2022-11-25 14:34:12 +00:00
|
|
|
# Copyright 2022 Lunar Ring. All rights reserved.
|
2023-01-11 11:58:59 +00:00
|
|
|
# Written by Johannes Stelzer, email stelzer@lunar-ring.ai twitter @j_stelzer
|
2022-11-25 14:34:12 +00:00
|
|
|
#
|
|
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
# you may not use this file except in compliance with the License.
|
|
|
|
# You may obtain a copy of the License at
|
|
|
|
#
|
|
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
#
|
|
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
# See the License for the specific language governing permissions and
|
|
|
|
# limitations under the License.
|
|
|
|
|
2023-02-22 09:15:03 +00:00
|
|
|
import os
|
2022-11-25 14:34:12 +00:00
|
|
|
import torch
|
|
|
|
torch.backends.cudnn.benchmark = False
|
2023-02-22 09:15:03 +00:00
|
|
|
torch.set_grad_enabled(False)
|
2022-11-25 14:34:12 +00:00
|
|
|
import numpy as np
|
|
|
|
import warnings
|
|
|
|
warnings.filterwarnings('ignore')
|
|
|
|
import warnings
|
|
|
|
import torch
|
|
|
|
from PIL import Image
|
|
|
|
import torch
|
2023-02-22 09:15:03 +00:00
|
|
|
from typing import Optional
|
2022-11-25 14:34:12 +00:00
|
|
|
from omegaconf import OmegaConf
|
|
|
|
from torch import autocast
|
|
|
|
from contextlib import nullcontext
|
|
|
|
from ldm.util import instantiate_from_config
|
|
|
|
from ldm.models.diffusion.ddim import DDIMSampler
|
2023-01-08 09:32:58 +00:00
|
|
|
from einops import repeat, rearrange
|
2023-02-22 09:15:03 +00:00
|
|
|
from utils import interpolate_spherical
|
2023-01-08 09:32:58 +00:00
|
|
|
|
|
|
|
|
|
|
|
def pad_image(input_image):
|
|
|
|
pad_w, pad_h = np.max(((2, 2), np.ceil(
|
|
|
|
np.array(input_image.size) / 64).astype(int)), axis=0) * 64 - input_image.size
|
|
|
|
im_padded = Image.fromarray(
|
|
|
|
np.pad(np.array(input_image), ((0, pad_h), (0, pad_w), (0, 0)), mode='edge'))
|
|
|
|
return im_padded
|
|
|
|
|
|
|
|
|
|
|
|
def make_batch_superres(
|
|
|
|
image,
|
|
|
|
txt,
|
|
|
|
device,
|
2023-02-22 09:15:03 +00:00
|
|
|
num_samples=1):
|
2023-01-08 09:32:58 +00:00
|
|
|
image = np.array(image.convert("RGB"))
|
|
|
|
image = torch.from_numpy(image).to(dtype=torch.float32) / 127.5 - 1.0
|
|
|
|
batch = {
|
|
|
|
"lr": rearrange(image, 'h w c -> 1 c h w'),
|
|
|
|
"txt": num_samples * [txt],
|
|
|
|
}
|
|
|
|
batch["lr"] = repeat(batch["lr"].to(device=device),
|
|
|
|
"1 ... -> n ...", n=num_samples)
|
|
|
|
return batch
|
|
|
|
|
|
|
|
|
|
|
|
def make_noise_augmentation(model, batch, noise_level=None):
|
|
|
|
x_low = batch[model.low_scale_key]
|
|
|
|
x_low = x_low.to(memory_format=torch.contiguous_format).float()
|
|
|
|
x_aug, noise_level = model.low_scale_model(x_low, noise_level)
|
|
|
|
return x_aug, noise_level
|
|
|
|
|
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
class StableDiffusionHolder:
|
2023-02-22 09:15:03 +00:00
|
|
|
def __init__(self,
|
|
|
|
fp_ckpt: str = None,
|
2022-11-25 14:34:12 +00:00
|
|
|
fp_config: str = None,
|
2023-02-22 09:15:03 +00:00
|
|
|
num_inference_steps: int = 30,
|
2022-11-25 14:34:12 +00:00
|
|
|
height: Optional[int] = None,
|
|
|
|
width: Optional[int] = None,
|
2023-01-08 09:32:58 +00:00
|
|
|
device: str = None,
|
2023-02-22 09:15:03 +00:00
|
|
|
precision: str = 'autocast',
|
2022-11-25 14:34:12 +00:00
|
|
|
):
|
2023-01-09 08:58:18 +00:00
|
|
|
r"""
|
|
|
|
Initializes the stable diffusion holder, which contains the models and sampler.
