Marathi_ASR_using_xlsr_wav2vec

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7582
• Wer: 0.7181

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 300
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	2.6667	200	0.7381	0.7263
0.336	5.3333	400	0.7472	0.7289
0.336	8.0	600	0.7452	0.7215
0.3237	10.6667	800	0.7449	0.7212
0.3237	13.3333	1000	0.7546	0.7192
0.3104	16.0	1200	0.7565	0.7210
0.3104	18.6667	1400	0.7550	0.7193
0.3089	21.3333	1600	0.7551	0.7186
0.3089	24.0	1800	0.7572	0.7185
0.2993	26.6667	2000	0.7571	0.7175
0.2993	29.3333	2200	0.7582	0.7181

Framework versions

• Transformers 4.40.0
• Pytorch 2.1.2
• Datasets 2.19.0
• Tokenizers 0.19.1