hindi_wav2vec2_final

This model is a fine-tuned version of Harveenchadha/vakyansh-wav2vec2-hindi-him-4200 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0000
• Wer: 0.0699

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	10.0	25	2.4458	0.9376
No log	20.0	50	0.5012	0.1820
No log	30.0	75	0.2451	0.1511
No log	40.0	100	0.0827	0.1211
No log	50.0	125	0.0438	0.0902
No log	60.0	150	0.0271	0.0827
No log	70.0	175	0.0282	0.0759
No log	80.0	200	0.0089	0.0677
No log	90.0	225	0.0217	0.0752
No log	100.0	250	0.0088	0.0609
No log	110.0	275	0.0076	0.0947
No log	120.0	300	0.0085	0.0624
No log	130.0	325	0.0060	0.0632
No log	140.0	350	0.0101	0.0654
No log	150.0	375	0.0031	0.0579
No log	160.0	400	0.0035	0.0624
No log	170.0	425	0.0048	0.0744
No log	180.0	450	0.0035	0.0744
No log	190.0	475	0.0196	0.0707
1.3647	200.0	500	0.0036	0.0624
1.3647	210.0	525	0.0019	0.0617
1.3647	220.0	550	0.0022	0.0714
1.3647	230.0	575	0.0020	0.0639
1.3647	240.0	600	0.0022	0.0594
1.3647	250.0	625	0.0014	0.0857
1.3647	260.0	650	0.0053	0.0744
1.3647	270.0	675	0.0011	0.0609
1.3647	280.0	700	0.0007	0.0602
1.3647	290.0	725	0.0034	0.0632
1.3647	300.0	750	0.0021	0.0609
1.3647	310.0	775	0.0009	0.0624
1.3647	320.0	800	0.0025	0.0609
1.3647	330.0	825	0.0004	0.0579
1.3647	340.0	850	0.0002	0.0594
1.3647	350.0	875	0.0002	0.0594
1.3647	360.0	900	0.0054	0.0617
1.3647	370.0	925	0.0007	0.0602
1.3647	380.0	950	0.0005	0.0609
1.3647	390.0	975	0.0002	0.0594
0.0095	400.0	1000	0.0005	0.0609
0.0095	410.0	1025	0.0002	0.0609
0.0095	420.0	1050	0.0002	0.0632
0.0095	430.0	1075	0.0003	0.0647
0.0095	440.0	1100	0.0010	0.0617
0.0095	450.0	1125	0.0055	0.0654
0.0095	460.0	1150	0.0002	0.0602
0.0095	470.0	1175	0.0001	0.0602
0.0095	480.0	1200	0.0002	0.0617
0.0095	490.0	1225	0.0001	0.0609
0.0095	500.0	1250	0.0001	0.0624
0.0095	510.0	1275	0.0001	0.0632
0.0095	520.0	1300	0.0001	0.0632
0.0095	530.0	1325	0.0001	0.0669
0.0095	540.0	1350	0.0001	0.0654
0.0095	550.0	1375	0.0001	0.0677
0.0095	560.0	1400	0.0001	0.0632
0.0095	570.0	1425	0.0001	0.0609
0.0095	580.0	1450	0.0001	0.0609
0.0095	590.0	1475	0.0001	0.0632
0.0031	600.0	1500	0.0001	0.0714
0.0031	610.0	1525	0.0001	0.0692
0.0031	620.0	1550	0.0001	0.0707
0.0031	630.0	1575	0.0001	0.0677
0.0031	640.0	1600	0.0001	0.0902
0.0031	650.0	1625	0.0000	0.0714
0.0031	660.0	1650	0.0000	0.0767
0.0031	670.0	1675	0.0000	0.0737
0.0031	680.0	1700	0.0000	0.0669
0.0031	690.0	1725	0.0000	0.0677
0.0031	700.0	1750	0.0000	0.0669
0.0031	710.0	1775	0.0000	0.0684
0.0031	720.0	1800	0.0000	0.0669
0.0031	730.0	1825	0.0000	0.0677
0.0031	740.0	1850	0.0000	0.0647
0.0031	750.0	1875	0.0000	0.0647
0.0031	760.0	1900	0.0000	0.0647
0.0031	770.0	1925	0.0000	0.0699
0.0031	780.0	1950	0.0000	0.0699
0.0031	790.0	1975	0.0000	0.0692
0.0014	800.0	2000	0.0000	0.0699

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1+cu118
• Datasets 1.18.3
• Tokenizers 0.14.1