CNEC1_1_Supertypes_xlm-roberta-large

This model is a fine-tuned version of FacebookAI/xlm-roberta-large on the cnec dataset. It achieves the following results on the evaluation set:

• Loss: 0.3156
• Precision: 0.8579
• Recall: 0.8890
• F1: 0.8732
• Accuracy: 0.9613

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 18

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.4473	0.85	500	0.1990	0.7879	0.8263	0.8066	0.9488
0.2061	1.7	1000	0.1800	0.8151	0.8537	0.8339	0.9544
0.1501	2.56	1500	0.1782	0.8145	0.8638	0.8384	0.9541
0.1257	3.41	2000	0.1613	0.8266	0.8767	0.8509	0.9606
0.1039	4.26	2500	0.1812	0.8359	0.8762	0.8556	0.9600
0.0859	5.11	3000	0.1949	0.8356	0.8811	0.8578	0.9594
0.0705	5.96	3500	0.1965	0.8323	0.8753	0.8533	0.9588
0.0549	6.81	4000	0.2135	0.8469	0.8899	0.8679	0.9619
0.0513	7.67	4500	0.2137	0.8488	0.8912	0.8695	0.9608
0.0374	8.52	5000	0.2099	0.8564	0.8908	0.8732	0.9625
0.0326	9.37	5500	0.2388	0.8617	0.8868	0.8741	0.9619
0.03	10.22	6000	0.2796	0.8569	0.8868	0.8716	0.9601
0.0258	11.07	6500	0.2669	0.8584	0.8899	0.8739	0.9607
0.018	11.93	7000	0.2855	0.8580	0.8815	0.8696	0.9592
0.0165	12.78	7500	0.2838	0.8612	0.8939	0.8772	0.9609
0.0133	13.63	8000	0.2903	0.8593	0.8855	0.8722	0.9605
0.0128	14.48	8500	0.3064	0.8529	0.8921	0.8721	0.9610
0.0092	15.33	9000	0.3078	0.8552	0.8904	0.8724	0.9607
0.0089	16.18	9500	0.3088	0.8570	0.8899	0.8731	0.9615
0.0077	17.04	10000	0.3099	0.8571	0.8912	0.8739	0.9612
0.0057	17.89	10500	0.3156	0.8579	0.8890	0.8732	0.9613

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0