liberta-large-upos

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

• Loss: 0.3346
• Precision: 0.8101
• Recall: 0.7466
• F1: 0.7542
• Accuracy: 0.8675

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	338	1.0412	0.5939	0.4306	0.4617	0.5790
No log	2.0	676	0.6850	0.6114	0.5788	0.5745	0.7115
No log	3.0	1014	0.6075	0.6787	0.6205	0.6241	0.7389
No log	4.0	1352	0.5585	0.7178	0.6393	0.6425	0.7608
No log	5.0	1690	0.4762	0.7424	0.6737	0.6874	0.7984
No log	6.0	2028	0.4203	0.7159	0.6962	0.6946	0.8228
No log	7.0	2366	0.4275	0.7403	0.7081	0.7028	0.8205
No log	8.0	2704	0.3789	0.7909	0.7189	0.7282	0.8470
No log	9.0	3042	0.3431	0.8051	0.7415	0.7484	0.8626
No log	10.0	3380	0.3346	0.8101	0.7466	0.7542	0.8675

Framework versions

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