bayerasif/speecht5_finetuned_voxpopuli_hu

microsoft/speecht5_tts_hu

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

• Loss: 0.4309

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: 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: 500
• training_steps: 4000

Training results

Training Loss	Epoch	Step	Validation Loss
0.6932	0.54	100	0.6017
0.6325	1.07	200	0.5632
0.5817	1.61	300	0.5078
0.5326	2.15	400	0.4830
0.5247	2.69	500	0.4703
0.5094	3.22	600	0.4630
0.5023	3.76	700	0.4568
0.4997	4.3	800	0.4541
0.4974	4.84	900	0.4504
0.4915	5.37	1000	0.4495
0.4885	5.91	1100	0.4475
0.4779	6.45	1200	0.4437
0.484	6.98	1300	0.4439
0.4799	7.52	1400	0.4419
0.4783	8.06	1500	0.4410
0.4764	8.6	1600	0.4401
0.4757	9.13	1700	0.4396
0.4742	9.67	1800	0.4378
0.4713	10.21	1900	0.4363
0.4747	10.75	2000	0.4370
0.4719	11.28	2100	0.4356
0.4694	11.82	2200	0.4349
0.4706	12.36	2300	0.4345
0.4757	12.89	2400	0.4341
0.466	13.43	2500	0.4334
0.4648	13.97	2600	0.4332
0.4663	14.51	2700	0.4329
0.4644	15.04	2800	0.4323
0.4646	15.58	2900	0.4324
0.4641	16.12	3000	0.4319
0.4644	16.66	3100	0.4316
0.463	17.19	3200	0.4312
0.4651	17.73	3300	0.4317
0.4637	18.27	3400	0.4315
0.4585	18.8	3500	0.4308
0.4605	19.34	3600	0.4310
0.4586	19.88	3700	0.4301
0.4636	20.42	3800	0.4308
0.4616	20.95	3900	0.4308
0.4593	21.49	4000	0.4309

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1