SpeechT5-tuned-bn

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

• Loss: 0.5572

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 1700
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.7601	0.1779	100	0.6865
0.7091	0.3559	200	0.6498
0.6819	0.5338	300	0.6345
0.6561	0.7117	400	0.6350
0.6353	0.8897	500	0.6044
0.6393	1.0676	600	0.5887
0.6402	1.2456	700	0.5906
0.6194	1.4235	800	0.5867
0.6127	1.6014	900	0.5788
0.608	1.7794	1000	0.5765
0.6129	1.9573	1100	0.5738
0.6044	2.1352	1200	0.5680
0.5894	2.3132	1300	0.5655
0.5952	2.4911	1400	0.5648
0.5963	2.6690	1500	0.5572
0.5889	2.8470	1600	0.5614
0.5897	3.0249	1700	0.5572

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.20.1