Datasets:

Rvcmodel
/

gbs

Tasks:

Audio-to-Audio

Languages:

Chinese

English

Japanese

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

Duplicate

Split (1)

train

The dataset viewer is not available for this split.

Cannot load the dataset split (in streaming mode) to extract the first rows.

Error code:   StreamingRowsError
Exception:    ValueError
Message:      Invalid string class label gbs@08444dd4c4fb5add19204cf3c33537e1ad653418
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2386, in __iter__
                  example = _apply_feature_types_on_example(
                      example, self.features, token_per_repo_id=self.token_per_repo_id
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2303, in _apply_feature_types_on_example
                  encoded_example = features.encode_example(example)
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 2178, in encode_example
                  return encode_nested_example(self, example)
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1460, in encode_nested_example
                  {k: encode_nested_example(schema[k], obj.get(k), level=level + 1) for k in schema}
                      ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1483, in encode_nested_example
                  return schema.encode_example(obj) if obj is not None else None
                         ~~~~~~~~~~~~~~~~~~~~~^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1158, in encode_example
                  example_data = self.str2int(example_data)
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1095, in str2int
                  output = [self._strval2int(value) for value in values]
                            ~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/features/features.py", line 1116, in _strval2int
                  raise ValueError(f"Invalid string class label {value}")
              ValueError: Invalid string class label gbs@08444dd4c4fb5add19204cf3c33537e1ad653418

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

RVC multilingual base-model pretraining

This bundle prepares one RVC v2, 40 kHz, F0-guided multi-speaker base model from:

AISHELL/AISHELL-3 on Hugging Face;
badayvedat/VCTK on Hugging Face;
the official JVS Google Drive archive;
hfm_whisper_110spk.zip in Rvcmodel/gbs.

The HFM package contains 13,303 mono 48 kHz/16-bit WAV files from 110 speakers, totaling 19.889 hours. Its SHA-256 is cfc382e9b9b176a3c9d854bbcd0edaf8647d3beaba13ca6ea5f84abdb899a3db.

The script pins upstream RVC commit 81eed5e8f68b6bed1789f682fe78cdd324495afc. That revision supports multi-speaker manifests but caps the helper at speaker ID 109, so the bootstrap raises only that manifest-validation limit. The model embedding size is generated from the actual assembled speaker IDs.

On a fresh Ubuntu NVIDIA server, download the small script bundle first (the 2 GiB HFM ZIP is downloaded later by the bootstrap):

mkdir -p rvc-pretrain-scripts && cd rvc-pretrain-scripts
BASE=https://huggingface.co/datasets/Rvcmodel/gbs/resolve/main/pretrain_tools
for NAME in bootstrap_rvc_pretrain.sh build_multilingual_dataset.py finalize_rvc_experiment.py; do
  curl -fL "$BASE/$NAME" -o "$NAME"
done

Then run:

chmod +x bootstrap_rvc_pretrain.sh
WORK_ROOT=/workspace/rvc-pretrain \
GPU_IDS=0 \
FEATURE_GPU=0 \
BATCH_SIZE=8 \
TOTAL_EPOCHS=200 \
bash bootstrap_rvc_pretrain.sh

For multiple training GPUs use IDs separated with a dash, such as GPU_IDS=0-1. Feature and RMVPE extraction use FEATURE_GPU.

The defaults save full, reusable G_*.pth and D_*.pth checkpoints after every epoch and retain every epoch:

-se 1 -l 0 -sw 0

Expect roughly 1.2 GiB per epoch for the G/D pair, so 200 retained epochs can require around 240 GiB just for checkpoints. -l 1 would overwrite the same large pair and therefore does not satisfy the requirement to retain each epoch.

The first run downloads and preprocesses all corpora. It is restartable: completed downloads and extracted archives are reused. Set RESUME=1 to resume an experiment that already contains G/D checkpoints. Choose a new EXPERIMENT value for an independent run.

JVS is downloaded directly from its official Google Drive file. Its official terms do not permit redistribution of the full audio corpus, so the bundle does not copy JVS into Rvcmodel/gbs.

Downloads last month: 173

Total file size:

2.16 GB