|
|
|
|
Args:
|
|
|
|
fp_ckpt: File pointer to the .ckpt model file
|
|
|
|
fp_config: File pointer to the .yaml config file
|
|
|
|
num_inference_steps: Number of diffusion iterations. Will be overwritten by latent blending.
|
2023-02-22 09:15:03 +00:00
|
|
|
height: Height of the resulting image.
|
|
|
|
width: Width of the resulting image.
|
2023-01-09 08:58:18 +00:00
|
|
|
device: Device to run the model on.
|
|
|
|
precision: Precision to run the model on.
|
|
|
|
"""
|
2022-11-25 14:34:12 +00:00
|
|
|
self.seed = 42
|
|
|
|
self.guidance_scale = 5.0
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
if device is None:
|
|
|
|
self.device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
|
|
|
|
else:
|
|
|
|
self.device = device
|
|
|
|
self.precision = precision
|
|
|
|
self.init_model(fp_ckpt, fp_config)
|
2023-02-22 09:15:03 +00:00
|
|
|
|
|
|
|
self.f = 8 # downsampling factor, most often 8 or 16"
|
2022-11-25 14:34:12 +00:00
|
|
|
self.C = 4
|
|
|
|
self.ddim_eta = 0
|
|
|
|
self.num_inference_steps = num_inference_steps
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2022-12-31 12:14:37 +00:00
|
|
|
if height is None and width is None:
|
2022-11-25 14:34:12 +00:00
|
|
|
self.init_auto_res()
|
2022-12-31 12:14:37 +00:00
|
|
|
else:
|
|
|
|
assert height is not None, "specify both width and height"
|
|
|
|
assert width is not None, "specify both width and height"
|
|
|
|
self.height = height
|
|
|
|
self.width = width
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-08 10:48:44 +00:00
|
|
|
self.negative_prompt = [""]
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
def init_model(self, fp_ckpt, fp_config):
|
2023-01-09 08:58:18 +00:00
|
|
|
r"""Loads the models and sampler.
|
|
|
|
"""
|
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
assert os.path.isfile(fp_ckpt), f"Your model checkpoint file does not exist: {fp_ckpt}"
|
2023-01-08 09:32:58 +00:00
|
|
|
self.fp_ckpt = fp_ckpt
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-12 09:06:02 +00:00
|
|
|
# Auto init the config?
|
|
|
|
if fp_config is None:
|
|
|
|
fn_ckpt = os.path.basename(fp_ckpt)
|
|
|
|
if 'depth' in fn_ckpt:
|
|
|
|
fp_config = 'configs/v2-midas-inference.yaml'
|
|
|
|
elif 'upscaler' in fn_ckpt:
|
2023-02-22 09:15:03 +00:00
|
|
|
fp_config = 'configs/x4-upscaling.yaml'
|
2023-01-12 09:06:02 +00:00
|
|
|
elif '512' in fn_ckpt:
|
2023-02-22 09:15:03 +00:00
|
|
|
fp_config = 'configs/v2-inference.yaml'
|
|
|
|
elif '768' in fn_ckpt:
|
|
|
|
fp_config = 'configs/v2-inference-v.yaml'
|
2023-01-14 20:04:35 +00:00
|
|
|
elif 'v1-5' in fn_ckpt:
|
2023-02-22 09:15:03 +00:00
|
|
|
fp_config = 'configs/v1-inference.yaml'
|
2023-01-12 09:06:02 +00:00
|
|
|
else:
|
|
|
|
raise ValueError("auto detect of config failed. please specify fp_config manually!")
