wavlm-base-plus-finetuned-linnut-sm

This model is a fine-tuned version of microsoft/wavlm-base-plus on the codymd/linnut_audio_sm dataset. It achieves the following results on the evaluation set:

• Loss: 1.4633
• Accuracy: 0.678
• F1: 0.5581

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Accuracy	F1	Validation Loss
2.2936	1.0	500	0.222	0.0430	2.4680
1.9148	2.0	1000	0.432	0.1492	1.9393
2.0218	3.0	1500	0.5	0.1981	1.6724
2.1636	4.0	2000	0.526	0.2262	1.6097
1.8098	5.0	2500	0.516	0.2431	2.0782
1.0826	6.0	3000	0.604	0.3281	1.3590
0.6267	7.0	3500	0.606	0.3441	1.3871
0.7986	8.0	4000	0.612	0.3829	1.4410
1.0745	9.0	4500	0.656	0.4504	1.3311
1.094	10.0	5000	0.664	0.4608	1.3141
0.9286	11.0	5500	1.2929	0.69	0.5016
1.1316	12.0	6000	1.5307	0.656	0.4794
0.1818	13.0	6500	1.3146	0.696	0.5485
0.1084	14.0	7000	1.3708	0.682	0.5621
0.3915	15.0	7500	1.4633	0.678	0.5581

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.5.0
• Tokenizers 0.21.0