BTX24/convnextv2-base-22k-224-finetuned-tekno24-upsampled-balanced

convnextv2-base-22k-224-finetuned-tekno24

This model is a fine-tuned version of facebook/convnextv2-base-22k-224 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.9781
• Accuracy: 0.5748
• F1: 0.5697
• Precision: 0.5822
• Recall: 0.5748

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: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
1.2643	0.9951	102	1.1487	0.5207	0.4764	0.4783	0.5207
1.1889	2.0	205	1.1038	0.5087	0.5191	0.5565	0.5087
1.215	2.9951	307	1.0810	0.4830	0.4795	0.5589	0.4830
1.1062	4.0	410	1.0103	0.5620	0.5281	0.5358	0.5620
1.089	4.9951	512	1.0459	0.5344	0.5440	0.5720	0.5344
1.0335	6.0	615	0.9781	0.5748	0.5697	0.5822	0.5748
1.0139	6.9951	717	0.9905	0.5592	0.5605	0.5625	0.5592
0.9047	8.0	820	0.9877	0.5629	0.5525	0.5482	0.5629
0.8856	8.9951	922	1.0060	0.5565	0.5569	0.5593	0.5565
0.8306	10.0	1025	0.9907	0.5666	0.5574	0.5531	0.5666
0.8458	10.9951	1127	1.0135	0.5500	0.5489	0.5506	0.5500
0.815	11.9415	1224	1.0185	0.5491	0.5520	0.5558	0.5491

Framework versions

• Transformers 4.42.4
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1