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-12 09:09:19 +00:00
|
|
|
assert os.path.isfile(fp_config), "Auto-init of the config file failed. Please specify manually."
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-12 09:06:02 +00:00
|
|
|
assert os.path.isfile(fp_config), f"Your config file does not exist: {fp_config}"
|
2022-11-25 14:34:12 +00:00
|
|
|
|
2023-01-08 09:32:58 +00:00
|
|
|
config = OmegaConf.load(fp_config)
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-08 09:32:58 +00:00
|
|
|
self.model = instantiate_from_config(config.model)
|
|
|
|
self.model.load_state_dict(torch.load(fp_ckpt)["state_dict"], strict=False)
|
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
self.model = self.model.to(self.device)
|
|
|
|
self.sampler = DDIMSampler(self.model)
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
def init_auto_res(self):
|
|
|
|
r"""Automatically set the resolution to the one used in training.
|
|
|
|
"""
|
|
|
|
if '768' in self.fp_ckpt:
|
|
|
|
self.height = 768
|
|
|
|
self.width = 768
|
|
|
|
else:
|
|
|
|
self.height = 512
|
|
|
|
self.width = 512
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-12 09:06:02 +00:00
|
|
|
def set_negative_prompt(self, negative_prompt):
|
|
|
|
r"""Set the negative prompt. Currenty only one negative prompt is supported
|
|
|
|
"""
|
|
|
|
|
|
|
|
if isinstance(negative_prompt, str):
|
|
|
|
self.negative_prompt = [negative_prompt]
|
|
|
|
else:
|
|
|
|
self.negative_prompt = negative_prompt
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-12 09:06:02 +00:00
|
|
|
if len(self.negative_prompt) > 1:
|
|
|
|
self.negative_prompt = [self.negative_prompt[0]]
|
2022-11-25 14:34:12 +00:00
|
|
|
|
|
|
|
def get_text_embedding(self, prompt):
|
|
|
|
c = self.model.get_learned_conditioning(prompt)
|
|
|
|
return c
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-08 09:32:58 +00:00
|
|
|
@torch.no_grad()
|
|
|
|
def get_cond_upscaling(self, image, text_embedding, noise_level):
|
|
|
|
r"""
|
|
|
|
Initializes the conditioning for the x4 upscaling model.
|
|
|
|
"""
|
|
|
|
image = pad_image(image) # resize to integer multiple of 32
|
|
|
|
w, h = image.size
|
|
|
|
noise_level = torch.Tensor(1 * [noise_level]).to(self.sampler.model.device).long()
|
|
|
|
batch = make_batch_superres(image, txt="placeholder", device=self.device, num_samples=1)
|
|
|
|
|
|
|
|
x_augment, noise_level = make_noise_augmentation(self.model, batch, noise_level)
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-01-08 09:32:58 +00:00
|
|
|
cond = {"c_concat": [x_augment], "c_crossattn": [text_embedding], "c_adm": noise_level}
|
|
|
|
# uncond cond
|
|
|
|
uc_cross = self.model.get_unconditional_conditioning(1, "")
|
|
|
|
uc_full = {"c_concat": [x_augment], "c_crossattn": [uc_cross], "c_adm": noise_level}
|
|
|
|
return cond, uc_full
|
2022-11-25 14:34:12 +00:00
|
|
|
|
|
|
|
@torch.no_grad()
|
|
|
|
def run_diffusion_standard(
|
2023-02-22 09:15:03 +00:00
|
|
|
self,
|
|
|
|
text_embeddings: torch.FloatTensor,
|
2023-02-16 10:48:45 +00:00
|
|
|
latents_start: torch.FloatTensor,
|
2023-02-22 09:15:03 +00:00
|
|
|
idx_start: int = 0,
|
|
|
|
list_latents_mixing=None,
|
|
|
|
mixing_coeffs=0.0,
|
|
|
|
spatial_mask=None,
|
|
|
|
return_image: Optional[bool] = False):
|
2022-11-25 14:34:12 +00:00
|
|
|
r"""
|
2023-02-22 09:15:03 +00:00
|
|
|
Diffusion standard version.
