zlm_b32_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: 0.3178

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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: 2000
• training_steps: 12000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.5499	0.2094	500	0.4841
0.4812	0.4188	1000	0.4274
0.451	0.6283	1500	0.4068
0.424	0.8377	2000	0.3928
0.4261	1.0471	2500	0.3783
0.4013	1.2565	3000	0.3741
0.3949	1.4660	3500	0.3681
0.3874	1.6754	4000	0.3603
0.3871	1.8848	4500	0.3590
0.3963	2.0942	5000	0.3481
0.3821	2.3037	5500	0.3421
0.3765	2.5131	6000	0.3407
0.3718	2.7225	6500	0.3373
0.3718	2.9319	7000	0.3368
0.3579	3.1414	7500	0.3339
0.3559	3.3508	8000	0.3300
0.3602	3.5602	8500	0.3268
0.3637	3.7696	9000	0.3272
0.355	3.9791	9500	0.3236
0.3616	4.1885	10000	0.3223
0.3509	4.3979	10500	0.3195
0.3481	4.6073	11000	0.3197
0.3544	4.8168	11500	0.3181
0.35	5.0262	12000	0.3178

Framework versions

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