zlm_b128_le4_s8000 / README.md
mikhail-panzo's picture
End of training
ba6aeb4 verified
metadata
license: mit
base_model: microsoft/speecht5_tts
tags:
  - generated_from_trainer
model-index:
  - name: zlm_b128_le4_s8000
    results: []

zlm_b128_le4_s8000

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: 0.3140

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: 8
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 2000
  • training_steps: 8000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
0.5174 0.8377 500 0.4794
0.4561 1.6754 1000 0.4022
0.4169 2.5131 1500 0.3854
0.4054 3.3508 2000 0.3649
0.3863 4.1885 2500 0.3528
0.3807 5.0262 3000 0.3480
0.3708 5.8639 3500 0.3432
0.366 6.7016 4000 0.3328
0.3575 7.5393 4500 0.3300
0.3607 8.3770 5000 0.3284
0.352 9.2147 5500 0.3232
0.3471 10.0524 6000 0.3189
0.3469 10.8901 6500 0.3171
0.3465 11.7277 7000 0.3161
0.3443 12.5654 7500 0.3159
0.3432 13.4031 8000 0.3140

Framework versions

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