forked from deforum/stable-diffusion
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathDeforum_Stable_Diffusion.py
1379 lines (1166 loc) · 54.2 KB
/
Deforum_Stable_Diffusion.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# %%
# !! {"metadata":{
# !! "id": "c442uQJ_gUgy"
# !! }}
"""
# **Deforum Stable Diffusion v0.3**
[Stable Diffusion](https://github.com/CompVis/stable-diffusion) by Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, Björn Ommer and the [Stability.ai](https://stability.ai/) Team. [K Diffusion](https://github.com/crowsonkb/k-diffusion) by [Katherine Crowson](https://twitter.com/RiversHaveWings). You need to get the ckpt file and put it on your Google Drive first to use this. It can be downloaded from [HuggingFace](https://huggingface.co/CompVis/stable-diffusion).
Notebook by [deforum](https://discord.gg/upmXXsrwZc)
"""
# %%
# !! {"metadata":{
# !! "id": "T4knibRpAQ06"
# !! }}
"""
# Setup
"""
# %%
# !! {"metadata":{
# !! "id": "2g-f7cQmf2Nt",
# !! "cellView": "form"
# !! }}
#@markdown **NVIDIA GPU**
import subprocess
sub_p_res = subprocess.run(['nvidia-smi', '--query-gpu=name,memory.total,memory.free', '--format=csv,noheader'], stdout=subprocess.PIPE).stdout.decode('utf-8')
print(sub_p_res)
# %%
# !! {"metadata":{
# !! "cellView": "form",
# !! "id": "TxIOPT0G5Lx1"
# !! }}
#@markdown **Model and Output Paths**
# ask for the link
print("Local Path Variables:\n")
models_path = "/content/models" #@param {type:"string"}
output_path = "/content/output" #@param {type:"string"}
#@markdown **Google Drive Path Variables (Optional)**
mount_google_drive = True #@param {type:"boolean"}
force_remount = False
if mount_google_drive:
from google.colab import drive # type: ignore
try:
drive_path = "/content/drive"
drive.mount(drive_path,force_remount=force_remount)
models_path_gdrive = "/content/drive/MyDrive/AI/models" #@param {type:"string"}
output_path_gdrive = "/content/drive/MyDrive/AI/StableDiffusion" #@param {type:"string"}
models_path = models_path_gdrive
output_path = output_path_gdrive
except:
print("...error mounting drive or with drive path variables")
print("...reverting to default path variables")
import os
os.makedirs(models_path, exist_ok=True)
os.makedirs(output_path, exist_ok=True)
print(f"models_path: {models_path}")
print(f"output_path: {output_path}")
# %%
# !! {"metadata":{
# !! "id": "VRNl2mfepEIe",
# !! "cellView": "form"
# !! }}
#@markdown **Setup Environment**
setup_environment = True #@param {type:"boolean"}
print_subprocess = False #@param {type:"boolean"}
if setup_environment:
import subprocess, time
print("Setting up environment...")
start_time = time.time()
all_process = [
['pip', 'install', 'torch==1.12.1+cu113', 'torchvision==0.13.1+cu113', '--extra-index-url', 'https://download.pytorch.org/whl/cu113'],
['pip', 'install', 'omegaconf==2.2.3', 'einops==0.4.1', 'pytorch-lightning==1.7.4', 'torchmetrics==0.9.3', 'torchtext==0.13.1', 'transformers==4.21.2', 'kornia==0.6.7'],
['git', 'clone', 'https://github.com/deforum/stable-diffusion'],
['pip', 'install', '-e', 'git+https://github.com/CompVis/taming-transformers.git@master#egg=taming-transformers'],
['pip', 'install', '-e', 'git+https://github.com/openai/CLIP.git@main#egg=clip'],
['pip', 'install', 'accelerate', 'ftfy', 'jsonmerge', 'matplotlib', 'resize-right', 'timm', 'torchdiffeq'],
['git', 'clone', 'https://github.com/shariqfarooq123/AdaBins.git'],
['git', 'clone', 'https://github.com/isl-org/MiDaS.git'],
['git', 'clone', 'https://github.com/MSFTserver/pytorch3d-lite.git'],
]
for process in all_process:
running = subprocess.run(process,stdout=subprocess.PIPE).stdout.decode('utf-8')
if print_subprocess:
print(running)
print(subprocess.run(['git', 'clone', 'https://github.com/deforum/k-diffusion/'], stdout=subprocess.PIPE).stdout.decode('utf-8'))
with open('k-diffusion/k_diffusion/__init__.py', 'w') as f:
f.write('')
end_time = time.time()
print(f"Environment set up in {end_time-start_time:.0f} seconds")
# %%
# !! {"metadata":{
# !! "id": "81qmVZbrm4uu",
# !! "cellView": "form"
# !! }}
#@markdown **Python Definitions**
import json
from IPython import display
import math, os, pathlib, shutil, subprocess, sys, time
import cv2
import numpy as np
import pandas as pd
import random
import requests
