[Perf]: Speedup VoxCPM2 TTS performance and Support PagedAttention (#2690)

Signed-off-by: Sy03 <1370724210@qq.com>
Signed-off-by: Yueqian Lin <linyueqian@outlook.com>
Co-authored-by: Yueqian Lin <linyueqian@outlook.com>
Co-authored-by: Yueqian Lin <70319226+linyueqian@users.noreply.github.com>
This commit is contained in:
authored and GitHub committed 2026-04-14 09:16:47 +12:00
1 parent c9e2e3e8d7
commit 14f7910900
9 files changed
+1325 -440

No files matched your search

+3 -3
View File
@@ -58,12 +58,12 @@ The script accepts the following arguments:
## Performance
Measured on a single H20 GPU (80 GB), voxcpm 0.0.0, PyTorch 2.10.0+cu128:
Measured on a single H20 GPU (80 GB):
| Input length | RTF | Sample rate |
|---|---|---|
| Short (~6 words) | ~0.81 | 48 kHz |
| Long (~50 words) | ~0.72 | 48 kHz |
| Short (~10 tokens) | ~0.28 | 48 kHz |
| Long (~100 tokens) | ~0.34 | 48 kHz |
RTF < 1.0 means faster than real time.
@@ -71,10 +71,10 @@ def parse_args():
def extract_audio(multimodal_output: dict) -> torch.Tensor:
"""Extract the final complete audio tensor from multimodal output.
The output processor accumulates per-step full audio under ``audio``
as a list. The last element is the complete waveform.
The output processor concatenates per-step delta tensors under
``model_outputs``. Falls back to ``audio`` for backwards compat.
"""
audio = multimodal_output.get("audio") or multimodal_output.get("model_outputs")
audio = multimodal_output.get("model_outputs") or multimodal_output.get("audio")
if audio is None:
raise ValueError(f"No audio key in multimodal_output: {list(multimodal_output.keys())}")