wav2vec2-large-xlsr-53-Hindi-Version3

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 0.7311
• Wer: 0.2821

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: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.362	6.7568	500	3.3795	1.0
0.5309	13.5135	1000	0.5572	0.4268
0.2342	20.2703	1500	0.5206	0.3341
0.1408	27.0270	2000	0.5516	0.3292
0.1126	33.7838	2500	0.6105	0.3199
0.0892	40.5405	3000	0.6489	0.3123
0.0721	47.2973	3500	0.6533	0.3067
0.0719	54.0541	4000	0.6898	0.3050
0.0592	60.8108	4500	0.7007	0.2990
0.0737	67.5676	5000	0.7106	0.2921
0.0399	74.3243	5500	0.7271	0.2916
0.0409	81.0811	6000	0.7298	0.2871
0.0322	87.8378	6500	0.7311	0.2835
0.0285	94.5946	7000	0.7311	0.2821

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1