speecht5_keshandataset_sinhala4

This model is a fine-tuned version of microsoft/speecht5_tts on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3713

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: 0.0001
• 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: 100
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.517	0.6849	100	0.4540
0.4639	1.3699	200	0.4232
0.4485	2.0548	300	0.4083
0.4364	2.7397	400	0.3961
0.431	3.4247	500	0.3981
0.4242	4.1096	600	0.3922
0.4212	4.7945	700	0.3884
0.4138	5.4795	800	0.3846
0.4059	6.1644	900	0.3856
0.4042	6.8493	1000	0.3797
0.4009	7.5342	1100	0.3793
0.4015	8.2192	1200	0.3765
0.3992	8.9041	1300	0.3781
0.3952	9.5890	1400	0.3733
0.3935	10.2740	1500	0.3731
0.3942	10.9589	1600	0.3708
0.3904	11.6438	1700	0.3727
0.3848	12.3288	1800	0.3719
0.3864	13.0137	1900	0.3719
0.3872	13.6986	2000	0.3713

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.6.0
• Tokenizers 0.21.0