wav2vec2-large-xls-r-300m-Hindi

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7786
• Wer: 0.5158

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.0146	0.7214	400	2.8409	0.9970
1.2129	1.4427	800	1.1656	0.7795
0.7592	2.1641	1200	1.0091	0.7199
0.5798	2.8855	1600	0.9187	0.6614
0.4609	3.6069	2000	0.8386	0.6084
0.3828	4.3282	2400	0.8442	0.6097
0.3242	5.0496	2800	0.7907	0.5744
0.2619	5.7710	3200	0.7661	0.5485
0.2132	6.4923	3600	0.7943	0.5388
0.1911	7.2137	4000	0.7835	0.5278
0.1637	7.9351	4400	0.7786	0.5158

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1