|
2022-11-25 14:34:12 +00:00
|
|
|
Args:
|
2023-02-22 09:15:03 +00:00
|
|
|
text_embeddings: torch.FloatTensor
|
2022-11-25 14:34:12 +00:00
|
|
|
Text embeddings used for diffusion
|
2023-02-15 17:21:00 +00:00
|
|
|
latents_for_injection: torch.FloatTensor or list
|
2022-11-25 14:34:12 +00:00
|
|
|
Latents that are used for injection
|
|
|
|
idx_start: int
|
|
|
|
Index of the diffusion process start and where the latents_for_injection are injected
|
2023-02-15 17:21:00 +00:00
|
|
|
mixing_coeff:
|
2023-02-20 10:26:04 +00:00
|
|
|
mixing coefficients for latent blending
|
|
|
|
spatial_mask:
|
|
|
|
experimental feature for enforcing pixels from list_latents_mixing
|
2022-11-25 14:34:12 +00:00
|
|
|
return_image: Optional[bool]
|
|
|
|
Optionally return image directly
|
|
|
|
"""
|
2023-02-16 10:48:45 +00:00
|
|
|
# Asserts
|
|
|
|
if type(mixing_coeffs) == float:
|
2023-02-22 09:15:03 +00:00
|
|
|
list_mixing_coeffs = self.num_inference_steps * [mixing_coeffs]
|
2023-02-16 10:48:45 +00:00
|
|
|
elif type(mixing_coeffs) == list:
|
|
|
|
assert len(mixing_coeffs) == self.num_inference_steps
|
|
|
|
list_mixing_coeffs = mixing_coeffs
|
2022-11-25 14:34:12 +00:00
|
|
|
else:
|
2023-02-16 10:48:45 +00:00
|
|
|
raise ValueError("mixing_coeffs should be float or list with len=num_inference_steps")
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-02-16 10:48:45 +00:00
|
|
|
if np.sum(list_mixing_coeffs) > 0:
|
|
|
|
assert len(list_latents_mixing) == self.num_inference_steps
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
precision_scope = autocast if self.precision == "autocast" else nullcontext
|
2022-12-09 14:03:20 +00:00
|
|
|
with precision_scope("cuda"):
|
|
|
|
with self.model.ema_scope():
|
|
|
|
if self.guidance_scale != 1.0:
|
2023-01-08 10:48:44 +00:00
|
|
|
uc = self.model.get_learned_conditioning(self.negative_prompt)
|
2022-12-09 14:03:20 +00:00
|
|
|
else:
|
|
|
|
uc = None
|
2023-02-22 09:15:03 +00:00
|
|
|
self.sampler.make_schedule(ddim_num_steps=self.num_inference_steps - 1, ddim_eta=self.ddim_eta, verbose=False)
|
2023-02-16 10:48:45 +00:00
|
|
|
latents = latents_start.clone()
|
2022-12-09 14:03:20 +00:00
|
|
|
timesteps = self.sampler.ddim_timesteps
|
|
|
|
time_range = np.flip(timesteps)
|
|
|
|
total_steps = timesteps.shape[0]
|
2023-02-22 09:15:03 +00:00
|
|
|
# Collect latents
|
2022-12-09 14:03:20 +00:00
|
|
|
list_latents_out = []
|
|
|
|
for i, step in enumerate(time_range):
|
2023-02-16 10:48:45 +00:00
|
|
|
# Set the right starting latents
|
|
|
|
if i < idx_start:
|
|
|
|
list_latents_out.append(None)
|
|
|
|
continue
|
|
|
|
elif i == idx_start:
|
|
|
|
latents = latents_start.clone()
|
2023-02-22 09:15:03 +00:00
|
|
|
# Mix latents
|
|
|
|
if i > 0 and list_mixing_coeffs[i] > 0:
|
|
|
|
latents_mixtarget = list_latents_mixing[i - 1].clone()
|
2023-02-16 10:48:45 +00:00
|
|
|
latents = interpolate_spherical(latents, latents_mixtarget, list_mixing_coeffs[i])
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-02-19 14:32:37 +00:00
|
|
|
if spatial_mask is not None and list_latents_mixing is not None:
|
2023-02-22 09:15:03 +00:00
|
|
|
latents = interpolate_spherical(latents, list_latents_mixing[i - 1], 1 - spatial_mask)
|
|
|
|
|
2022-12-09 14:03:20 +00:00
|
|
|
index = total_steps - i - 1
|
2023-02-16 10:48:45 +00:00
|
|
|
