wav2vec2-large-xls-r-300m-hsb

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the Upper Sorbian common_voice_11_0 dataset.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0006
• 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: 50
• num_epochs: 1

Training results

Step	Training Loss	Validation Loss	Wer
200	4.408800	2.981909	0.979601
400	1.089200	0.900384	0.821805
600	0.166900	0.946962	0.755920
800	0.091500	0.877633	0.682767
1000	0.064100	0.883517	0.657913
1200	0.053000	0.865288	0.630715
1400	0.037800	0.867455	0.615475
1600	0.028900	0.834865	0.590621
1800	0.023800	0.845873	0.589215
2000	0.019600	0.830817	0.561313
2200	0.016300	0.836810	0.560610

Framework versions

• Transformers 4.32.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3