ai-light-dance_drums_ft_pretrain_wav2vec2-base-new-v6

This model is a fine-tuned version of gary109/ai-light-dance_drums_pretrain_wav2vec2-base-new on the GARY109/AI_LIGHT_DANCE - ONSET-IDMT-MDB-ENST dataset. It achieves the following results on the evaluation set:

• Loss: 0.6823
• Wer: 0.3851

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: 0.0004
• 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
17.274	0.99	35	2.9045	1.0
1.8443	1.99	70	3.5065	1.0
1.709	2.99	105	2.0072	1.0
1.4981	3.99	140	1.9510	0.9688
1.2977	4.99	175	1.8863	0.5534
1.1257	5.99	210	1.9137	0.4833
1.1218	6.99	245	1.9707	0.4960
0.8878	7.99	280	1.4179	0.4774
0.8562	8.99	315	1.5276	0.4635
1.5769	9.99	350	1.1270	0.4509
0.796	10.99	385	1.2706	0.4496
0.8776	11.99	420	1.2372	0.4471
0.7417	12.99	455	1.2826	0.4382
0.8273	13.99	490	1.2425	0.4542
0.7164	14.99	525	1.1415	0.4192
0.7061	15.99	560	1.2315	0.4407
0.6553	16.99	595	0.9983	0.4112
0.7114	17.99	630	1.1510	0.4382
0.6467	18.99	665	1.0612	0.4049
0.6035	19.99	700	1.0360	0.4188
0.6058	20.99	735	1.0008	0.4137
0.682	21.99	770	1.1948	0.4209
0.566	22.99	805	1.0555	0.4133
0.5952	23.99	840	0.8615	0.4095
0.5889	24.99	875	1.0740	0.4302
0.5954	25.99	910	1.1465	0.4167
0.5615	26.99	945	0.8980	0.4074
0.5385	27.99	980	0.8443	0.4062
0.5097	28.99	1015	1.1464	0.4049
0.5224	29.99	1050	1.0213	0.4003
0.5226	30.99	1085	0.8601	0.4091
0.5303	31.99	1120	1.0191	0.3986
0.6457	32.99	1155	1.2443	0.4306
0.5305	33.99	1190	0.9872	0.4171
0.5179	34.99	1225	1.0433	0.3935
0.471	35.99	1260	1.0011	0.4074
0.473	36.99	1295	0.8887	0.3901
0.5465	37.99	1330	0.8612	0.3897
0.4584	38.99	1365	0.9581	0.4070
0.565	39.99	1400	1.0735	0.4083
0.4916	40.99	1435	0.8890	0.3906
0.4643	41.99	1470	0.7317	0.4040
0.4633	42.99	1505	0.9384	0.4142
0.4867	43.99	1540	0.8899	0.4074
0.4892	44.99	1575	0.8419	0.4053
0.4338	45.99	1610	0.8297	0.4024
0.4038	46.99	1645	0.9689	0.3825
0.4519	47.99	1680	0.8536	0.4053
0.4298	48.99	1715	0.9737	0.3796
0.4622	49.99	1750	0.9054	0.4074
0.4358	50.99	1785	0.7809	0.3813
0.4277	51.99	1820	0.8464	0.3922
0.4186	52.99	1855	0.8106	0.3956
0.413	53.99	1890	0.9219	0.3813
0.4262	54.99	1925	0.9600	0.3990
0.4542	55.99	1960	0.8444	0.4057
0.3966	56.99	1995	0.7814	0.3914
0.444	57.99	2030	0.8331	0.3771
0.4673	58.99	2065	0.7872	0.3960
0.483	59.99	2100	1.0760	0.4036
0.5059	60.99	2135	0.8133	0.3981
0.3927	61.99	2170	0.8601	0.4032
0.4297	62.99	2205	0.7363	0.3880
0.4034	63.99	2240	0.7639	0.4028
0.3731	64.99	2275	0.8137	0.3686
0.3793	65.99	2310	0.7646	0.3787
0.3593	66.99	2345	0.7878	0.3952
0.3616	67.99	2380	0.7936	0.4045
0.3991	68.99	2415	0.7425	0.3775
0.3709	69.99	2450	0.6933	0.3834
0.3886	70.99	2485	0.7044	0.3728
0.3624	71.99	2520	0.6916	0.3922
0.3477	72.99	2555	0.7245	0.3872
0.4116	73.99	2590	0.6823	0.3851
0.3956	74.99	2625	0.7743	0.3846
0.386	75.99	2660	0.7772	0.3943
0.3755	76.99	2695	0.7823	0.3741
0.3569	77.99	2730	0.7801	0.3880
0.3403	78.99	2765	0.7619	0.3783
0.3623	79.99	2800	0.7294	0.3834
0.4157	80.99	2835	0.7345	0.3855
0.3569	81.99	2870	0.7349	0.3804
0.3988	82.99	2905	0.7232	0.3834
0.3425	83.99	2940	0.7239	0.3792
0.353	84.99	2975	0.7367	0.3758
0.3756	85.99	3010	0.7283	0.3728
0.3702	86.99	3045	0.7044	0.3792
0.3339	87.99	3080	0.7279	0.3766
0.3161	88.99	3115	0.7680	0.3796
0.3573	89.99	3150	0.7498	0.3733
0.3557	90.99	3185	0.7433	0.3779
0.3563	91.99	3220	0.7249	0.3787
0.3304	92.99	3255	0.7543	0.3783
0.3596	93.99	3290	0.7329	0.3733
0.3548	94.99	3325	0.7531	0.3720
0.3269	95.99	3360	0.7377	0.3712
0.3289	96.99	3395	0.7378	0.3749
0.2978	97.99	3430	0.7200	0.3728
0.3075	98.99	3465	0.7210	0.3724
0.3402	99.99	3500	0.7173	0.3737

Framework versions

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