imaginAIry/imaginairy/samplers/ddim.py
Bryce 95a8fa31a9 fix: inpainting producing blurry images
while the previous version did produce much better blending it also makes images that lack detail for some reason.

tests: Added more tests to help catch this sort of thing earlies

fix: found that median blur is really slow, so I made sure we only do it on downsampled masks.  Was taking like 3 minutes to run on the large pearl girl picture on M1

- docs: update examples
2022-09-27 17:19:25 -07:00

392 lines
13 KiB
Python

# pylama:ignore=W0613
"""SAMPLING ONLY."""
import logging
import numpy as np
import torch
from tqdm import tqdm
from imaginairy.img_log import log_latent
from imaginairy.modules.diffusion.util import (
extract_into_tensor,
make_ddim_sampling_parameters,
make_ddim_timesteps,
noise_like,
)
from imaginairy.utils import get_device
logger = logging.getLogger(__name__)
class DDIMSampler:
"""
Denoising Diffusion Implicit Models
https://arxiv.org/abs/2010.02502
"""
def __init__(self, model):
self.model = model
def make_schedule(self, ddim_num_steps, ddim_discretize="uniform", ddim_eta=0.0):
buffers = self._make_schedule(
model_num_timesteps=self.model.num_timesteps,
model_alphas_cumprod=self.model.alphas_cumprod,
model_betas=self.model.betas,
model_alphas_cumprod_prev=self.model.alphas_cumprod_prev,
ddim_num_steps=ddim_num_steps,
ddim_discretize=ddim_discretize,
ddim_eta=ddim_eta,
device=self.model.device,
)
for k, v in buffers.items():
setattr(self, k, v)
@staticmethod
def _make_schedule(
model_num_timesteps,
model_alphas_cumprod,
model_betas,
model_alphas_cumprod_prev,
ddim_num_steps,
ddim_discretize="uniform",
ddim_eta=0.0,
device=get_device(),
):
ddim_timesteps = make_ddim_timesteps(
ddim_discr_method=ddim_discretize,
num_ddim_timesteps=ddim_num_steps,
num_ddpm_timesteps=model_num_timesteps,
)
alphas_cumprod = model_alphas_cumprod
if not alphas_cumprod.shape[0] == model_num_timesteps:
raise ValueError("alphas have to be defined for each timestep")
def to_torch(x):
return x.clone().detach().to(torch.float32).to(device)
# ddim sampling parameters
ddim_sigmas, ddim_alphas, ddim_alphas_prev = make_ddim_sampling_parameters(
alphacums=alphas_cumprod.cpu(),
ddim_timesteps=ddim_timesteps,
eta=ddim_eta,
)
buffers = {
"ddim_timesteps": ddim_timesteps,
"betas": to_torch(model_betas),
"alphas_cumprod": to_torch(alphas_cumprod),
"alphas_cumprod_prev": to_torch(model_alphas_cumprod_prev),
# calculations for diffusion q(x_t | x_{t-1}) and others
"sqrt_alphas_cumprod": to_torch(np.sqrt(alphas_cumprod.cpu())),
"sqrt_one_minus_alphas_cumprod": to_torch(
np.sqrt(1.0 - alphas_cumprod.cpu())
),
"log_one_minus_alphas_cumprod": to_torch(
np.log(1.0 - alphas_cumprod.cpu())
),
"sqrt_recip_alphas_cumprod": to_torch(np.sqrt(1.0 / alphas_cumprod.cpu())),
"sqrt_recipm1_alphas_cumprod": to_torch(
np.sqrt(1.0 / alphas_cumprod.cpu() - 1)
),
"ddim_sigmas": ddim_sigmas.to(torch.float32).to(device),
"ddim_alphas": ddim_alphas.to(torch.float32).to(device),
"ddim_alphas_prev": ddim_alphas_prev,
"ddim_sqrt_one_minus_alphas": np.sqrt(1.0 - ddim_alphas)
.to(torch.float32)
.to(device),
}
sigmas_for_original_sampling_steps = ddim_eta * torch.sqrt(
(1 - buffers["alphas_cumprod_prev"])
/ (1 - buffers["alphas_cumprod"])
* (1 - buffers["alphas_cumprod"] / buffers["alphas_cumprod_prev"])
)
buffers[
"ddim_sigmas_for_original_num_steps"
] = sigmas_for_original_sampling_steps
return buffers
@torch.no_grad()
def sample(
