ThuraAung1601/speecht5_for_thai_with_ipa_tts_v2

SpeechT5-TTS with IPA v2 for Thai

This model is a fine-tuned version of microsoft/speecht5_tts on the Processed Thai Speech Data dataset. It achieves the following results on the evaluation set:

• Loss: 0.4837

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: 16
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• 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: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.6046	0.2664	1000	0.5522
0.564	0.5328	2000	0.5281
0.5567	0.7993	3000	0.5057
0.5458	1.0655	4000	0.5033
0.527	1.3320	5000	0.4917
0.5286	1.5984	6000	0.4874
0.5282	1.8648	7000	0.4839
0.5115	2.1311	8000	0.4847
0.5247	2.3975	9000	0.4837
0.528	2.6639	10000	0.4837

Framework versions

• Transformers 4.51.1
• Pytorch 2.5.1+cu124
• Datasets 3.5.0
• Tokenizers 0.21.0