wav2vec2-base-vios-google-colab

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: 0.5647
• Wer: 0.4970

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: 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: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.7292	2.0	500	3.4159	1.0
3.0762	4.0	1000	1.3005	0.9615
0.8812	6.0	1500	0.4664	0.4740
0.5076	8.0	2000	0.4101	0.4180
0.4075	10.0	2500	0.3815	0.3802
0.3724	12.0	3000	0.3785	0.3741
0.3762	14.0	3500	0.4404	0.3766
0.4541	16.0	4000	0.4671	0.3822
0.6391	18.0	4500	0.5643	0.4200
0.7681	20.0	5000	0.6564	0.5214
0.8131	22.0	5500	0.5786	0.4934
0.7448	24.0	6000	0.5561	0.4920
0.7337	26.0	6500	0.5631	0.4964
0.7359	28.0	7000	0.5647	0.4968
0.7397	30.0	7500	0.5647	0.4970

Framework versions

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