|
import torch |
|
|
|
from diffusers import DDIMInverseScheduler |
|
|
|
from .test_schedulers import SchedulerCommonTest |
|
|
|
|
|
class DDIMInverseSchedulerTest(SchedulerCommonTest): |
|
scheduler_classes = (DDIMInverseScheduler,) |
|
forward_default_kwargs = (("num_inference_steps", 50),) |
|
|
|
def get_scheduler_config(self, **kwargs): |
|
config = { |
|
"num_train_timesteps": 1000, |
|
"beta_start": 0.0001, |
|
"beta_end": 0.02, |
|
"beta_schedule": "linear", |
|
"clip_sample": True, |
|
} |
|
|
|
config.update(**kwargs) |
|
return config |
|
|
|
def full_loop(self, **config): |
|
scheduler_class = self.scheduler_classes[0] |
|
scheduler_config = self.get_scheduler_config(**config) |
|
scheduler = scheduler_class(**scheduler_config) |
|
|
|
num_inference_steps = 10 |
|
|
|
model = self.dummy_model() |
|
sample = self.dummy_sample_deter |
|
|
|
scheduler.set_timesteps(num_inference_steps) |
|
|
|
for t in scheduler.timesteps: |
|
residual = model(sample, t) |
|
sample = scheduler.step(residual, t, sample).prev_sample |
|
|
|
return sample |
|
|
|
def test_timesteps(self): |
|
for timesteps in [100, 500, 1000]: |
|
self.check_over_configs(num_train_timesteps=timesteps) |
|
|
|
def test_steps_offset(self): |
|
for steps_offset in [0, 1]: |
|
self.check_over_configs(steps_offset=steps_offset) |
|
|
|
scheduler_class = self.scheduler_classes[0] |
|
scheduler_config = self.get_scheduler_config(steps_offset=1) |
|
scheduler = scheduler_class(**scheduler_config) |
|
scheduler.set_timesteps(5) |
|
assert torch.equal(scheduler.timesteps, torch.LongTensor([1, 201, 401, 601, 801])) |
|
|
|
def test_betas(self): |
|
for beta_start, beta_end in zip([0.0001, 0.001, 0.01, 0.1], [0.002, 0.02, 0.2, 2]): |
|
self.check_over_configs(beta_start=beta_start, beta_end=beta_end) |
|
|
|
def test_schedules(self): |
|
for schedule in ["linear", "squaredcos_cap_v2"]: |
|
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_clip_sample(self): |
|
for clip_sample in [True, False]: |
|
self.check_over_configs(clip_sample=clip_sample) |
|
|
|
def test_timestep_spacing(self): |
|
for timestep_spacing in ["trailing", "leading"]: |
|
self.check_over_configs(timestep_spacing=timestep_spacing) |
|
|
|
def test_rescale_betas_zero_snr(self): |
|
for rescale_betas_zero_snr in [True, False]: |
|
self.check_over_configs(rescale_betas_zero_snr=rescale_betas_zero_snr) |
|
|
|
def test_thresholding(self): |
|
self.check_over_configs(thresholding=False) |
|
for threshold in [0.5, 1.0, 2.0]: |
|
for prediction_type in ["epsilon", "v_prediction"]: |
|
self.check_over_configs( |
|
thresholding=True, |
|
prediction_type=prediction_type, |
|
sample_max_value=threshold, |
|
) |
|
|
|
def test_time_indices(self): |
|
for t in [1, 10, 49]: |
|
self.check_over_forward(time_step=t) |
|
|
|
def test_inference_steps(self): |
|
for t, num_inference_steps in zip([1, 10, 50], [10, 50, 500]): |
|
self.check_over_forward(time_step=t, num_inference_steps=num_inference_steps) |
|
|
|
def test_add_noise_device(self): |
|
pass |
|
|
|
def test_full_loop_no_noise(self): |
|
sample = self.full_loop() |
|
|
|
result_sum = torch.sum(torch.abs(sample)) |
|
result_mean = torch.mean(torch.abs(sample)) |
|
|
|
assert abs(result_sum.item() - 671.6816) < 1e-2 |
|
assert abs(result_mean.item() - 0.8746) < 1e-3 |
|
|
|
def test_full_loop_with_v_prediction(self): |
|
sample = self.full_loop(prediction_type="v_prediction") |
|
|
|
result_sum = torch.sum(torch.abs(sample)) |
|
result_mean = torch.mean(torch.abs(sample)) |
|
|
|
assert abs(result_sum.item() - 1394.2185) < 1e-2 |
|
assert abs(result_mean.item() - 1.8154) < 1e-3 |
|
|
|
def test_full_loop_with_set_alpha_to_one(self): |
|
|
|
sample = self.full_loop(set_alpha_to_one=True, beta_start=0.01) |
|
result_sum = torch.sum(torch.abs(sample)) |
|
result_mean = torch.mean(torch.abs(sample)) |
|
|
|
assert abs(result_sum.item() - 539.9622) < 1e-2 |
|
assert abs(result_mean.item() - 0.7031) < 1e-3 |
|
|
|
def test_full_loop_with_no_set_alpha_to_one(self): |
|
|
|
sample = self.full_loop(set_alpha_to_one=False, beta_start=0.01) |
|
result_sum = torch.sum(torch.abs(sample)) |
|
result_mean = torch.mean(torch.abs(sample)) |
|
|
|
assert abs(result_sum.item() - 542.6722) < 1e-2 |
|
assert abs(result_mean.item() - 0.7066) < 1e-3 |
|
|
