hindi_beekeeping_wav2vec2-2

This model is a fine-tuned version of Harveenchadha/hindi_base_wav2vec2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5244
• Wer: 0.3984

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: 10
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3815	4.55	25	1.0849	0.3984
0.1143	9.09	50	1.0752	0.3659
0.0827	13.64	75	1.4514	0.4390
0.0546	18.18	100	1.6510	0.4146
0.0506	22.73	125	1.5120	0.4593
0.0394	27.27	150	1.5948	0.4797
0.0356	31.82	175	1.4317	0.4228
0.0314	36.36	200	1.7475	0.4553
0.0278	40.91	225	1.6500	0.4634
0.0335	45.45	250	1.5310	0.4756
0.0213	50.0	275	1.4817	0.4593
0.0177	54.55	300	1.6921	0.4228
0.0115	59.09	325	1.6241	0.4106
0.0152	63.64	350	1.5167	0.4106
0.0109	68.18	375	1.7510	0.4553
0.011	72.73	400	1.6425	0.4146
0.0101	77.27	425	1.6253	0.4146
0.0091	81.82	450	1.6383	0.3984
0.008	86.36	475	1.5251	0.3984
0.0069	90.91	500	1.5244	0.3984

Framework versions

• Transformers 4.34.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1