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: 0.3866
• Wer: 0.3363

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
3.9949	3.67	400	0.7055	0.6984
0.4192	7.34	800	0.4530	0.4711
0.1987	11.01	1200	0.4319	0.4384
0.1317	14.68	1600	0.4332	0.4179
0.0988	18.35	2000	0.4201	0.3755
0.0791	22.02	2400	0.3968	0.3723
0.0628	25.69	2800	0.3998	0.3477
0.0501	29.36	3200	0.3866	0.3363

Framework versions

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