self,
num_steps,
batch_size,
shape,
conditioning,
callback=None,
normals_sequence=None,
img_callback=None,
quantize_x0=False,
eta=0.0,
mask=None,
x0=None,
temperature=1.0,
noise_dropout=0.0,
score_corrector=None,
corrector_kwargs=None,
x_T=None,
unconditional_guidance_scale=1.0,
unconditional_conditioning=None,
**kwargs,
# this has to come in the same format as the conditioning, # e.g. as encoded tokens, ...
):
if isinstance(conditioning, dict):
cbs = conditioning[list(conditioning.keys())[0]].shape[0]
if cbs != batch_size:
logger.warning(
f"Warning: Got {cbs} conditionings but batch-size is {batch_size}"
)
else:
if conditioning.shape[0] != batch_size:
logger.warning(
f"Warning: Got {conditioning.shape[0]} conditionings but batch-size is {batch_size}"
)
self.make_schedule(ddim_num_steps=num_steps, ddim_eta=eta)
samples = self.ddim_sampling(
conditioning,
shape=(batch_size, *shape),
callback=callback,
img_callback=img_callback,
quantize_denoised=quantize_x0,
mask=mask,
x0=x0,
noise_dropout=noise_dropout,
temperature=temperature,
score_corrector=score_corrector,
corrector_kwargs=corrector_kwargs,
x_T=x_T,
unconditional_guidance_scale=unconditional_guidance_scale,
unconditional_conditioning=unconditional_conditioning,
)
return samples
@torch.no_grad()
def ddim_sampling(
self,
cond,
shape,
x_T=None,
callback=None,
timesteps=None,
quantize_denoised=False,
mask=None,
x0=None,
img_callback=None,
temperature=1.0,
noise_dropout=0.0,
score_corrector=None,
corrector_kwargs=None,
unconditional_guidance_scale=1.0,
unconditional_conditioning=None,
):
device = self.model.betas.device
b = shape[0]
if x_T is None:
# run on CPU for seed consistency. M1/mps runs were not consistent otherwise
img = torch.randn(shape, device="cpu").to(device)
else:
img = x_T
log_latent(img, "initial noise")
if timesteps is None:
timesteps = self.ddim_timesteps
else:
subset_end = (
int(
min(timesteps / self.ddim_timesteps.shape[0], 1)
* self.ddim_timesteps.shape[0]
)
- 1
)
timesteps = self.ddim_timesteps[:subset_end]
time_range = np.flip(timesteps)
total_steps = timesteps.shape[0]
logger.info(f"Running DDIM Sampling with {total_steps} timesteps")
iterator = tqdm(time_range, desc="DDIM Sampler", total=total_steps)
for i, step in enumerate(iterator):
index = total_steps - i - 1
ts = torch.full((b,), step, device=device, dtype=torch.long)
if mask is not None:
assert x0 is not None
img_orig = self.model.q_sample(
x0, ts
) # TODO: deterministic forward pass?
img = img_orig * mask + (1.0 - mask) * img
img, pred_x0 = self.p_sample_ddim(
img,
cond,
ts,
index=index,
quantize_denoised=quantize_denoised,
temperature=temperature,
noise_dropout=noise_dropout,
unconditional_guidance_scale=unconditional_guidance_scale,
unconditional_conditioning=unconditional_conditioning,
)
if callback:
callback(i)
log_latent(img, "img")
log_latent(pred_x0, "pred_x0")
return img
def p_sample_ddim(
self,
x,
c,
t,
index,
repeat_noise=False,
quantize_denoised=False,
temperature=1.0,
noise_dropout=0.0,
unconditional_guidance_scale=1.0,
unconditional_conditioning=None,
loss_function=None,
):
assert unconditional_guidance_scale >= 1
x_in = torch.cat([x] * 2)
t_in = torch.cat([t] * 2)
c_in = torch.cat([unconditional_conditioning, c])
# with torch.no_grad():
noise_pred_uncond, noise_pred = self.model.apply_model(x_in, t_in, c_in).chunk(
2
)
noise_pred = noise_pred_uncond + unconditional_guidance_scale * (
noise_pred - noise_pred_uncond
)
b = x.shape[0]
