wav2vec2-large-xls-r-300m-slovenian

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - SL dataset. It achieves the following results on the evaluation set:

• Loss: 0.2093
• Wer: 0.1907

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: 7e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 100.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.785	12.5	1000	0.7465	0.6812
0.8989	25.0	2000	0.2495	0.2732
0.7118	37.5	3000	0.2126	0.2284
0.6367	50.0	4000	0.2049	0.2049
0.5763	62.5	5000	0.2116	0.2055
0.5196	75.0	6000	0.2111	0.1910
0.4949	87.5	7000	0.2131	0.1931
0.4797	100.0	8000	0.2093	0.1907

Framework versions

• Transformers 4.16.0.dev0
• Pytorch 1.10.1+cu102
• Datasets 1.18.3
• Tokenizers 0.11.0