SSL-Harveen-Chadda-Fine-Tuning

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: 0.0032
• Wer: 0.1008

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: 500
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.3793	4.17	400	0.5347	0.2584
0.2137	8.33	800	0.8339	0.2664
0.1282	12.5	1200	0.1785	0.1414
0.0698	16.66	1600	0.0135	0.1066
0.0354	20.83	2000	0.0351	0.1179
0.0212	24.99	2400	0.0104	0.1035
0.0066	29.17	2800	0.0032	0.1008

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu111
• Datasets 1.18.3
• Tokenizers 0.10.3