wav2vec2-large-xls-r-300m-turkish-colab

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

• Loss: 0.3965
• Wer: 0.3807

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
3.974	3.67	400	0.7102	0.7318
0.4216	7.34	800	0.4273	0.4941
0.1891	11.01	1200	0.4548	0.4864
0.1267	14.68	1600	0.4208	0.4082
0.0958	18.35	2000	0.4236	0.4033
0.0799	22.02	2400	0.4052	0.3829
0.0624	25.69	2800	0.4088	0.3875
0.0491	29.36	3200	0.3965	0.3807

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu111
• Datasets 1.18.3
• Tokenizers 0.10.3