esc50-wav2vec2-attn

This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6556
• Accuracy: 0.875
• F1 Macro: 0.8752

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 450
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro
3.7226	1.0	225	3.6235	0.205	0.1414
3.0002	2.0	450	2.8299	0.435	0.3782
2.2889	3.0	675	2.0599	0.585	0.5339
1.6197	4.0	900	1.5775	0.7	0.66
1.1835	5.0	1125	1.3031	0.72	0.7070
0.7858	6.0	1350	1.2474	0.7	0.6953
0.5843	7.0	1575	0.9818	0.76	0.7385
0.4295	8.0	1800	0.8253	0.8	0.7958
0.3041	9.0	2025	0.8176	0.8	0.7926
0.2178	10.0	2250	0.8450	0.795	0.7861
0.1874	11.0	2475	0.7450	0.81	0.8045
0.1225	12.0	2700	0.7663	0.845	0.8409
0.0818	13.0	2925	0.7127	0.855	0.8531
0.0874	14.0	3150	0.7242	0.84	0.8396
0.0469	15.0	3375	0.6220	0.855	0.8562
0.0531	16.0	3600	0.5916	0.875	0.8743
0.0351	17.0	3825	0.6738	0.85	0.8485
0.0205	18.0	4050	0.6656	0.865	0.8666
0.0207	19.0	4275	0.6556	0.875	0.8752
0.0194	20.0	4500	0.6624	0.875	0.8752

Framework versions

• Transformers 4.56.1
• Pytorch 2.8.0+cu128
• Datasets 2.19.0
• Tokenizers 0.22.0