Update src/pipeline.py

src/pipeline.py

@@ -208,9 +208,7 @@ def empty_cache():
     print(f"Flush took: {time.time() - start}")
 
 def load_pipeline() -> Pipeline:    
-    buffer = torch.empty((1024, 1024), device="cuda")
     empty_cache()
-    buffer = torch.empty((1024, 1024), device="cuda")
     dtype, device = torch.bfloat16, "cuda"
 
     text_encoder_2 = T5EncoderModel.from_pretrained(
@@ -228,11 +226,11 @@ def load_pipeline() -> Pipeline:
     torch.cuda.set_per_process_memory_fraction(0.99)
     pipeline.text_encoder.to(memory_format=torch.channels_last)
     pipeline.transformer.to(memory_format=torch.channels_last)
-    
-
     pipeline.vae.to(memory_format=torch.channels_last)
+    pipeline.vae.enable_tiling()
     pipeline.vae = torch.compile(pipeline.vae)
     
+    
     pipeline._exclude_from_cpu_offload = ["vae"]
     pipeline.enable_sequential_cpu_offload()
     for _ in range(2):