Skip to content

fix(fish_speech): use from_indices() instead of decode() for DAC decoder #2668

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ianliuy wants to merge 3 commits into
base: main
Choose a base branch
from
Open

fix(fish_speech): use from_indices() instead of decode() for DAC decoder #2668

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
67cfe97
fix(fish_speech): use from_indices() instead of decode() for DAC decoder
ianliuy Apr 10, 2026
591d980
Merge branch 'main' into fix/issue-2643
lishunyang12 Apr 11, 2026
92299ab
chore: retrigger CI (build timed out)
ianliuy Apr 14, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 3 additions & 5 deletions tests/model_executor/models/test_fish_speech_regressions.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,11 +13,9 @@


class _FakeCodec:
def decode(self, codes_bqf: torch.Tensor, feature_lengths: torch.Tensor):
del codes_bqf, feature_lengths
wav = torch.arange(100, dtype=torch.float32).view(1, 1, 100)
audio_lengths = torch.tensor([100], dtype=torch.long)
return wav, audio_lengths
def from_indices(self, codes_bqf: torch.Tensor):
del codes_bqf
return torch.arange(100, dtype=torch.float32).view(1, 1, 100)


class _FakeTokenizer:
Expand Down
6 changes: 5 additions & 1 deletion vllm_omni/model_executor/models/fish_speech/fish_speech_dac_decoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -296,7 +296,11 @@ def forward(
codes_bqf[i, :, :frame_count] = codes_qf

with torch.amp.autocast("cuda", enabled=False):
wav_batch, audio_lengths = self._codec.decode(codes_bqf, feature_lengths)
wav_batch = self._codec.from_indices(codes_bqf)
audio_lengths = torch.clamp(
feature_lengths * self._hop_length,
max=wav_batch.shape[-1],
)

audios: list[torch.Tensor] = [empty] * num_req
srs = [sr_tensor] * num_req
Expand Down
Loading