wav2vec2-large-xls-r-300m-rw-KinyarwandaTTSDataset-10hr-v1

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

• Loss: 0.0032
• Wer: 0.0374
• Cer: 0.0061

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: 0.0005
• 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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.8359	1.3986	100	2.9224	1.0	1.0
2.8486	2.7972	200	2.6198	1.0	0.9721
1.159	4.1958	300	0.3714	0.6408	0.1162
0.3918	5.5944	400	0.1782	0.3063	0.0493
0.26	6.9930	500	0.1190	0.2531	0.0450
0.2008	8.3916	600	0.0753	0.1472	0.0230
0.1459	9.7902	700	0.0401	0.0921	0.0140
0.1144	11.1888	800	0.0271	0.0718	0.0112
0.0903	12.5874	900	0.0176	0.0474	0.0069
0.0751	13.9860	1000	0.0119	0.0573	0.0084
0.0619	15.3846	1100	0.0083	0.0340	0.0053
0.0509	16.7832	1200	0.0051	0.0437	0.0068
0.0435	18.1818	1300	0.0036	0.0374	0.0059
0.0363	19.5804	1400	0.0032	0.0374	0.0060

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1