speecht5_pt_full

This model is a fine-tuned version of microsoft/speecht5_tts on the mozilla-foundation/common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5021

DISCLAIMER: This model is trained for the sole purpose of finishing the HuggingFace Audio course. It doesn't have any usability and outputs pure noise. If you have an idea of how to improve the model, feel free to create a post in the Community tab of this model. Thank you!

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: 4
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 300
• training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss
0.8882	0.18	100	0.7494
0.7726	0.36	200	0.6657
0.642	0.54	300	0.5767
0.6042	0.71	400	0.5545
0.5972	0.89	500	0.5342
0.5832	1.07	600	0.5337
0.5851	1.25	700	0.5291
0.5744	1.43	800	0.5245
0.5638	1.61	900	0.5186
0.5562	1.78	1000	0.5174
0.56	1.96	1100	0.5133
0.5446	2.14	1200	0.5113
0.5556	2.32	1300	0.5099
0.5457	2.5	1400	0.5071
0.5504	2.68	1500	0.5087
0.5497	2.85	1600	0.5039
0.545	3.03	1700	0.5034
0.5503	3.21	1800	0.5051
0.5621	3.39	1900	0.5040
0.5347	3.57	2000	0.5021

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1