andrewbawitlung/SpeechT5-Mizo-Lus-v24.11.19

SpeechT5-v21

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

• Loss: 0.4759

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 12000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.5196	6.47	500	0.4665
0.4779	12.94	1000	0.4381
0.4525	19.42	1500	0.4259
0.438	25.89	2000	0.4207
0.4272	32.36	2500	0.4287
0.4211	38.83	3000	0.4359
0.4118	45.31	3500	0.4403
0.4074	51.78	4000	0.4431
0.4036	58.25	4500	0.4504
0.3972	64.72	5000	0.4571
0.3949	71.2	5500	0.4549
0.3906	77.67	6000	0.4554
0.3903	84.14	6500	0.4825
0.3858	90.61	7000	0.4600
0.3848	97.09	7500	0.4741
0.3838	103.56	8000	0.4686
0.3845	110.03	8500	0.4707
0.3819	116.5	9000	0.4610
0.3793	122.98	9500	0.4685
0.3808	129.45	10000	0.4666
0.377	135.92	10500	0.4766
0.3768	142.39	11000	0.4714
0.3762	148.87	11500	0.4792
0.3749	155.34	12000	0.4759

Framework versions

• Transformers 4.37.2
• Pytorch 2.3.1+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1