wav2vec2-large-xlsr-53-ur

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: inf
• Wer: 0.4817

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: 6
• eval_batch_size: 6
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 12
• total_eval_batch_size: 12
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.0981	0.48	300	inf	0.9981
2.0031	0.97	600	inf	0.8283
0.7476	1.45	900	inf	0.6584
0.8585	1.94	1200	inf	0.5823
0.4978	2.42	1500	inf	0.5564
0.5423	2.9	1800	inf	0.5209
0.3504	3.39	2100	inf	0.5396
0.3185	3.87	2400	inf	0.4865
0.3337	4.35	2700	inf	0.4733
0.4935	4.84	3000	inf	0.4721
0.4022	5.32	3300	inf	0.4692
0.3517	5.81	3600	inf	0.4585
0.1838	6.29	3900	inf	0.4567
0.2635	6.77	4200	inf	0.4459
0.1163	7.26	4500	inf	0.4495
0.1776	7.74	4800	inf	0.4657
0.262	8.23	5100	inf	0.4562
0.1853	8.71	5400	inf	0.4724
0.3173	9.19	5700	inf	0.4752
0.4985	9.68	6000	inf	0.4604
0.3707	10.16	6300	inf	0.4769
0.4214	10.65	6600	inf	0.5246
0.3443	11.13	6900	inf	0.5391
0.3302	11.61	7200	inf	0.5051
0.327	12.1	7500	inf	0.5389
0.2489	12.58	7800	inf	0.5355
0.2328	13.06	8100	inf	0.5111
0.2488	13.55	8400	inf	0.4794
0.3255	14.03	8700	inf	0.4959
0.3056	14.52	9000	inf	0.4895
0.1758	15.0	9300	inf	0.4817

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2