wav2vec2-large-xls-r-300m-hindi-colab

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: 2.0956
• Wer: 0.7285

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: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.247	42.11	400	1.8828	0.9513
0.3283	84.21	800	1.7075	0.8399
0.0915	126.32	1200	1.7553	0.7715
0.0492	168.42	1600	1.8279	0.7645
0.0283	210.53	2000	1.9725	0.7970
0.0176	252.63	2400	1.9678	0.7483
0.013	294.74	2800	1.9957	0.7645
0.0082	336.84	3200	2.1776	0.7575
0.0051	378.95	3600	2.0943	0.7506
0.0039	421.05	4000	2.0855	0.7227
0.0026	463.16	4400	2.0956	0.7285

Framework versions

• Transformers 4.30.2
• Pytorch 1.11.0+cu113
• Datasets 1.18.3
• Tokenizers 0.13.3