log_latent(noise_pred, "noise prediction")
# select parameters corresponding to the currently considered timestep
a_t = torch.full((b, 1, 1, 1), self.ddim_alphas[index], device=x.device)
a_prev = torch.full((b, 1, 1, 1), self.ddim_alphas_prev[index], device=x.device)
sigma_t = torch.full((b, 1, 1, 1), self.ddim_sigmas[index], device=x.device)
sqrt_one_minus_at = torch.full(
(b, 1, 1, 1), self.ddim_sqrt_one_minus_alphas[index], device=x.device
)
return self._p_sample_ddim_formula(
x,
noise_pred,
sqrt_one_minus_at,
a_t,
sigma_t,
a_prev,
noise_dropout,
repeat_noise,
temperature,
)
@staticmethod
def _p_sample_ddim_formula(
x,
noise_pred,
sqrt_one_minus_at,
a_t,
sigma_t,
a_prev,
noise_dropout,
repeat_noise,
temperature,
):
# current prediction for x_0
pred_x0 = (x - sqrt_one_minus_at * noise_pred) / a_t.sqrt()
# direction pointing to x_t
dir_xt = (1.0 - a_prev - sigma_t**2).sqrt() * noise_pred
noise = sigma_t * noise_like(x.shape, x.device, repeat_noise) * temperature
if noise_dropout > 0.0:
noise = torch.nn.functional.dropout(noise, p=noise_dropout)
x_prev = a_prev.sqrt() * pred_x0 + dir_xt + noise
return x_prev, pred_x0
@torch.no_grad()
def stochastic_encode(self, init_latent, t, noise=None):
# fast, but does not allow for exact reconstruction
# t serves as an index to gather the correct alphas
t = t.clamp(0, 1000)
sqrt_alphas_cumprod = torch.sqrt(self.ddim_alphas)
sqrt_one_minus_alphas_cumprod = self.ddim_sqrt_one_minus_alphas
if noise is None:
noise = torch.randn_like(init_latent, device="cpu").to(get_device())
return (
extract_into_tensor(sqrt_alphas_cumprod, t, init_latent.shape) * init_latent
+ extract_into_tensor(sqrt_one_minus_alphas_cumprod, t, init_latent.shape)
* noise
)
@torch.no_grad()
def decode(
self,
x_latent,
cond,
t_start,
unconditional_guidance_scale=1.0,
unconditional_conditioning=None,
img_callback=None,
score_corrector=None,
temperature=1.0,
mask=None,
orig_latent=None,
):
timesteps = self.ddim_timesteps[:t_start]
time_range = np.flip(timesteps)
total_steps = timesteps.shape[0]
logger.debug(f"Running DDIM Sampling with {total_steps} timesteps")
iterator = tqdm(time_range, desc="Decoding image", total=total_steps)
x_dec = x_latent
for i, step in enumerate(iterator):
index = total_steps - i - 1
ts = torch.full(
(x_latent.shape[0],), step, device=x_latent.device, dtype=torch.long
)
if mask is not None:
assert orig_latent is not None
xdec_orig = self.model.q_sample(orig_latent, ts)
log_latent(xdec_orig, "xdec_orig")
# this helps prevent the weird disjointed images that can happen with masking
hint_strength = 0.8
if i < 2:
xdec_orig_with_hints = (
xdec_orig * (1 - hint_strength) + orig_latent * hint_strength
)
else:
xdec_orig_with_hints = xdec_orig
x_dec = xdec_orig_with_hints * mask + (1.0 - mask) * x_dec
log_latent(x_dec, "x_dec")
x_dec, pred_x0 = self.p_sample_ddim(
x_dec,
cond,
ts,
index=index,
unconditional_guidance_scale=unconditional_guidance_scale,
unconditional_conditioning=unconditional_conditioning,
temperature=temperature,
)
# original_loss = ((x_dec - x_latent).abs().mean()*70)
# sigma_t = torch.full((1, 1, 1, 1), self.ddim_sigmas[index], device=get_device())
# x_dec = x_dec.detach() + (original_loss * 0.1) ** 2
# cond_grad = -torch.autograd.grad(original_loss, x_dec)[0]
# x_dec = x_dec.detach() + cond_grad * sigma_t ** 2
# x_dec_alt = x_dec + (original_loss * 0.1) ** 2
log_latent(x_dec, f"x_dec {i}")
log_latent(pred_x0, f"pred_x0 {i}")
return x_dec