ts = torch.full((1,), step, device=self.device, dtype=torch.long)
|
2022-12-09 14:03:20 +00:00
|
|
|
outs = self.sampler.p_sample_ddim(latents, text_embeddings, ts, index=index, use_original_steps=False,
|
2023-02-22 09:15:03 +00:00
|
|
|
quantize_denoised=False, temperature=1.0,
|
|
|
|
noise_dropout=0.0, score_corrector=None,
|
|
|
|
corrector_kwargs=None,
|
|
|
|
unconditional_guidance_scale=self.guidance_scale,
|
|
|
|
unconditional_conditioning=uc,
|
|
|
|
dynamic_threshold=None)
|
2022-12-09 14:03:20 +00:00
|
|
|
latents, pred_x0 = outs
|
|
|
|
list_latents_out.append(latents.clone())
|
2023-02-22 09:15:03 +00:00
|
|
|
if return_image:
|
2022-12-09 14:03:20 +00:00
|
|
|
return self.latent2image(latents)
|
|
|
|
else:
|
|
|
|
return list_latents_out
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
@torch.no_grad()
|
2023-02-18 06:56:30 +00:00
|
|
|
def run_diffusion_upscaling(
|
2023-02-22 09:15:03 +00:00
|
|
|
self,
|
2023-02-18 06:56:30 +00:00
|
|
|
cond,
|
|
|
|
uc_full,
|
2023-02-22 09:15:03 +00:00
|
|
|
latents_start: torch.FloatTensor,
|
|
|
|
idx_start: int = -1,
|
|
|
|
list_latents_mixing: list = None,
|
|
|
|
mixing_coeffs: float = 0.0,
|
|
|
|
return_image: Optional[bool] = False):
|
2022-11-25 14:34:12 +00:00
|
|
|
r"""
|
2023-02-22 09:15:03 +00:00
|
|
|
Diffusion upscaling version.
|
2023-02-18 06:56:30 +00:00
|
|
|
"""
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-02-18 06:56:30 +00:00
|
|
|
# Asserts
|
|
|
|
if type(mixing_coeffs) == float:
|
2023-02-22 09:15:03 +00:00
|
|
|
list_mixing_coeffs = self.num_inference_steps * [mixing_coeffs]
|
2023-02-18 06:56:30 +00:00
|
|
|
elif type(mixing_coeffs) == list:
|
|
|
|
assert len(mixing_coeffs) == self.num_inference_steps
|
|
|
|
list_mixing_coeffs = mixing_coeffs
|
2022-11-25 14:34:12 +00:00
|
|
|
else:
|
2023-02-18 06:56:30 +00:00
|
|
|
raise ValueError("mixing_coeffs should be float or list with len=num_inference_steps")
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2023-02-18 06:56:30 +00:00
|
|
|
if np.sum(list_mixing_coeffs) > 0:
|
|
|
|
assert len(list_latents_mixing) == self.num_inference_steps
|
2023-02-22 09:15:03 +00:00
|
|
|
|
2022-11-25 14:34:12 +00:00
|
|
|
precision_scope = autocast if self.precision == "autocast" else nullcontext
|
2023-02-22 09:15:03 +00:00
|
|
|
h = uc_full['c_concat'][0].shape[2]
|
|
|
|
w = uc_full['c_concat'][0].shape[3]
|
2022-12-09 14:03:20 +00:00
|
|
|
with precision_scope("cuda"):
|
|
|
|
with self.model.ema_scope():
|
2023-02-18 06:56:30 +00:00
|
|
|
|
|
|
|
shape_latents = [self.model.channels, h, w]
|
2023-02-22 09:15:03 +00:00
|
|
|
self.sampler.make_schedule(ddim_num_steps=self.num_inference_steps - 1, ddim_eta=self.ddim_eta, verbose=False)
|
2022-12-09 14:03:20 +00:00
|
|
|
C, H, W = shape_latents
|
|
|
|
size = (1, C, H, W)
|
|
|
|
b = size[0]
|
2023-02-18 06:56:30 +00:00
|
|
|
latents = latents_start.clone()
|
2022-12-09 14:03:20 +00:00
|
|
|
timesteps = self.sampler.ddim_timesteps
|
|
|
|
time_range = np.flip(timesteps)
|
|
|
|
total_steps = timesteps.shape[0]
|
|
|
|
# collect latents
|
|
|
|
list_latents_out = []
|
|
|
|
for i, step in enumerate(time_range):
|
2023-02-18 06:56:30 +00:00
|
|
|
# Set the right starting latents
|
|
|
|
if i < idx_start:
|
|
|
|
list_latents_out.append(None)
|
|
|
|
continue
|
|
|
|
elif i == idx_start:
|
|
|
|
latents = latents_start.clone()
