Spaces:
Running
Running
File size: 4,042 Bytes
1d8403e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 | # ==================================================================================================
# DEEPFAKE AUDIO - vocoder_train.py (Waveform Reconstruction Neural Training)
# ==================================================================================================
#
# 📝 DESCRIPTION
# This script manages the training of the WaveRNN-based Vocoder. It optimizes a
# recurrent neural network to reconstruct high-fidelity time-domain audio from
# mel-spectrogram sequences. For optimal results, it uses the GTA (Ground-Truth
# Aligned) spectrograms generated during the preprocessing phase, allowing the
# vocoder to compensate for common synthesizer artifacts.
#
# 👤 AUTHORS
# - Amey Thakur (https://github.com/Amey-Thakur)
# - Mega Satish (https://github.com/msatmod)
#
# 🤝🏻 CREDITS
# Original Real-Time Voice Cloning methodology by CorentinJ
# Repository: https://github.com/CorentinJ/Real-Time-Voice-Cloning
#
# 🔗 PROJECT LINKS
# Repository: https://github.com/Amey-Thakur/DEEPFAKE-AUDIO
# Video Demo: https://youtu.be/i3wnBcbHDbs
# Research: https://github.com/Amey-Thakur/DEEPFAKE-AUDIO/blob/main/DEEPFAKE-AUDIO.ipynb
#
# 📜 LICENSE
# Released under the MIT License
# Release Date: 2021-02-06
# ==================================================================================================
import argparse
from pathlib import Path
# --- CORE VOCATION ENGINE ---
from utils.argutils import print_args
from vocoder.train import train
if __name__ == "__main__":
# --- TRAINING PARAMETERS ---
parser = argparse.ArgumentParser(
description="Vocoder Training Hub: Reconstructing waveforms from synthetic spectrograms.",
formatter_class=argparse.ArgumentDefaultsHelpFormatter
)
# --- SESSION DEFINITION ---
parser.add_argument("run_id", type=str,
help="Identifier for this experimental run. Weights and metrics will be archived under this ID.")
parser.add_argument("datasets_root", type=Path,
help="Root path to the SV2TTS training directory.")
# --- PATH OVERRIDES ---
parser.add_argument("--syn_dir", type=Path, default=argparse.SUPPRESS,
help="Custom path to synthesizer-level data. Defaults to <datasets_root>/SV2TTS/synthesizer/.")
parser.add_argument("--voc_dir", type=Path, default=argparse.SUPPRESS,
help="Custom path to vocoder-level GTA data. Defaults to <datasets_root>/SV2TTS/vocoder/.")
parser.add_argument("-m", "--models_dir", type=Path, default="saved_models",
help="Destination for saved weights, rolling backups, and audio samples.")
# --- TRAINING LOGIC ---
parser.add_argument("-g", "--ground_truth", action="store_true",
help="Bypass GTA spectrograms and train directly on original ground truth data.")
parser.add_argument("-s", "--save_every", type=int, default=1000,
help="Iteration interval for weight persistence on disk.")
parser.add_argument("-b", "--backup_every", type=int, default=25000,
help="Frequency of immutable state backups.")
parser.add_argument("-f", "--force_restart", action="store_true",
help="Initialize weights from scratch, ignoring existing checkpoints.")
args = parser.parse_args()
# --- ARCHITECTURAL RESOLUTION ---
if not hasattr(args, "syn_dir"):
args.syn_dir = args.datasets_root / "SV2TTS" / "synthesizer"
if not hasattr(args, "voc_dir"):
args.voc_dir = args.datasets_root / "SV2TTS" / "vocoder"
# Finalize workspace preparations.
del args.datasets_root
args.models_dir.mkdir(exist_ok=True)
# --- EXECUTION ---
print_args(args, parser)
print("🤝🏻 Scholarly Partnership: Amey Thakur & Mega Satish")
print("🚀 Initiating Vocoder Training Pipeline - Optimizing wave reconstruction...")
# Delegate to the WaveRNN training module.
train(**vars(args))
|