wav2vec2-base-mirst500-ac

This model is a fine-tuned version of facebook/wav2vec2-base on the /workspace/datasets/datasets/MIR_ST500/MIR_ST500.py dataset. It achieves the following results on the evaluation set:

• Loss: 0.7566
• Accuracy: 0.7570

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: 16
• eval_batch_size: 1
• seed: 0
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.3718	1.0	1304	1.4422	0.4255
1.1285	2.0	2608	1.1061	0.5869
1.0275	3.0	3912	0.8825	0.6724
0.9982	4.0	5216	0.9181	0.6713
0.9482	5.0	6520	0.8717	0.6971
0.8687	6.0	7824	0.8041	0.7164
0.8841	7.0	9128	0.8869	0.7034
0.8094	8.0	10432	0.8216	0.7172
0.7733	9.0	11736	0.8018	0.7298
0.7892	10.0	13040	0.7517	0.7426
0.8736	11.0	14344	0.7482	0.7482
0.7035	12.0	15648	0.7730	0.7488
0.7361	13.0	16952	0.7677	0.7510
0.7808	14.0	18256	0.7765	0.7512
0.7359	15.0	19560	0.7566	0.7570

Framework versions

• Transformers 4.18.0
• Pytorch 1.9.1+cu102
• Datasets 2.0.0
• Tokenizers 0.11.6