load in advance

app.py

@@ -18,7 +18,7 @@ pipe_turbo = StableDiffusionXLPipeline_Turbo.from_pretrained(model_ckpt_turbo, t
 torch.cuda.empty_cache()
 
 @spaces.GPU(duration=120)
-def infer_gpu_part(pipe, seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps):
+def infer_gpu_part(seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps):
     generator = torch.Generator(device='cuda')
     generator = generator.manual_seed(seed)
     if not disable_freeu:
@@ -32,13 +32,13 @@ def infer_gpu_part(pipe, seed, prompt, negative_prompt, ddim_steps, guidance_sca
     return result
 
 @spaces.GPU(duration=40)
-def infer_gpu_part_turbo(pipe, seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps):
+def infer_gpu_part_turbo(seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps):
     generator = torch.Generator(device='cuda')
     generator = generator.manual_seed(seed)
     if not disable_freeu:
-        register_free_upblock2d(pipe, b1=1.1, b2=1.2, s1=0.6, s2=0.4)
-        register_free_crossattn_upblock2d(pipe, b1=1.1, b2=1.2, s1=0.6, s2=0.4)
-    result = pipe(prompt, negative_prompt=negative_prompt, generator=generator,
+        register_free_upblock2d(pipe_turbo, b1=1.1, b2=1.2, s1=0.6, s2=0.4)
+        register_free_crossattn_upblock2d(pipe_turbo, b1=1.1, b2=1.2, s1=0.6, s2=0.4)
+    result = pipe_turbo(prompt, negative_prompt=negative_prompt, generator=generator,
                 num_inference_steps=ddim_steps, guidance_scale=guidance_scale,
                 resolutions_list=resolutions_list, fast_mode=fast_mode, cosine_scale=cosine_scale,
                 restart_steps=restart_steps,
@@ -67,7 +67,7 @@ def infer(prompt, output_size, ddim_steps, guidance_scale, cosine_scale, seed, o
         restart_steps = [int(ddim_steps * 0.3)]
 
         # print('GPU starts')
-        result = infer_gpu_part(pipe, seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps)
+        result = infer_gpu_part(seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps)
         # print('GPU ends')
 
     else:
@@ -87,7 +87,7 @@ def infer(prompt, output_size, ddim_steps, guidance_scale, cosine_scale, seed, o
         restart_steps = [int(ddim_steps * 0.5)] * 2
 
         # print('GPU starts')
-        result = infer_gpu_part_turbo(pipe_turbo, seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps)
+        result = infer_gpu_part_turbo(seed, prompt, negative_prompt, ddim_steps, guidance_scale, resolutions_list, fast_mode, cosine_scale, disable_freeu, restart_steps)
         # print('GPU ends')
 
     return result