Update sonic.py

sonic.py

@@ -89,13 +89,11 @@ def test(
             pad = torch.zeros_like(cond_clip[:, :, : 10 - cond_clip.shape[2]])
             cond_clip = torch.cat([cond_clip, pad], dim=2)
 
-        bucket_clip = last_prompts[:, start : start + 50]
-        if bucket_clip.shape[2] < 50:                             # [★ 수정] 패딩
-            pad = torch.zeros_like(bucket_clip[:, :, : 50 - bucket_clip.shape[2]])
-            bucket_clip = torch.cat([bucket_clip, pad], dim=2)
-
+        bucket_clip = last_prompts[:, start : start + 50]           # (1 , 50 , 384)
+        bucket_clip = bucket_clip.unsqueeze(0).unsqueeze(-2)        # (1 , 1 , 50 , 1 , 384) ✔
         motion = audio2bucket(bucket_clip, image_embeds) * 16 + 16
 
+
         ref_list.append(ref_img[0])
         audio_list.append(audio_pe(cond_clip).squeeze(0)[0])
         uncond_list.append(audio_pe(torch.zeros_like(cond_clip)).squeeze(0)[0])