Setosm/speecht5_finetuned_afrispeech_east_african

SpeechT5_tts_ea_eng

This model is a fine-tuned version of microsoft/speecht5_tts on the Afrispeech-200 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5899

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 400
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.798	2.8937	200	0.7320
0.7224	5.7928	400	0.6457
0.6826	8.6919	600	0.6195
0.6541	11.5910	800	0.6079
0.6493	14.4901	1000	0.6030
0.6454	17.3892	1200	0.5985
0.6491	20.2883	1400	0.5952
0.6273	23.1874	1600	0.5925
0.6416	26.0865	1800	0.5918
0.6195	28.9802	2000	0.5899

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1