Lagyamfi/speecht5_tts_lagyamfi_yor_v2

SpeechT5 TTS YOR_v2

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

• Loss: 0.4262

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: 0.0002
• train_batch_size: 6
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 20
• total_train_batch_size: 480
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.9966	58	0.4736
No log	1.9931	116	0.4692
No log	2.9897	174	0.4660
No log	3.9863	232	0.4698
No log	5.0	291	0.4874
No log	5.9966	349	0.4755
No log	6.9931	407	0.4630
No log	7.9897	465	0.4591
No log	8.9863	523	0.4646
No log	10.0	582	0.4536
No log	10.9966	640	0.4585
No log	11.9931	698	0.4745
No log	12.9897	756	0.4645
No log	13.9863	814	0.4618
No log	15.0	873	0.4444
No log	15.9966	931	0.4517
No log	16.9931	989	0.4395
0.4803	17.9897	1047	0.4523
0.4803	18.9863	1105	0.4335
0.4803	19.9313	1160	0.4262

Framework versions

• Transformers 4.40.1
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1