0.9629
No log	5.0	2190	0.0782	0.8965	0.8704	0.8764	0.9666
No log	6.0	2628	0.0766	0.8973	0.8704	0.8767	0.9691
No log	7.0	3066	0.0634	0.9171	0.8811	0.8923	0.9790
No log	8.0	3504	0.0626	0.9139	0.8909	0.8989	0.9796
No log	9.0	3942	0.0675	0.9131	0.8792	0.8893	0.9803
No log	10.0	4380	0.0611	0.9166	0.8826	0.8928	0.9828

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1