wav2vec2-large-xls-r-300m-finetuned-hindi-common-voice-9-0

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 0.7392
• Wer: 1.0141

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: 4.42184e-05
• 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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
9.2217	3.03	400	4.0314	1.0
3.2902	6.06	800	2.1356	1.0001
0.9858	9.09	1200	0.8566	1.0037
0.5131	12.12	1600	0.7481	1.0074
0.3781	15.15	2000	0.7437	1.008
0.2998	18.18	2400	0.7310	1.0162
0.2553	21.21	2800	0.7384	1.0159
0.2216	24.24	3200	0.7537	1.0100
0.2048	27.27	3600	0.7392	1.0141

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu113
• Datasets 2.2.2
• Tokenizers 0.10.3