mirror of
https://github.com/Nighthawk42/soprano-factory.git
synced 2026-08-30 04:30:21 +00:00
Nighthawk
9b7dd275c7
- Move hyperparameters from hardcoded script values to `config.yaml` - Replace pip requirements with `pyproject.toml` and `uv` support (CUDA 12.6) - Refactor all scripts to use `pathlib` for robust path handling - Optimize `generate_dataset.py` with GPU acceleration - Register quantizer constants as buffers for proper device mapping - Update README with new installation and usage instructions
24 lines
757 B
Python
24 lines
757 B
Python
import json
|
|
import pathlib
|
|
from torch.utils.data import Dataset
|
|
|
|
|
|
class AudioDataset(Dataset):
|
|
def __init__(self, path):
|
|
# Convert string path to Path object if necessary for consistency
|
|
self.path = pathlib.Path(path)
|
|
with open(self.path, encoding='utf-8') as f:
|
|
self.dataset = json.load(f)
|
|
|
|
def __len__(self):
|
|
return len(self.dataset)
|
|
|
|
def __getitem__(self, idx):
|
|
text, audio = self.dataset[idx]
|
|
|
|
# Format: [STOP][TEXT]<text prompt>[START]<audio tokens>[STOP]
|
|
# Optimization: Use a generator expression for joining tokens
|
|
audio_tokens = ''.join(f'[{x}]' for x in audio)
|
|
res = f"[STOP][TEXT]{text}[START]{audio_tokens}[STOP]"
|
|
|
|
return res |