up

run_xl_ediffi.py

@@ -18,27 +18,53 @@ from torch.nn.functional import fractional_max_pool2d_with_indices
 api = HfApi()
 start_time = time.time()
 
-random_generator = torch.Generator()
-random_generator.manual_seed(12345)
+scheduler = DPMSolverMultistepScheduler(
+    beta_start=0.00085,
+    beta_end=0.012,
+    beta_schedule="scaled_linear",
+    prediction_type="epsilon",
+    num_train_timesteps=1000,
+    trained_betas=None,
+    thresholding=False,
+    algorithm_type="dpmsolver++",
+    solver_type="midpoint",
+    lower_order_final=True,
+    use_karras_sigmas=True,
+)
 
 model_id = "stabilityai/stable-diffusion-xl-base-0.9"
 pipe_high_noise = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16, variant="fp16", use_safetensors=True, local_files_only=True)
-pipe_high_noise.scheduler = DPMSolverMultistepScheduler.from_config(pipe_high_noise.scheduler.config)
+pipe_high_noise.scheduler = scheduler
 pipe_high_noise.to("cuda")
 
+scheduler = DPMSolverMultistepScheduler(
+    beta_start=0.00085,
+    beta_end=0.012,
+    beta_schedule="scaled_linear",
+    prediction_type="epsilon",
+    num_train_timesteps=1000,
+    trained_betas=None,
+    thresholding=False,
+    algorithm_type="dpmsolver++",
+    solver_type="midpoint",
+    lower_order_final=True,
+    use_karras_sigmas=True,
+)
+
 pipe_low_noise = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-refiner-0.9", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
-pipe_low_noise.scheduler = pipe_high_noise.scheduler
+pipe_low_noise.scheduler = scheduler
 pipe_low_noise.to("cuda")
 
 prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
 
-num_inference_steps = 40
-high_noise_frac = 0.8
 
-# seed = 0
-# seed_everything(seed)
+random_generator = torch.Generator()
+random_generator.manual_seed(0)
+
+num_inference_steps = 100
+high_noise_frac = 0.8
 
-image = pipe_high_noise(prompt=prompt, num_inference_steps=num_inference_steps, denoising_end=high_noise_frac, output_type="pt").images
+image = pipe_high_noise(prompt=prompt, num_inference_steps=num_inference_steps, denoising_end=high_noise_frac, output_type="latent").images
 image = pipe_low_noise(prompt=prompt, num_inference_steps=num_inference_steps, denoising_start=high_noise_frac, image=image).images[0]
 
 file_name = f"aaa_1"