stable-diffusion-webui/modules/sd_samplers_compvis.py

221 lines
11 KiB
Python
Raw Normal View History

import math
import ldm.models.diffusion.ddim
import ldm.models.diffusion.plms
import numpy as np
import torch
from modules.shared import state
from modules import sd_samplers_common, prompt_parser, shared
2023-02-10 14:30:20 +03:00
import modules.models.diffusion.uni_pc
samplers_data_compvis = [
sd_samplers_common.SamplerData('DDIM', lambda model: VanillaStableDiffusionSampler(ldm.models.diffusion.ddim.DDIMSampler, model), [], {}),
sd_samplers_common.SamplerData('PLMS', lambda model: VanillaStableDiffusionSampler(ldm.models.diffusion.plms.PLMSSampler, model), [], {}),
2023-02-10 14:30:20 +03:00
sd_samplers_common.SamplerData('UniPC', lambda model: VanillaStableDiffusionSampler(modules.models.diffusion.uni_pc.UniPCSampler, model), [], {}),
]
class VanillaStableDiffusionSampler:
def __init__(self, constructor, sd_model):
self.sampler = constructor(sd_model)
2023-02-10 14:30:20 +03:00
self.is_ddim = hasattr(self.sampler, 'p_sample_ddim')
2022-11-26 16:10:46 +03:00
self.is_plms = hasattr(self.sampler, 'p_sample_plms')
2023-02-10 14:30:20 +03:00
self.is_unipc = isinstance(self.sampler, modules.models.diffusion.uni_pc.UniPCSampler)
self.orig_p_sample_ddim = None
if self.is_plms:
self.orig_p_sample_ddim = self.sampler.p_sample_plms
elif self.is_ddim:
self.orig_p_sample_ddim = self.sampler.p_sample_ddim
self.mask = None
self.nmask = None
self.init_latent = None
self.sampler_noises = None
2022-09-15 13:10:16 +03:00
self.step = 0
self.stop_at = None
self.eta = None
2022-10-06 14:12:52 +03:00
self.config = None
self.last_latent = None
2022-09-15 13:10:16 +03:00
self.conditioning_key = sd_model.model.conditioning_key
def number_of_needed_noises(self, p):
return 0
def launch_sampling(self, steps, func):
state.sampling_steps = steps
state.sampling_step = 0
try:
return func()
except sd_samplers_common.InterruptedException:
return self.last_latent
2022-09-15 13:10:16 +03:00
def p_sample_ddim_hook(self, x_dec, cond, ts, unconditional_conditioning, *args, **kwargs):
2023-02-10 14:30:20 +03:00
x_dec, ts, cond, unconditional_conditioning = self.before_sample(x_dec, ts, cond, unconditional_conditioning)
2023-05-10 11:19:16 +03:00
res = self.orig_p_sample_ddim(x_dec, cond, ts, *args, unconditional_conditioning=unconditional_conditioning, **kwargs)
2023-02-10 14:30:20 +03:00
x_dec, ts, cond, unconditional_conditioning, res = self.after_sample(x_dec, ts, cond, unconditional_conditioning, res)
return res
def before_sample(self, x, ts, cond, unconditional_conditioning):
if state.interrupted or state.skipped:
raise sd_samplers_common.InterruptedException
if self.stop_at is not None and self.step > self.stop_at:
raise sd_samplers_common.InterruptedException
# Have to unwrap the inpainting conditioning here to perform pre-processing
image_conditioning = None
uc_image_conditioning = None
if isinstance(cond, dict):
if self.conditioning_key == "crossattn-adm":
image_conditioning = cond["c_adm"]
uc_image_conditioning = unconditional_conditioning["c_adm"]
else:
image_conditioning = cond["c_concat"][0]
cond = cond["c_crossattn"][0]
unconditional_conditioning = unconditional_conditioning["c_crossattn"][0]
conds_list, tensor = prompt_parser.reconstruct_multicond_batch(cond, self.step)
2022-09-15 13:10:16 +03:00
