zikangzheng/ast-finetuned-audioset-10-10-0.4593-gtzan-optimized

ast-finetuned-audioset-10-10-0.4593-finetuned-gtzan-optimized

This model is a fine-tuned version of MIT/ast-finetuned-audioset-10-10-0.4593 on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 0.8353
• Accuracy: 0.875
• Precision: 0.8852
• Recall: 0.875
• F1: 0.8757

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
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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: polynomial
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Accuracy	F1	Validation Loss	Precision	Recall
9.9362	1.0	22	0.1333	0.0927	2.4114	0.0742	0.1333
8.8573	2.0	44	0.3167	0.3088	2.0453	0.3933	0.3167
7.2272	3.0	66	0.6722	0.6791	1.5591	0.6966	0.6722
5.3183	4.0	88	0.7556	0.7557	1.1642	0.7839	0.7556
3.527	5.0	110	0.8111	0.8108	0.9838	0.8235	0.8111
3.074	6.0	132	0.7778	0.7820	0.9769	0.8054	0.7778
2.7507	7.0	154	0.8389	0.8384	0.8948	0.8423	0.8389
2.4184	8.0	176	0.8222	0.8230	0.9135	0.8423	0.8222
2.2014	9.0	198	0.8333	0.8351	0.8896	0.8448	0.8333
2.093	10.0	220	0.8333	0.8344	0.8813	0.8388	0.8333
2.0531	11.0	242	0.8333	0.8342	0.8742	0.8415	0.8333
2.0729	12.0	264	0.8222	0.8224	0.8972	0.8335	0.8222
2.3772	13.0	286	0.8980	0.8278	0.8375	0.8278	0.8290
2.1131	14.0	308	0.8904	0.8278	0.8408	0.8278	0.8293
2.0553	15.0	330	0.8555	0.8389	0.8546	0.8389	0.8428
2.0367	16.0	352	0.8485	0.8444	0.8533	0.8444	0.8469
2.0353	17.0	374	0.8616	0.8444	0.8509	0.8444	0.8466
2.0233	18.0	396	0.8422	0.8556	0.8600	0.8556	0.8570
2.0208	19.0	418	0.8436	0.85	0.8599	0.85	0.8524
2.0145	20.0	440	0.8409	0.8556	0.8658	0.8556	0.8587
2.0121	21.0	462	0.8345	0.8667	0.8759	0.8667	0.8694
2.0096	22.0	484	0.8434	0.8722	0.8861	0.8722	0.8753
2.0093	23.0	506	0.8341	0.8444	0.8498	0.8444	0.8460
2.0058	24.0	528	0.8301	0.8444	0.8515	0.8444	0.8465
2.0053	25.0	550	0.8269	0.8556	0.8636	0.8556	0.8575
2.0037	26.0	572	0.8362	0.8556	0.8630	0.8556	0.8577
2.0026	27.0	594	0.8334	0.8611	0.8661	0.8611	0.8628

Framework versions

• Transformers 4.57.0.dev0
• Pytorch 2.9.0.dev20250716+cu129
• Datasets 4.0.0
• Tokenizers 0.22.0