ck8 / optimization_report.md
memevis's picture
Upload folder using huggingface_hub
e30f2e4 verified

TTS Model Optimization Report

Overview

  • Optimization Date: 2025-05-20 00:54:58
  • Iterations Performed: 3
  • Total Strategies Tested: 34
  • Best Overall Score: 0.8655

Performance Metrics

  • Mel Cepstral Distortion: 0.9073
  • Word Error Rate: 0.0575
  • Naturalness: 0.9089
  • Intelligibility: 0.9582
  • Speaker Similarity: 0.9349
  • Prosody: 0.9579
  • Overall Quality: 0.8781

Optimization Insights

Most Effective Parameter Settings

  • attention_scale: 1.5467
  • output_scale: 1.3266
  • projection_scale: 1.4734
  • encoder_scale: 1.2266
  • decoder_scale: 1.4467
  • base_enhancement: 0.0021
  • importance_factor: 1.7601

Optimization Journey

Iteration 1: Best strategy 'attention_decoder_focus_var1.5' - Score 0.8648

Iteration 2: Best strategy 'attention_decoder_focus_var1.5_iter2_noise_scale_0.8' - Score 0.8628

Iteration 3: Best strategy 'ensemble_top3_iter3' - Score 0.8655