wav2vec2-base-vios-v1

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6352
• Wer: 0.5161

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: 0.0001
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.7944	3.98	1000	1.7427	1.0387
0.7833	7.97	2000	0.4026	0.4364
0.4352	11.95	3000	0.3967	0.4042
0.4988	15.94	4000	0.5446	0.4632
0.7822	19.92	5000	0.6563	0.5491
0.8496	23.9	6000	0.5828	0.5045
0.8072	27.89	7000	0.6318	0.5109
0.8336	31.87	8000	0.6352	0.5161
0.8311	35.86	9000	0.6352	0.5161
0.839	39.84	10000	0.6352	0.5161
0.8297	43.82	11000	0.6352	0.5161
0.8288	47.81	12000	0.6352	0.5161

Framework versions

• Transformers 4.17.0
• Pytorch 1.11.0+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1