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

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: 0.5678
• Accuracy: 0.87

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.8868	1.0	50	0.8761	0.72
0.4771	2.0	100	0.7632	0.76
0.3415	3.0	150	1.0356	0.72
0.2508	4.0	200	0.5432	0.82
0.1699	5.0	250	0.6632	0.81
0.024	6.0	300	0.8745	0.82
0.0353	7.0	350	0.8643	0.79
0.0341	8.0	400	0.5614	0.86
0.0411	9.0	450	0.6230	0.86
0.0345	10.0	500	0.9361	0.76
0.0304	11.0	550	0.6329	0.87
0.0504	12.0	600	1.0623	0.81
0.0526	13.0	650	0.7261	0.83
0.0007	14.0	700	0.8432	0.8
0.0041	15.0	750	0.8342	0.86
0.0002	16.0	800	0.6246	0.89
0.0092	17.0	850	0.5784	0.89
0.0001	18.0	900	0.6059	0.87
0.0001	19.0	950	0.5561	0.86
0.0001	20.0	1000	0.5483	0.85
0.0192	21.0	1050	0.5678	0.87

Framework versions

• Transformers 4.55.0
• Pytorch 2.6.0+cu124
• Datasets 2.16.0
• Tokenizers 0.21.4