unconditional_conditioning = prompt_parser.reconstruct_cond_batch(unconditional_conditioning, self.step)
2023-05-10 11:05:02 +03:00
assert all(len(conds) == 1 for conds in conds_list), 'composition via AND is not supported for DDIM/PLMS samplers'
cond = tensor
# for DDIM, shapes must match, we can't just process cond and uncond independently;
# filling unconditional_conditioning with repeats of the last vector to match length is
# not 100% correct but should work well enough
if unconditional_conditioning.shape[1] < cond.shape[1]:
last_vector = unconditional_conditioning[:, -1:]
last_vector_repeated = last_vector.repeat([1, cond.shape[1] - unconditional_conditioning.shape[1], 1])
unconditional_conditioning = torch.hstack([unconditional_conditioning, last_vector_repeated])
elif unconditional_conditioning.shape[1] > cond.shape[1]:
unconditional_conditioning = unconditional_conditioning[:, :cond.shape[1]]
2022-09-15 13:10:16 +03:00
if self.mask is not None:
img_orig = self.sampler.model.q_sample(self.init_latent, ts)
2023-02-10 14:30:20 +03:00
x = img_orig * self.mask + self.nmask * x
2022-09-15 13:10:16 +03:00
# Wrap the image conditioning back up since the DDIM code can accept the dict directly.
# Note that they need to be lists because it just concatenates them later.
if image_conditioning is not None:
if self.conditioning_key == "crossattn-adm":
cond = {"c_adm": image_conditioning, "c_crossattn": [cond]}
unconditional_conditioning = {"c_adm": uc_image_conditioning, "c_crossattn": [unconditional_conditioning]}
else:
cond = {"c_concat": [image_conditioning], "c_crossattn": [cond]}
unconditional_conditioning = {"c_concat": [image_conditioning], "c_crossattn": [unconditional_conditioning]}
2023-02-10 14:30:20 +03:00
return x, ts, cond, unconditional_conditioning
2023-02-10 15:47:08 +03:00
def update_step(self, last_latent):
2022-09-15 13:10:16 +03:00
if self.mask is not None:
2023-02-10 15:47:08 +03:00
self.last_latent = self.init_latent * self.mask + self.nmask * last_latent
2022-09-15 13:10:16 +03:00
else:
2023-02-10 15:47:08 +03:00
self.last_latent = last_latent
sd_samplers_common.store_latent(self.last_latent)
2022-09-15 13:10:16 +03:00
self.step += 1
state.sampling_step = self.step
shared.total_tqdm.update()
2023-02-10 15:47:08 +03:00
def after_sample(self, x, ts, cond, uncond, res):
if not self.is_unipc:
self.update_step(res[1])
2023-02-10 14:30:20 +03:00
return x, ts, cond, uncond, res
2023-02-10 15:47:08 +03:00
def unipc_after_update(self, x, model_x):
self.update_step(x)
def initialize(self, p):
self.eta = p.eta if p.eta is not None else shared.opts.eta_ddim
if self.eta != 0.0:
p.extra_generation_params["Eta DDIM"] = self.eta
2023-03-11 12:09:36 +03:00
if self.is_unipc:
keys = [
('UniPC variant', 'uni_pc_variant'),
('UniPC skip type', 'uni_pc_skip_type'),
('UniPC order', 'uni_pc_order'),
('UniPC lower order final', 'uni_pc_lower_order_final'),
]
for name, key in keys:
v = getattr(shared.opts, key)
if v != shared.opts.get_default(key):
p.extra_generation_params[name] = v
for fieldname in ['p_sample_ddim', 'p_sample_plms']:
if hasattr(self.sampler, fieldname):
setattr(self.sampler, fieldname, self.p_sample_ddim_hook)
2023-02-10 14:30:20 +03:00
if self.is_unipc:
2023-02-10 15:47:08 +03:00
self.sampler.set_hooks(lambda x, t, c, u: self.before_sample(x, t, c, u), lambda x, t, c, u, r: self.after_sample(x, t, c, u, r), lambda x, mx: self.unipc_after_update(x, mx))
