speechT5_tts-finetuned-cml-tts2

This model is a fine-tuned version of microsoft/speechT5_tts on the cml-tts dataset. It achieves the following results on the evaluation set:

• Loss: 0.4595

Model description

SpeechT5 model trained for Audio course Unit 6 hands-on on Portugues language cml-tts2 dataset for 5 hours. Honestly it is not that good but definetly better then initial SpeechT5. More information here https://outleys.site/en/development/AI/hugface-audio-course-handson-unit-6-exercise/

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.99) and epsilon=1e-07 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 16000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.4819	0.0625	1000	0.5007
0.4364	0.125	2000	0.4965
0.4224	0.1875	3000	0.4841
0.4006	1.0473	4000	0.4782
0.3993	1.1098	5000	0.4728
0.3993	1.1723	6000	0.4687
0.389	2.032	7000	0.4684
0.3827	2.0945	8000	0.4665
0.3895	2.157	9000	0.4702
0.3829	3.0168	10000	0.4648
0.3717	3.0793	11000	0.4631
0.384	3.1418	12000	0.4627
0.3802	4.0015	13000	0.4601
0.3667	4.064	14000	0.4610
0.3757	4.1265	15000	0.4606
0.375	4.189	16000	0.4595

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3