|
2023-02-22 09:15:03 +00:00
|
|
|
# Mix the latents.
|
|
|
|
if i > 0 and list_mixing_coeffs[i] > 0:
|
|
|
|
latents_mixtarget = list_latents_mixing[i - 1].clone()
|
2023-02-18 06:56:30 +00:00
|
|
|
latents = interpolate_spherical(latents, latents_mixtarget, list_mixing_coeffs[i])
|
|
|
|
# print(f"diffusion iter {i}")
|
2022-12-09 14:03:20 +00:00
|
|
|
index = total_steps - i - 1
|
2023-02-18 06:56:30 +00:00
|
|
|
ts = torch.full((b,), step, device=self.device, dtype=torch.long)
|
2022-12-09 14:03:20 +00:00
|
|
|
outs = self.sampler.p_sample_ddim(latents, cond, ts, index=index, use_original_steps=False,
|
2023-02-22 09:15:03 +00:00
|
|
|
quantize_denoised=False, temperature=1.0,
|
|
|
|
noise_dropout=0.0, score_corrector=None,
|
|
|
|
corrector_kwargs=None,
|
|
|
|
unconditional_guidance_scale=self.guidance_scale,
|
|
|
|
unconditional_conditioning=uc_full,
|
|
|
|
dynamic_threshold=None)
|
2022-12-09 14:03:20 +00:00
|
|
|
latents, pred_x0 = outs
|
|
|
|
list_latents_out.append(latents.clone())
|
2023-02-18 06:56:30 +00:00
|
|
|
|
2023-02-22 09:15:03 +00:00
|
|
|
if return_image:
|
2023-01-08 09:32:58 +00:00
|
|
|
return self.latent2image(latents)
|
|
|
|
else:
|
2023-02-18 06:56:30 +00:00
|
|
|
return list_latents_out
|
2022-11-25 14:34:12 +00:00
|
|
|
|
|
|
|
@torch.no_grad()
|
|
|
|
def latent2image(
|
2023-02-22 09:15:03 +00:00
|
|
|
self,
|
|
|
|
latents: torch.FloatTensor):
|
2022-11-25 14:34:12 +00:00
|
|
|
r"""
|
|
|
|
Returns an image provided a latent representation from diffusion.
|
|
|
|
Args:
|
|
|
|
latents: torch.FloatTensor
|
2023-02-22 09:15:03 +00:00
|
|
|
Result of the diffusion process.
|
2022-11-25 14:34:12 +00:00
|
|
|
"""
|
|
|
|
x_sample = self.model.decode_first_stage(latents)
|
|
|
|
x_sample = torch.clamp((x_sample + 1.0) / 2.0, min=0.0, max=1.0)
|
2023-02-22 09:15:03 +00:00
|
|
|
x_sample = 255 * x_sample[0, :, :].permute([1, 2, 0]).cpu().numpy()
|
2022-11-25 14:34:12 +00:00
|
|
|
image = x_sample.astype(np.uint8)
|
|
|
|
return image
|