self.mask = p.mask if hasattr(p, 'mask') else None
self.nmask = p.nmask if hasattr(p, 'nmask') else None
2023-03-11 12:09:36 +03:00
def adjust_steps_if_invalid(self, p, num_steps):
if ((self.config.name == 'DDIM') and p.ddim_discretize == 'uniform') or (self.config.name == 'PLMS') or (self.config.name == 'UniPC'):
2023-03-11 03:56:14 +03:00
if self.config.name == 'UniPC' and num_steps < shared.opts.uni_pc_order:
num_steps = shared.opts.uni_pc_order
valid_step = 999 / (1000 // num_steps)
if valid_step == math.floor(valid_step):
return int(valid_step) + 1
2023-03-11 03:56:14 +03:00
return num_steps
def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None, image_conditioning=None):
steps, t_enc = sd_samplers_common.setup_img2img_steps(p, steps)
steps = self.adjust_steps_if_invalid(p, steps)
2022-09-28 22:30:52 +03:00
self.initialize(p)
self.sampler.make_schedule(ddim_num_steps=steps, ddim_eta=self.eta, ddim_discretize=p.ddim_discretize, verbose=False)
x1 = self.sampler.stochastic_encode(x, torch.tensor([t_enc] * int(x.shape[0])).to(shared.device), noise=noise)
2022-09-19 16:42:56 +03:00
self.init_latent = x
self.last_latent = x
self.step = 0
2022-10-20 00:14:24 +03:00
# Wrap the conditioning models with additional image conditioning for inpainting model
if image_conditioning is not None:
if self.conditioning_key == "crossattn-adm":
conditioning = {"c_adm": image_conditioning, "c_crossattn": [conditioning]}
unconditional_conditioning = {"c_adm": torch.zeros_like(image_conditioning), "c_crossattn": [unconditional_conditioning]}
else:
conditioning = {"c_concat": [image_conditioning], "c_crossattn": [conditioning]}
unconditional_conditioning = {"c_concat": [image_conditioning], "c_crossattn": [unconditional_conditioning]}
2022-10-21 17:46:32 +03:00
samples = self.launch_sampling(t_enc + 1, lambda: self.sampler.decode(x1, conditioning, t_enc, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning))
return samples
def sample(self, p, x, conditioning, unconditional_conditioning, steps=None, image_conditioning=None):
self.initialize(p)
2022-09-08 19:20:41 +03:00
self.init_latent = None
self.last_latent = x
self.step = 0
2022-09-08 19:20:41 +03:00
steps = self.adjust_steps_if_invalid(p, steps or p.steps)
2022-09-19 16:42:56 +03:00
# Wrap the conditioning models with additional image conditioning for inpainting model
2022-11-26 16:10:46 +03:00
# dummy_for_plms is needed because PLMS code checks the first item in the dict to have the right shape
if image_conditioning is not None:
if self.conditioning_key == "crossattn-adm":
conditioning = {"dummy_for_plms": np.zeros((conditioning.shape[0],)), "c_crossattn": [conditioning], "c_adm": image_conditioning}
unconditional_conditioning = {"c_crossattn": [unconditional_conditioning], "c_adm": torch.zeros_like(image_conditioning)}
else:
conditioning = {"dummy_for_plms": np.zeros((conditioning.shape[0],)), "c_crossattn": [conditioning], "c_concat": [image_conditioning]}
unconditional_conditioning = {"c_crossattn": [unconditional_conditioning], "c_concat": [image_conditioning]}
samples_ddim = self.launch_sampling(steps, lambda: self.sampler.sample(S=steps, conditioning=conditioning, batch_size=int(x.shape[0]), shape=x[0].shape, verbose=False, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning, x_T=x, eta=self.eta)[0])
2022-09-13 20:12:24 +03:00
return samples_ddim