xlsr_hindi

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_16_1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9242
• Wer : 0.4032

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: 10
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 3
• total_train_batch_size: 30
• 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
12.3878	3.61	200	3.7452	1.0
3.3396	7.23	400	2.7621	1.0
1.1465	10.84	600	0.9738	0.5791
0.4158	14.46	800	0.8970	0.4873
0.2417	18.07	1000	0.8884	0.4374
0.1703	21.69	1200	0.8942	0.4164
0.1293	25.3	1400	0.9219	0.4093
0.1027	28.92	1600	0.9242	0.4032

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.15.2