Hindi_SpeechT5_finetuned

This model is a fine-tuned version of microsoft/speecht5_tts on the Validated split of Hindi data of common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4524

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 1500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.6856	0.3442	100	0.5976
0.5929	0.6885	200	0.5453
0.5554	1.0327	300	0.5130
0.5407	1.3769	400	0.5052
0.5318	1.7212	500	0.4847
0.5213	2.0654	600	0.4796
0.514	2.4096	700	0.4728
0.5065	2.7539	800	0.4703
0.5046	3.0981	900	0.4684
0.4976	3.4423	1000	0.4621
0.4929	3.7866	1100	0.4583
0.4791	4.1308	1200	0.4550
0.4823	4.4750	1300	0.4529
0.485	4.8193	1400	0.4506
0.4774	5.1635	1500	0.4524

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.0.2
• Tokenizers 0.19.1