Spaces:
Running
on
A10G
Running
on
A10G
import torch | |
from diffusers import KDPM2DiscreteScheduler | |
from diffusers.utils import torch_device | |
from .test_schedulers import SchedulerCommonTest | |
class KDPM2DiscreteSchedulerTest(SchedulerCommonTest): | |
scheduler_classes = (KDPM2DiscreteScheduler,) | |
num_inference_steps = 10 | |
def get_scheduler_config(self, **kwargs): | |
config = { | |
"num_train_timesteps": 1100, | |
"beta_start": 0.0001, | |
"beta_end": 0.02, | |
"beta_schedule": "linear", | |
} | |
config.update(**kwargs) | |
return config | |
def test_timesteps(self): | |
for timesteps in [10, 50, 100, 1000]: | |
self.check_over_configs(num_train_timesteps=timesteps) | |
def test_betas(self): | |
for beta_start, beta_end in zip([0.00001, 0.0001, 0.001], [0.0002, 0.002, 0.02]): | |
self.check_over_configs(beta_start=beta_start, beta_end=beta_end) | |
def test_schedules(self): | |
for schedule in ["linear", "scaled_linear"]: | |
self.check_over_configs(beta_schedule=schedule) | |
def test_prediction_type(self): | |
for prediction_type in ["epsilon", "v_prediction"]: | |
self.check_over_configs(prediction_type=prediction_type) | |
def test_full_loop_with_v_prediction(self): | |
scheduler_class = self.scheduler_classes[0] | |
scheduler_config = self.get_scheduler_config(prediction_type="v_prediction") | |
scheduler = scheduler_class(**scheduler_config) | |
scheduler.set_timesteps(self.num_inference_steps) | |
model = self.dummy_model() | |
sample = self.dummy_sample_deter * scheduler.init_noise_sigma | |
sample = sample.to(torch_device) | |
for i, t in enumerate(scheduler.timesteps): | |
sample = scheduler.scale_model_input(sample, t) | |
model_output = model(sample, t) | |
output = scheduler.step(model_output, t, sample) | |
sample = output.prev_sample | |
result_sum = torch.sum(torch.abs(sample)) | |
result_mean = torch.mean(torch.abs(sample)) | |
if torch_device in ["cpu", "mps"]: | |
assert abs(result_sum.item() - 4.6934e-07) < 1e-2 | |
assert abs(result_mean.item() - 6.1112e-10) < 1e-3 | |
else: | |
# CUDA | |
assert abs(result_sum.item() - 4.693428650170972e-07) < 1e-2 | |
assert abs(result_mean.item() - 0.0002) < 1e-3 | |
def test_full_loop_no_noise(self): | |
if torch_device == "mps": | |
return | |
scheduler_class = self.scheduler_classes[0] | |
scheduler_config = self.get_scheduler_config() | |
scheduler = scheduler_class(**scheduler_config) | |
scheduler.set_timesteps(self.num_inference_steps) | |
model = self.dummy_model() | |
sample = self.dummy_sample_deter * scheduler.init_noise_sigma | |
sample = sample.to(torch_device) | |
for i, t in enumerate(scheduler.timesteps): | |
sample = scheduler.scale_model_input(sample, t) | |
model_output = model(sample, t) | |
output = scheduler.step(model_output, t, sample) | |
sample = output.prev_sample | |
result_sum = torch.sum(torch.abs(sample)) | |
result_mean = torch.mean(torch.abs(sample)) | |
if torch_device in ["cpu", "mps"]: | |
assert abs(result_sum.item() - 20.4125) < 1e-2 | |
assert abs(result_mean.item() - 0.0266) < 1e-3 | |
else: | |
# CUDA | |
assert abs(result_sum.item() - 20.4125) < 1e-2 | |
assert abs(result_mean.item() - 0.0266) < 1e-3 | |
def test_full_loop_device(self): | |
if torch_device == "mps": | |
return | |
scheduler_class = self.scheduler_classes[0] | |
scheduler_config = self.get_scheduler_config() | |
scheduler = scheduler_class(**scheduler_config) | |
scheduler.set_timesteps(self.num_inference_steps, device=torch_device) | |
model = self.dummy_model() | |
sample = self.dummy_sample_deter.to(torch_device) * scheduler.init_noise_sigma | |
for t in scheduler.timesteps: | |
sample = scheduler.scale_model_input(sample, t) | |
model_output = model(sample, t) | |
output = scheduler.step(model_output, t, sample) | |
sample = output.prev_sample | |
result_sum = torch.sum(torch.abs(sample)) | |
result_mean = torch.mean(torch.abs(sample)) | |
if str(torch_device).startswith("cpu"): | |
# The following sum varies between 148 and 156 on mps. Why? | |
assert abs(result_sum.item() - 20.4125) < 1e-2 | |
assert abs(result_mean.item() - 0.0266) < 1e-3 | |
else: | |
# CUDA | |
assert abs(result_sum.item() - 20.4125) < 1e-2 | |
assert abs(result_mean.item() - 0.0266) < 1e-3 | |