Skip to content

SOTA Open Source TTS

Optimize webui & better concat repeat dataloader

Lengyue3ddf3d975c71
diff --git a/fish_speech/datasets/concat_repeat.py b/fish_speech/datasets/concat_repeat.py
index ce2671c..bda88e3 100644
--- a/fish_speech/datasets/concat_repeat.py
+++ b/fish_speech/datasets/concat_repeat.py
@@ -1,4 +1,5 @@
import bisect
+import random
from typing import Iterable
from torch.utils.data import Dataset, IterableDataset
@@ -32,7 +33,7 @@ class ConcatRepeatDataset(Dataset):
for d in self.datasets:
assert not isinstance(
d, IterableDataset
- ), "ConcatDataset does not support IterableDataset"
+ ), "ConcatRepeatDataset does not support IterableDataset"
self.cumulative_sizes = self.cumsum(self.datasets, self.repeats)
@@ -50,3 +51,38 @@ class ConcatRepeatDataset(Dataset):
dataset = self.datasets[dataset_idx]
return dataset[sample_idx % len(dataset)]
+
+
+class ConcatWeightedIterableDataset(IterableDataset):
+ datasets: list[IterableDataset]
+ weights: list[float]
+
+ def __init__(self, datasets: Iterable[IterableDataset], weights: list[float]):
+ super().__init__()
+
+ total_weight = sum(weights)
+ self.weights = [w / total_weight for w in weights]
+ self.datasets = list(datasets)
+
+ assert len(self.datasets) > 0, "datasets should not be an empty iterable"
+ assert len(self.datasets) == len(
+ weights
+ ), "datasets and repeats should have the same length"
+
+ for d in self.datasets:
+ assert isinstance(
+ d, IterableDataset
+ ), "ConcatRepeatIterableDataset only supports IterableDataset"
+
+ def __iter__(self):
+ all_datasets = [iter(dataset) for dataset in self.datasets]
+ ids = list(range(len(self.datasets)))
+
+ while True:
+ chosen_dataset = random.choices(ids, self.weights)[0]
+
+ try:
+ yield next(all_datasets[chosen_dataset])
+ except StopIteration:
+ all_datasets[chosen_dataset] = iter(self.datasets[chosen_dataset])
+ yield next(all_datasets[chosen_dataset])
diff --git a/tools/webui.py b/tools/webui.py
index 0e12e85..0571fd7 100644
--- a/tools/webui.py
+++ b/tools/webui.py
@@ -44,7 +44,6 @@ HEADER_MD = f"""# Fish Speech
TEXTBOX_PLACEHOLDER = i18n("Put your text here.")
SPACE_IMPORTED = False
-cached_audio = np.zeros((1,))
def build_html_error_message(error):