wav2vec2-classic-300m-norwegian-colab

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

• Loss: 2.2190
• Wer: 0.7528

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
4.8141	2.57	400	3.0571	1.0
3.0777	5.14	800	2.9987	1.0
2.7311	7.72	1200	2.5705	0.9829
2.1302	10.29	1600	1.8399	0.9225
1.6827	12.86	2000	1.6372	0.8559
1.312	15.43	2400	1.8908	0.9172
0.9979	18.01	2800	1.7908	0.7890
0.7456	20.58	3200	1.8110	0.7720
0.592	23.15	3600	2.0024	0.7686
0.4946	25.72	4000	2.1173	0.7702
0.4093	28.3	4400	2.2190	0.7528

Framework versions

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