speecht5_finetuned_om_vishesh

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

• Loss: 0.4650

Model description

Base Model: The project is centered around the SpeechT5 model from Hugging Face, designed for text-to-speech (TTS) tasks. Dataset: Utilized the Common Voice dataset, specifically focusing on the Hindi subset, to provide a diverse range of speech samples for training and fine-tuning

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: 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: 1500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.6519	1.0999	150	0.5697
0.5707	2.1998	300	0.5258
0.5473	3.2997	450	0.5129
0.5335	4.3996	600	0.4908
0.5131	5.4995	750	0.4855
0.5055	6.5995	900	0.4766
0.5005	7.6994	1050	0.4711
0.4926	8.7993	1200	0.4650
0.4873	9.8992	1350	0.4677
0.4882	10.9991	1500	0.4650

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.1+cu118
• Datasets 3.0.0
• Tokenizers 0.20.0