ai-light-dance_drums_ft_pretrain_wav2vec2-base-new_onset-idmt-mdb-enst-2

This model is a fine-tuned version of gary109/ai-light-dance_drums_ft_pretrain_wav2vec2-base-new_onset-idmt-mdb-2 on the GARY109/AI_LIGHT_DANCE - ONSET-IDMT-MDB-ENST-2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6379
• Wer: 0.3792

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 30
• num_epochs: 100.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4226	0.99	35	2.0435	0.4154
0.8744	1.99	70	1.7193	0.4382
0.9474	2.99	105	1.7853	0.4374
0.8316	3.99	140	1.2827	0.4306
0.8336	4.99	175	1.0676	0.4040
0.7345	5.99	210	1.5364	0.4264
0.6666	6.99	245	1.4284	0.4585
0.6677	7.99	280	0.9475	0.4003
0.6779	8.99	315	1.1172	0.4209
0.6503	9.99	350	0.8999	0.3834
0.6159	10.99	385	1.1501	0.4386
0.6831	11.99	420	1.0860	0.3825
0.5959	12.99	455	0.9410	0.4045
0.7154	13.99	490	1.0463	0.3821
0.6094	14.99	525	0.8598	0.3965
0.6929	15.99	560	0.9494	0.3931
0.7627	16.99	595	0.8060	0.3948
0.601	17.99	630	0.9890	0.3965
0.546	18.99	665	0.8059	0.3990
0.5222	19.99	700	0.6379	0.3792
0.5802	20.99	735	0.6995	0.3661
0.5731	21.99	770	0.8405	0.3606
0.5462	22.99	805	0.6667	0.3965
0.6057	23.99	840	0.8396	0.3762
0.5323	24.99	875	0.9054	0.3952
0.683	25.99	910	0.6898	0.4062
0.525	26.99	945	0.7245	0.3884
0.4885	27.99	980	0.8076	0.4049
0.4653	28.99	1015	0.8100	0.3838
0.4827	29.99	1050	0.7247	0.3863
0.4839	30.99	1085	0.7009	0.3817
0.4982	31.99	1120	0.7637	0.3914
0.6105	32.99	1155	0.7343	0.3914
0.4936	33.99	1190	0.7390	0.3762
0.4674	34.99	1225	0.6724	0.3581
0.4677	35.99	1260	0.6730	0.3488
0.516	36.99	1295	0.6956	0.3728
0.4507	37.99	1330	0.6483	0.3615
0.4207	38.99	1365	0.7718	0.3484
0.4803	39.99	1400	0.8316	0.3775
0.3946	40.99	1435	0.8322	0.3568
0.411	41.99	1470	0.9933	0.3707
0.4405	42.99	1505	0.8789	0.3943
0.5124	43.99	1540	0.9030	0.3707
0.5959	44.99	1575	0.7809	0.3948
0.3841	45.99	1610	0.7716	0.3965
0.3975	46.99	1645	0.7064	0.3931
1.4983	47.99	1680	3.2866	0.3627
0.3962	48.99	1715	0.6486	0.3648
0.4422	49.99	1750	0.8450	0.3779
0.4198	50.99	1785	0.7628	0.3564
0.3577	51.99	1820	0.7553	0.3678
0.4425	52.99	1855	0.7566	0.3716
0.3492	53.99	1890	0.7710	0.3631
0.3731	54.99	1925	0.7737	0.3627
0.3868	55.99	1960	0.7021	0.3572
0.3311	56.99	1995	0.6603	0.3518
0.3993	57.99	2030	0.6664	0.3581
0.4226	58.99	2065	0.6813	0.3551
0.4143	59.99	2100	0.6567	0.3568
0.3623	60.99	2135	0.6568	0.3454
0.3228	61.99	2170	0.7326	0.3568
0.3204	62.99	2205	0.7277	0.3640
0.377	63.99	2240	0.7145	0.3585
0.3487	64.99	2275	0.6943	0.3505
0.343	65.99	2310	0.7461	0.3395
0.3251	66.99	2345	0.7442	0.3564
0.3135	67.99	2380	0.7331	0.3530
0.381	68.99	2415	0.7306	0.3513
0.3319	69.99	2450	0.8495	0.3484
0.3552	70.99	2485	0.7546	0.3551
0.3292	71.99	2520	0.7483	0.3450
0.3041	72.99	2555	0.7305	0.3522
0.3606	73.99	2590	0.7358	0.3484
0.3629	74.99	2625	0.7709	0.3446
0.3409	75.99	2660	0.7568	0.3585
0.3315	76.99	2695	0.7466	0.3475
0.2934	77.99	2730	0.7351	0.3496
0.3366	78.99	2765	0.8014	0.3484
0.3176	79.99	2800	0.8014	0.3420
0.3319	80.99	2835	0.7996	0.3437
0.2967	81.99	2870	0.8156	0.3412
0.3137	82.99	2905	0.8025	0.3361
0.3133	83.99	2940	0.7784	0.3416
0.3134	84.99	2975	0.7894	0.3336
0.3216	85.99	3010	0.8331	0.3395
0.365	86.99	3045	0.7980	0.3353
0.2962	87.99	3080	0.7965	0.3404
0.3126	88.99	3115	0.7470	0.3420
0.2843	89.99	3150	0.7788	0.3404
0.2967	90.99	3185	0.7902	0.3374
0.3171	91.99	3220	0.8022	0.3404
0.3069	92.99	3255	0.7999	0.3345
0.3571	93.99	3290	0.7896	0.3404
0.2805	94.99	3325	0.7831	0.3391
0.3099	95.99	3360	0.7909	0.3366
0.2868	96.99	3395	0.7918	0.3395
0.2626	97.99	3430	0.7766	0.3429
0.2634	98.99	3465	0.7770	0.3437
0.288	99.99	3500	0.7762	0.3429

Framework versions

• Transformers 4.25.0.dev0
• Pytorch 1.8.1+cu111
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2