wav2vec2-classic-300m-norwegian-colab-hung

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the voxpopuli dataset. It achieves the following results on the evaluation set:

• Loss: 3.8820
• Wer: 1.7882

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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.7686	2.57	400	2.9953	1.0
2.5005	5.14	800	2.2739	1.9808
1.6554	7.72	1200	2.4720	1.6708
1.1995	10.29	1600	2.2613	1.2480
0.8972	12.86	2000	2.7599	1.8873
0.6962	15.43	2400	3.2783	1.9560
0.5554	18.01	2800	3.2272	1.7544
0.4234	20.58	3200	3.0755	1.5645
0.3341	23.15	3600	3.5022	1.7442
0.2832	25.72	4000	3.7905	1.8324
0.2293	28.3	4400	3.8820	1.7882

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0