Update app.py

app.py

@@ -1,79 +1,47 @@
 import gradio as gr
-from gradio_client import Client
-from diffusers import AutoencoderKL, StableDiffusionXLPipeline
+from diffusers import StableDiffusionXLPipeline, UNet2DConditionModel, EulerDiscreteScheduler, LCMScheduler
 import torch
-import concurrent.futures
-import spaces
-
-client_lightning = Client("AP123/SDXL-Lightning")
-client_hyper = Client("ByteDance/Hyper-SDXL-1Step-T2I")
-
-vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
+from huggingface_hub import hf_hub_download
+from safetensors.torch import load_file
 
 ### SDXL Turbo #### 
-pipe_turbo = StableDiffusionXLPipeline.from_pretrained("stabilityai/sdxl-turbo",
-                                                       vae=vae,
-                                                       torch_dtype=torch.float16,
-                                                       variant="fp16"
-                                                      )
+pipe_turbo = StableDiffusionXLPipeline.from_pretrained("stabilityai/sdxl-turbo", torch_dtype=torch.float16, variant="fp16")
 pipe_turbo.to("cuda")
 
+### SDXL Lightning ### 
+base = "stabilityai/stable-diffusion-xl-base-1.0"
+repo = "ByteDance/SDXL-Lightning"
+ckpt = "sdxl_lightning_1step_unet_x0.safetensors" 
 
-def get_lighting_result(prompt):
-    result_lighting = client_lightning.predict(
-        prompt,  # Your prompt
-        "1-Step",  # Number of inference steps
-        api_name="/generate_image"
-    )
-    return result_lighting
+unet = UNet2DConditionModel.from_config(base, subfolder="unet").to("cuda", torch.float16)
+unet.load_state_dict(load_file(hf_hub_download(repo, ckpt), device="cuda"))
+pipe_lightning = StableDiffusionXLPipeline.from_pretrained(base, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")
+pipe_lightning.scheduler = EulerDiscreteScheduler.from_config(pipe_lightning.scheduler.config, timestep_spacing="trailing", prediction_type="sample")
+pipe_lightning.to("cuda")
 
-def get_hyper_result(prompt):
-    result_hyper = client_hyper.predict(
-        num_images=1,
-        height=1024,
-        width=1024,
-        prompt=prompt,
-        seed=3413,
-        api_name="/process_image"
-    )
-    return result_hyper
+### Hyper SDXL ### 
+repo_name = "ByteDance/Hyper-SD"
+ckpt_name = "Hyper-SDXL-1step-Unet.safetensors"
 
-@spaces.GPU
-def get_turbo_result(prompt):
-    image_turbo = pipe_turbo(prompt=prompt, num_inference_steps=1, guidance_scale=0).images[0]
-    return image_turbo
+unet = UNet2DConditionModel.from_config(base, subfolder="unet").to("cuda", torch.float16)
+unet.load_state_dict(load_file(hf_hub_download(repo_name, ckpt_name), device="cuda"))
+pipe_hyper = StableDiffusionXLPipeline.from_pretrained(base, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")
+pipe_hyper.scheduler = LCMScheduler.from_config(pipe_hyper.scheduler.config)
+pipe_hyper.to("cuda")
 
 def run_comparison(prompt):
-    with concurrent.futures.ThreadPoolExecutor() as executor:
-        # Submit tasks to the executor
-        future_lighting = executor.submit(get_lighting_result, prompt)
-        future_hyper = executor.submit(get_hyper_result, prompt)
-        future_turbo = executor.submit(get_turbo_result, prompt)
-
-        # Wait for all futures to complete
-        results = concurrent.futures.wait(
-            [future_lighting, future_hyper, future_turbo],
-            return_when=concurrent.futures.ALL_COMPLETED
-        )
-
-        # Extract results from futures
-        result_lighting = future_lighting.result()
-        result_hyper = future_hyper.result()
-        image_turbo = future_turbo.result()
-        print(result_lighting)
-        print(result_hyper)
-        return image_turbo, result_lighting, result_hyper
+    image_turbo=pipe_turbo(prompt=prompt, num_inference_steps=1, guidance_scale=0).images[0]
+    image_lightning=pipe_lightning(prompt=prompt, num_inference_steps=1, guidance_scale=0).images[0]
+    image_hyper=pipe_hyper(prompt=prompt, num_inference_steps=1, guidance_scale=0, timesteps=[800]).images[0]
+    return image_turbo, image_lightning, image_hyper
 
-css = '''
-.gradio-container{max-width: 768px !important}
-'''
-with gr.Blocks(css=css) as demo:
+with gr.Blocks() as demo:
     prompt = gr.Textbox(label="Prompt")
     run = gr.Button("Run")
     with gr.Row():
         image_turbo = gr.Image(label="SDXL Turbo")
         image_lightning = gr.Image(label="SDXL Lightning")
-        image_hyper = gr.Image("Hyper SDXL")
+        image_hyper = gr.Image(label="Hyper SDXL")
     
     run.click(fn=run_comparison, inputs=prompt, outputs=[image_turbo, image_lightning, image_hyper])
 demo.launch()