cdp-multi-classifier-sub-classes-weighted

This model is a fine-tuned version of alex-miller/ODABert. It achieves the following results on the evaluation set:

• Loss: 0.6494
• Accuracy: 0.8423
• F1: 0.7847
• Precision: 0.7488
• Recall: 0.8243

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.8824	1.0	206	0.9109	0.6793	0.5494	0.5385	0.5607
0.816	2.0	412	0.8348	0.7865	0.7127	0.6712	0.7597
0.7455	3.0	618	0.7694	0.7712	0.7053	0.64	0.7855
0.6808	4.0	824	0.7323	0.7874	0.7217	0.6638	0.7907
0.6246	5.0	1030	0.6920	0.7829	0.7175	0.6567	0.7907
0.5689	6.0	1236	0.6926	0.7874	0.7230	0.6624	0.7959
0.5237	7.0	1442	0.6642	0.8072	0.7476	0.6876	0.8191
0.4842	8.0	1648	0.6419	0.8081	0.7461	0.6925	0.8088
0.4519	9.0	1854	0.6498	0.8225	0.7612	0.7169	0.8114
0.4264	10.0	2060	0.6496	0.8315	0.7728	0.7294	0.8217
0.4078	11.0	2266	0.6669	0.8378	0.7800	0.7401	0.8243
0.398	12.0	2472	0.6508	0.8405	0.7828	0.7453	0.8243
0.3808	13.0	2678	0.6539	0.8387	0.7809	0.7419	0.8243
0.3757	14.0	2884	0.6497	0.8423	0.7847	0.7488	0.8243
0.3691	15.0	3090	0.6494	0.8423	0.7847	0.7488	0.8243

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1