EGY1.5K

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: 0.4590

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: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 3000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.5688	1.48	100	0.5112
0.5301	2.96	200	0.4908
0.5023	4.44	300	0.4707
0.524	5.93	400	0.4797
0.5001	7.41	500	0.4712
0.4774	8.89	600	0.4690
0.4793	10.37	700	0.4627
0.4666	11.85	800	0.4657
0.4649	13.33	900	0.4599
0.4659	14.81	1000	0.4616
0.4557	16.3	1100	0.4532
0.4516	17.78	1200	0.4535
0.4489	19.26	1300	0.4572
0.4431	20.74	1400	0.4504
0.4488	22.22	1500	0.4543
0.4452	23.7	1600	0.4557
0.4386	25.19	1700	0.4549
0.4297	26.67	1800	0.4487
0.4327	28.15	1900	0.4559
0.425	29.63	2000	0.4572
0.4251	31.11	2100	0.4531
0.4295	32.59	2200	0.4500
0.4258	34.07	2300	0.4561
0.4222	35.56	2400	0.4550
0.4119	37.04	2500	0.4569
0.4208	38.52	2600	0.4573
0.4145	40.0	2700	0.4568
0.4215	41.48	2800	0.4585
0.4141	42.96	2900	0.4594
0.4136	44.44	3000	0.4590

Framework versions

• Transformers 4.39.3
• Pytorch 2.3.0+cu118
• Datasets 3.0.0
• Tokenizers 0.15.2