AST_speechcommandsV2_final

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

• Loss: 0.4825
• Accuracy: 0.8890

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: 72
• eval_batch_size: 72
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 288
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.3557	1.0	294	0.7017	0.8354
0.1948	2.0	589	0.6838	0.8397
0.1219	3.0	884	0.5752	0.8699
0.0704	4.0	1179	0.5554	0.8675
0.0404	5.0	1473	0.5437	0.8663
0.0136	6.0	1768	0.5247	0.8759
0.0072	7.0	2063	0.5235	0.8759
0.0026	8.0	2358	0.5035	0.8859
0.0007	9.0	2652	0.4800	0.8896
0.0005	9.97	2940	0.4825	0.8890

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1