Use fp32 for CPU diffusion pipeline.

app.py

@@ -27,11 +27,14 @@ gpu = False
 
 AUTH_TOKEN = os.environ.get('AUTH_TOKEN')
 
-pipeline = DiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", custom_pipeline="stable_diffusion_mega", torch_dtype=torch.float16, revision="fp16", use_auth_token=AUTH_TOKEN)
 
 if gpu:
+    pipeline = DiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", custom_pipeline="stable_diffusion_mega", torch_dtype=torch.float16, revision="fp16", use_auth_token=AUTH_TOKEN)
     pipeline.to("cuda")
-
+else:
+    pipeline = DiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5",
+                                                 custom_pipeline="stable_diffusion_mega", torch_dtype=torch.float32,
+                                                 revision="fp32", use_auth_token=AUTH_TOKEN)
 
 # Huggingface Spaces have 16GB RAM and 8 CPU cores
 # See https://huggingface.co/docs/hub/spaces-overview#hardware-resources