multilabel_transfer_learning_transformer

This model is a fine-tuned version of xlm-roberta-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0217
• F1: 0.9924
• Roc Auc: 0.9955
• Accuracy: 0.9887

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 4
• seed: 123
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 300
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Roc Auc	Accuracy
0.5454	1.0	136	0.4135	0.0125	0.5030	0.0
0.3917	2.0	272	0.3582	0.2939	0.5855	0.0338
0.3405	3.0	408	0.3048	0.4862	0.6649	0.0827
0.2918	4.0	544	0.2753	0.5913	0.7250	0.1278
0.2531	5.0	680	0.2285	0.7261	0.8065	0.2406
0.214	6.0	816	0.1971	0.7684	0.8328	0.3233
0.181	7.0	952	0.1663	0.8199	0.8624	0.4173
0.1529	8.0	1088	0.1431	0.8591	0.8905	0.4774
0.1307	9.0	1224	0.1224	0.8979	0.9260	0.6090
0.1108	10.0	1360	0.1034	0.9195	0.9329	0.6955
0.0961	11.0	1496	0.0920	0.9435	0.9553	0.7744
0.0821	12.0	1632	0.0793	0.9559	0.9627	0.8346
0.0719	13.0	1768	0.0682	0.9636	0.9732	0.8759
0.0612	14.0	1904	0.0618	0.9651	0.9760	0.8947
0.0526	15.0	2040	0.0519	0.9757	0.9796	0.9135
0.0456	16.0	2176	0.0468	0.9778	0.9835	0.9248
0.0394	17.0	2312	0.0396	0.9854	0.9885	0.9586
0.0343	18.0	2448	0.0372	0.9855	0.9911	0.9586
0.0299	19.0	2584	0.0329	0.9854	0.9885	0.9586
0.0266	20.0	2720	0.0289	0.9887	0.9932	0.9887
0.0233	21.0	2856	0.0264	0.9874	0.9919	0.9812
0.0212	22.0	2992	0.0258	0.9887	0.9932	0.9887
0.02	23.0	3128	0.0242	0.9887	0.9932	0.9887
0.0177	24.0	3264	0.0217	0.9924	0.9955	0.9887
0.0162	25.0	3400	0.0200	0.9887	0.9932	0.9887
0.0146	26.0	3536	0.0201	0.9906	0.9951	0.9887
0.0136	27.0	3672	0.0192	0.9906	0.9951	0.9887
0.0127	28.0	3808	0.0169	0.9924	0.9955	0.9887

Framework versions

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