Update app.py

app.py

@@ -7,7 +7,7 @@ import time
 from diffusers import DiffusionPipeline, AutoencoderTiny
 from custom_pipeline import FluxWithCFGPipeline
 
-torch.set_float32_matmul_precision("medium")
+torch.backends.cuda.matmul.allow_tf32 = True
 
 # Constants
 MAX_SEED = np.iinfo(np.int32).max
@@ -22,14 +22,9 @@ pipe = FluxWithCFGPipeline.from_pretrained(
     "black-forest-labs/FLUX.1-schnell", torch_dtype=dtype
 )
 pipe.vae = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype)
-# pipe.load_lora_weights("ostris/OpenFLUX.1", weight_name="openflux1-v0.1.0-fast-lora.safetensors", adapter_name="fast")
-# pipe.set_adapters("fast")
-# pipe.fuse_lora(adapter_names=["fast"], lora_scale=1.0)
 pipe.to("cuda")
-# pipe.transformer.to(memory_format=torch.channels_last)
-# pipe.transformer = torch.compile(
-#     pipe.transformer, mode="max-autotune", fullgraph=True
-# )
+pipe.enable_xformers_memory_efficient_attention()
+
 torch.cuda.empty_cache()
 
 # Inference function