urdu_text_to_speech_tts

This model is a fine-tuned version of microsoft/speecht5_tts on an common_voice_17_0 urdu dataset with very small amount. It's trained using only 4200 sentences, for business use model need to be trained on large datasets. It achieves the following results on the evaluation set:

• Loss: 0.4936

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: 8
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.6365	1.0	486	0.5707
0.6045	2.0	972	0.5319
0.591	3.0	1458	0.5265
0.5711	4.0	1944	0.5178
0.5528	5.0	2430	0.5142
0.5335	6.0	2916	0.5073
0.5316	7.0	3402	0.5015
0.5308	8.0	3888	0.4992
0.5381	9.0	4374	0.5022
0.5292	10.0	4860	0.4977
0.5242	11.0	5346	0.4975
0.5129	12.0	5832	0.4970
0.5122	13.0	6318	0.4937
0.5329	14.0	6804	0.4943
0.5189	15.0	7290	0.4921
0.5164	16.0	7776	0.4946
0.5097	17.0	8262	0.4931
0.5858	18.0	8748	0.4948
0.5128	19.0	9234	0.4936
0.5203	20.0	9720	0.4936

Framework versions

• Transformers 4.42.3
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1