From 90eac14f720cf66ca1e28f1cc4af32df44806bc7 Mon Sep 17 00:00:00 2001
From: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>
Date: Fri, 21 Apr 2023 18:24:37 +0000
Subject: [PATCH] [AudioLDM] Fix dtype of returned waveform (#3189)

---
 src/diffusers/pipelines/audioldm/pipeline_audioldm.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/diffusers/pipelines/audioldm/pipeline_audioldm.py b/src/diffusers/pipelines/audioldm/pipeline_audioldm.py
index 86a8fd6590..c80e6ef992 100644
--- a/src/diffusers/pipelines/audioldm/pipeline_audioldm.py
+++ b/src/diffusers/pipelines/audioldm/pipeline_audioldm.py
@@ -293,7 +293,7 @@ class AudioLDMPipeline(DiffusionPipeline):
 
         waveform = self.vocoder(mel_spectrogram)
         # we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
-        waveform = waveform.cpu()
+        waveform = waveform.cpu().float()
         return waveform
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs