Skip to content

SOTA Open Source TTS

Fix build dataset bug & update document

Lengyue4d6f6d0d2460
diff --git a/docs/en/finetune.md b/docs/en/finetune.md
index 8fb47e9..7d51571 100644
--- a/docs/en/finetune.md
+++ b/docs/en/finetune.md
@@ -37,7 +37,7 @@ python tools/vqgan/create_train_split.py data/demo
This command will create `data/demo/vq_train_filelist.txt` and `data/demo/vq_val_filelist.txt` in the `data/demo` directory, to be used for training and validation respectively.
!!!info
- For the VITS format, you can specify a file list using `--filelist xxx.list`.
+ For the VITS format, you can specify a file list using `--filelist xxx.list`.
Please note that the audio files in `filelist` must also be located in the `data/demo` folder.
### 3. Start Training
@@ -100,7 +100,8 @@ python tools/vqgan/extract_vq.py data/demo \
```
!!! note
- You can adjust `--num-workers` and `--batch-size` to increase extraction speed, but please make sure not to exceed your GPU memory limit.
+ You can adjust `--num-workers` and `--batch-size` to increase extraction speed, but please make sure not to exceed your GPU memory limit.
+ For the VITS format, you can specify a file list using `--filelist xxx.list`.
This command will create `.npy` files in the `data/demo` directory, as shown below:
@@ -134,7 +135,6 @@ After the command finishes executing, you should see the `quantized-dataset-ft.p
!!!info
For the VITS format, you can specify a file list using `--filelist xxx.list`.
- Please note that the audio files referenced in `filelist` must also be located in the `data/demo` folder.
### 4. Start the Rust data server
diff --git a/docs/zh/finetune.md b/docs/zh/finetune.md
index cdee5fb..714bf0c 100644
--- a/docs/zh/finetune.md
+++ b/docs/zh/finetune.md
@@ -37,7 +37,7 @@ python tools/vqgan/create_train_split.py data/demo
该命令会在 `data/demo` 目录下创建 `data/demo/vq_train_filelist.txt` 和 `data/demo/vq_val_filelist.txt` 文件, 分别用于训练和验证.
!!!info
- 对于 VITS 格式, 你可以使用 `--filelist xxx.list` 来指定文件列表.
+ 对于 VITS 格式, 你可以使用 `--filelist xxx.list` 来指定文件列表.
请注意, `filelist` 所指向的音频文件必须也位于 `data/demo` 文件夹下.
### 3. 启动训练
@@ -112,7 +112,7 @@ python tools/vqgan/extract_vq.py data/demo \
```
!!! note
- 你可以调整 `--num-workers` 和 `--batch-size` 来提高提取速度, 但是请注意不要超过你的显存限制.
+ 你可以调整 `--num-workers` 和 `--batch-size` 来提高提取速度, 但是请注意不要超过你的显存限制.
对于 VITS 格式, 你可以使用 `--filelist xxx.list` 来指定文件列表.
该命令会在 `data/demo` 目录下创建 `.npy` 文件, 如下所示:
diff --git a/tools/llama/build_dataset.py b/tools/llama/build_dataset.py
index 7da3a13..46ccdd6 100644
--- a/tools/llama/build_dataset.py
+++ b/tools/llama/build_dataset.py
@@ -44,31 +44,30 @@ def task_generator_yaml(config):
logger.info(f"Found {len(grouped_files)} groups in {root}")
for name, subset in grouped_files.items():
- yield name, subset, source, languages, extension, None
+ yield name, subset, source, languages, extension
def task_generator_filelist(filelist):
grouped_files = defaultdict(list)
for filename, speaker, languages, text in load_filelist(filelist):
- if speaker in grouped_files:
- assert (
- languages == grouped_files[speaker][0][2]
- ), f"Speaker {speaker} has different languages"
-
grouped_files[speaker].append((Path(filename), text, languages))
logger.info(f"Found {len(grouped_files)} groups in {filelist}")
for speaker, values in grouped_files.items():
- for filename, txt, languages in values:
- yield speaker, filename, "filelist", languages, None, txt
+ yield speaker, values, "filelist", languages, None
def run_task(task):
- name, subset, source, languages, extension, text = task
+ name, subset, source, languages, extension = task
# Parse the files
sentences = []
for file in subset:
+ if isinstance(file, tuple):
+ file, text, languages = file
+ else:
+ text = None
+
np_file = file.with_suffix(".npy")
if np_file.exists() is False:
logger.warning(f"Can't find {np_file}")