From ee5f2567ac684739bf254c5f09a1e9bcd5cd9555 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E9=91=AB?= <liuxin@modelbest.cn>
Date: Mon, 15 Dec 2025 20:37:02 +0800
Subject: [PATCH] FIX:When a prompt is present, concatenate two patches as the
 context for VAE decoding

---
 src/voxcpm/model/voxcpm.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/src/voxcpm/model/voxcpm.py b/src/voxcpm/model/voxcpm.py
index 5664353..3ce59eb 100644
--- a/src/voxcpm/model/voxcpm.py
+++ b/src/voxcpm/model/voxcpm.py
@@ -452,7 +452,6 @@ class VoxCPMModel(nn.Module):
                 patch_len = self.patch_size * self.chunk_size
                 for latent_pred, _ in inference_result:
                     decode_audio = self.audio_vae.decode(latent_pred.to(torch.float32))
-                    print(decode_audio.shape)
                     decode_audio = decode_audio[..., -patch_len:].squeeze(1).cpu()
                     yield decode_audio
                 break