zlm_b64_le4_s12000

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: 0.3114

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: 32
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2000
• training_steps: 12000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.5487	0.4188	500	0.4746
0.483	0.8375	1000	0.4227
0.432	1.2563	1500	0.3983
0.429	1.6750	2000	0.3953
0.4168	2.0938	2500	0.3701
0.4021	2.5126	3000	0.3613
0.3925	2.9313	3500	0.3509
0.3839	3.3501	4000	0.3506
0.3798	3.7688	4500	0.3423
0.3693	4.1876	5000	0.3375
0.3712	4.6064	5500	0.3367
0.3668	5.0251	6000	0.3316
0.3635	5.4439	6500	0.3291
0.3543	5.8626	7000	0.3250
0.3526	6.2814	7500	0.3221
0.3525	6.7002	8000	0.3218
0.3513	7.1189	8500	0.3182
0.346	7.5377	9000	0.3163
0.3448	7.9564	9500	0.3162
0.3563	8.3752	10000	0.3145
0.3449	8.7940	10500	0.3126
0.3436	9.2127	11000	0.3128
0.3413	9.6315	11500	0.3121
0.3397	10.0503	12000	0.3114

Framework versions

• Transformers 4.41.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1