Update app.py

app.py

@@ -5,7 +5,7 @@ import spaces
 import os
 import torch
 
-from diffusers import DiffusionPipeline, AutoencoderTiny, AutoencoderKL
+from diffusers import DiffusionPipeline, AutoencoderTiny
 from huggingface_hub import login
 from live_preview_helpers import flux_pipe_call_that_returns_an_iterable_of_images
 
@@ -17,11 +17,9 @@ if hf_token:
 dtype = torch.bfloat16
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Load VAEs
+# ✅ DO NOT CHANGE: Working pipeline using taef1
 taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
-good_vae = AutoencoderKL.from_pretrained("black-forest-labs/FLUX.1-dev", subfolder="vae", torch_dtype=dtype).to(device)
 
-# Load base model with TAEF1 injected
 pipe = DiffusionPipeline.from_pretrained(
     "black-forest-labs/FLUX.1-dev",
     torch_dtype=dtype,
@@ -29,7 +27,6 @@ pipe = DiffusionPipeline.from_pretrained(
     vae=taef1
 ).to(device)
 
-# Inject method + apply LoRA
 pipe.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipe)
 pipe.load_lora_weights("ZennyKenny/flux_lora_natalie-diffusion")
 
@@ -42,7 +39,6 @@ def infer(prompt, seed=42, randomize_seed=False, width=1024, height=1024, guidan
         seed = random.randint(0, MAX_SEED)
     generator = torch.Generator().manual_seed(seed)
 
-    # Prepend XTON
     prompt = f"XTON {prompt}"
 
     for img in pipe.flux_pipe_call_that_returns_an_iterable_of_images(
@@ -115,7 +111,7 @@ Generate images in the surreal style of artist [Natalie Kav](https://www.behance
                     guidance_scale = gr.Slider(label="Guidance Scale", minimum=1, maximum=15, step=0.1, value=3.5)
                     num_inference_steps = gr.Slider(label="Number of inference steps", minimum=1, maximum=50, step=1, value=28)
 
-            # Define hidden result image just for example caching
+            # Use hidden result for examples only
             result_example = gr.Image(visible=False)
 
             gr.Examples(