ai-light-dance_singing3_ft_wav2vec2-large-xlsr-53-v1-5gram

This model is a fine-tuned version of gary109/ai-light-dance_singing3_ft_wav2vec2-large-xlsr-53-v1-5gram on the GARY109/AI_LIGHT_DANCE - ONSET-SINGING3 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4505
• Wer: 0.2119

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: 1e-07
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 100.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3355	1.0	144	0.4505	0.2119
0.3069	2.0	288	0.4509	0.2124
0.3049	3.0	432	0.4511	0.2119
0.3028	4.0	576	0.4521	0.2114
0.3092	5.0	720	0.4532	0.2112
0.3043	6.0	864	0.4536	0.2117
0.2903	7.0	1008	0.4543	0.2114
0.3124	8.0	1152	0.4538	0.2118
0.3079	9.0	1296	0.4541	0.2121
0.3093	10.0	1440	0.4537	0.2117
0.3093	11.0	1584	0.4544	0.2111
0.3202	12.0	1728	0.4549	0.2110
0.3086	13.0	1872	0.4546	0.2104
0.2947	14.0	2016	0.4542	0.2119
0.3145	15.0	2160	0.4539	0.2115
0.3292	16.0	2304	0.4532	0.2115
0.3049	17.0	2448	0.4547	0.2117
0.3177	18.0	2592	0.4544	0.2111
0.3108	19.0	2736	0.4547	0.2114
0.2944	20.0	2880	0.4560	0.2105
0.3232	21.0	3024	0.4560	0.2113
0.3196	22.0	3168	0.4559	0.2107
0.3207	23.0	3312	0.4563	0.2106
0.3039	24.0	3456	0.4555	0.2110
0.3157	25.0	3600	0.4560	0.2117
0.3285	26.0	3744	0.4561	0.2102
0.3125	27.0	3888	0.4553	0.2107
0.3051	28.0	4032	0.4560	0.2103
0.3166	29.0	4176	0.4560	0.2103
0.321	30.0	4320	0.4551	0.2101
0.3146	31.0	4464	0.4552	0.2100
0.323	32.0	4608	0.4551	0.2105
0.3223	33.0	4752	0.4554	0.2101
0.3105	34.0	4896	0.4549	0.2102
0.3134	35.0	5040	0.4552	0.2101
0.3054	36.0	5184	0.4550	0.2103
0.3162	37.0	5328	0.4554	0.2106
0.3094	38.0	5472	0.4551	0.2099
0.3174	39.0	5616	0.4553	0.2105
0.3218	40.0	5760	0.4553	0.2106
0.3134	41.0	5904	0.4552	0.2101
0.3019	42.0	6048	0.4552	0.2101
0.3169	43.0	6192	0.4552	0.2095
0.3209	44.0	6336	0.4550	0.2090
0.3035	45.0	6480	0.4550	0.2100
0.3181	46.0	6624	0.4550	0.2104
0.3133	47.0	6768	0.4546	0.2096
0.3173	48.0	6912	0.4556	0.2099
0.3174	49.0	7056	0.4552	0.2101
0.313	50.0	7200	0.4553	0.2100
0.3139	51.0	7344	0.4555	0.2101
0.3054	52.0	7488	0.4555	0.2100
0.3212	53.0	7632	0.4554	0.2097
0.3252	54.0	7776	0.4553	0.2097
0.3063	55.0	7920	0.4554	0.2106
0.3206	56.0	8064	0.4551	0.2097
0.3176	57.0	8208	0.4552	0.2101
0.3179	58.0	8352	0.4554	0.2099
0.3064	59.0	8496	0.4559	0.2092
0.301	60.0	8640	0.4559	0.2103
0.3103	61.0	8784	0.4559	0.2102
0.3169	62.0	8928	0.4559	0.2103
0.3081	63.0	9072	0.4559	0.2101
0.3249	64.0	9216	0.4555	0.2106
0.3031	65.0	9360	0.4553	0.2105
0.3017	66.0	9504	0.4556	0.2105
0.3261	67.0	9648	0.4551	0.2100
0.3196	68.0	9792	0.4553	0.2096
0.3085	69.0	9936	0.4554	0.2095
0.3235	70.0	10080	0.4552	0.2096
0.3194	71.0	10224	0.4550	0.2102
0.3243	72.0	10368	0.4546	0.2098
0.3115	73.0	10512	0.4542	0.2101
0.3307	74.0	10656	0.4545	0.2100
0.3072	75.0	10800	0.4547	0.2100
0.3218	76.0	10944	0.4545	0.2102
0.3116	77.0	11088	0.4540	0.2103
0.3021	78.0	11232	0.4542	0.2101
0.3165	79.0	11376	0.4539	0.2109
0.327	80.0	11520	0.4539	0.2090
0.3268	81.0	11664	0.4540	0.2110
0.304	82.0	11808	0.4537	0.2097
0.3256	83.0	11952	0.4537	0.2102
0.3208	84.0	12096	0.4544	0.2101
0.3199	85.0	12240	0.4541	0.2094
0.3104	86.0	12384	0.4543	0.2097
0.3218	87.0	12528	0.4542	0.2106
0.3301	88.0	12672	0.4538	0.2098
0.3055	89.0	12816	0.4540	0.2101
0.3154	90.0	12960	0.4533	0.2098
0.3169	91.0	13104	0.4543	0.2098
0.3122	92.0	13248	0.4541	0.2098
0.319	93.0	13392	0.4536	0.2094
0.307	94.0	13536	0.4538	0.2092
0.3132	95.0	13680	0.4540	0.2094
0.3185	96.0	13824	0.4536	0.2099
0.2996	97.0	13968	0.4541	0.2100
0.3193	98.0	14112	0.4539	0.2092
0.3091	99.0	14256	0.4538	0.2096
0.315	100.0	14400	0.4544	0.2100

Framework versions

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