ai-light-dance_stepmania_ft_wav2vec2-large-xlsr-53-v7

This model is a fine-tuned version of gary109/ai-light-dance_stepmania_ft_wav2vec2-large-xlsr-53-v6 on the GARY109/AI_LIGHT_DANCE - ONSET-STEPMANIA2 dataset. It achieves the following results on the evaluation set:

• Loss: 1.0424
• Wer: 0.6512

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: 4e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9303	1.0	12031	1.1160	0.6712
0.8181	2.0	24062	1.0601	0.6608
0.7861	3.0	36093	1.0478	0.6520
0.767	4.0	48124	1.0617	0.6526
0.797	5.0	60155	1.0424	0.6512
0.834	6.0	72186	1.0519	0.6542
0.7915	7.0	84217	1.0508	0.6494
0.8106	8.0	96248	1.0753	0.6449
0.7512	9.0	108279	1.1223	0.6592
0.777	10.0	120310	1.1201	0.6535
0.7631	11.0	132341	1.0780	0.6512
0.7465	12.0	144372	1.0822	0.6499
0.826	13.0	156403	1.0706	0.6445
0.7552	14.0	168434	1.0862	0.6449
0.8279	15.0	180465	1.1162	0.6461
0.7769	16.0	192496	1.1023	0.6420
0.7918	17.0	204527	1.1085	0.6456
0.6941	18.0	216558	1.1139	0.6417
0.7379	19.0	228589	1.1126	0.6410
0.7467	20.0	240620	1.1102	0.6369
0.8045	21.0	252651	1.1191	0.6376
0.7059	22.0	264682	1.1285	0.6381
0.7008	23.0	276713	1.1328	0.6377
0.7816	24.0	288744	1.1326	0.6366
0.7426	25.0	300775	1.1420	0.6362
0.7226	26.0	312806	1.1326	0.6350
0.665	27.0	324837	1.1419	0.6346
0.7184	28.0	336868	1.1480	0.6346
0.77	29.0	348899	1.1476	0.6343
0.727	30.0	360930	1.1494	0.6348

Framework versions

• Transformers 4.21.0.dev0
• Pytorch 1.9.1+cu102
• Datasets 2.3.3.dev0
• Tokenizers 0.12.1