import torch, torchvision
import torch.nn as nn
import torchvision.transforms as T
import torchvision.transforms.functional as TF
from contextlib import contextmanager, nullcontext
from einops import rearrange, repeat
from itertools import islice
from omegaconf import OmegaConf
from PIL import Image
from pytorch_lightning import seed_everything
from skimage.exposure import match_histograms
from torchvision.utils import make_grid
from tqdm import tqdm, trange
from types import SimpleNamespace
from torch import autocast
sys.path.extend([
'src/taming-transformers',
'src/clip',
'stable-diffusion/',
'k-diffusion',
'pytorch3d-lite',
'AdaBins',
'MiDaS',
])
import py3d_tools as p3d
from helpers import save_samples, sampler_fn
from infer import InferenceHelper
from k_diffusion import sampling
from k_diffusion.external import CompVisDenoiser
from ldm.util import instantiate_from_config
from ldm.models.diffusion.ddim import DDIMSampler
from ldm.models.diffusion.plms import PLMSSampler
from midas.dpt_depth import DPTDepthModel
from midas.transforms import Resize, NormalizeImage, PrepareForNet
def sanitize(prompt):
whitelist = set('abcdefghijklmnopqrstuvwxyz ABCDEFGHIJKLMNOPQRSTUVWXYZ')
tmp = ''.join(filter(whitelist.__contains__, prompt))
return '_'.join(prompt.split(" "))
def anim_frame_warp_2d(prev_img_cv2, args, anim_args, keys, frame_idx):
angle = keys.angle_series[frame_idx]
zoom = keys.zoom_series[frame_idx]
translation_x = keys.translation_x_series[frame_idx]
translation_y = keys.translation_y_series[frame_idx]
center = (args.W // 2, args.H // 2)
trans_mat = np.float32([[1, 0, translation_x], [0, 1, translation_y]])
rot_mat = cv2.getRotationMatrix2D(center, angle, zoom)
trans_mat = np.vstack([trans_mat, [0,0,1]])
rot_mat = np.vstack([rot_mat, [0,0,1]])
xform = np.matmul(rot_mat, trans_mat)
return cv2.warpPerspective(
prev_img_cv2,
xform,
(prev_img_cv2.shape[1], prev_img_cv2.shape[0]),
borderMode=cv2.BORDER_WRAP if anim_args.border == 'wrap' else cv2.BORDER_REPLICATE
)
def anim_frame_warp_3d(prev_img_cv2, anim_args, keys, frame_idx, adabins_helper, midas_model, midas_transform):
TRANSLATION_SCALE = 1.0/200.0 # matches Disco
translate_xyz = [
-keys.translation_x_series[frame_idx] * TRANSLATION_SCALE,
keys.translation_y_series[frame_idx] * TRANSLATION_SCALE,
-keys.translation_z_series[frame_idx] * TRANSLATION_SCALE
]
rotate_xyz = [
math.radians(keys.rotation_3d_x_series[frame_idx]),
math.radians(keys.rotation_3d_y_series[frame_idx]),
math.radians(keys.rotation_3d_z_series[frame_idx])
]
rot_mat = p3d.euler_angles_to_matrix(torch.tensor(rotate_xyz, device=device), "XYZ").unsqueeze(0)
result = transform_image_3d(prev_img_cv2, adabins_helper, midas_model, midas_transform, rot_mat, translate_xyz, anim_args)
torch.cuda.empty_cache()
return result
def add_noise(sample: torch.Tensor, noise_amt: float):
return sample + torch.randn(sample.shape, device=sample.device) * noise_amt
def download_depth_models():
def wget(url, outputdir):
print(subprocess.run(['wget', url, '-P', outputdir], stdout=subprocess.PIPE).stdout.decode('utf-8'))
if not os.path.exists(os.path.join(models_path, 'dpt_large-midas-2f21e586.pt')):
print("Downloading dpt_large-midas-2f21e586.pt...")
wget("https://github.com/intel-isl/DPT/releases/download/1_0/dpt_large-midas-2f21e586.pt", models_path)
if not os.path.exists('pretrained/AdaBins_nyu.pt'):
print("Downloading AdaBins_nyu.pt...")
os.makedirs('pretrained', exist_ok=True)
wget("https://cloudflare-ipfs.com/ipfs/Qmd2mMnDLWePKmgfS8m6ntAg4nhV5VkUyAydYBp8cWWeB7/AdaBins_nyu.pt", 'pretrained')
def get_output_folder(output_path, batch_folder):
out_path = os.path.join(output_path,time.strftime('%Y-%m'))
if batch_folder != "":
out_path = os.path.join(out_path, batch_folder)
os.makedirs(out_path, exist_ok=True)
return out_path
def load_depth_model(optimize=True):
midas_model = DPTDepthModel(
path=f"{models_path}/dpt_large-midas-2f21e586.pt",
backbone="vitl16_384",
non_negative=True,
)
normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
midas_transform = T.Compose([
Resize(
384, 384,
resize_target=None,
keep_aspect_ratio=True,
ensure_multiple_of=32,
resize_method="minimal",
image_interpolation_method=cv2.INTER_CUBIC,
),
normalization,
PrepareForNet()
])
midas_model.eval()
if optimize:
if device == torch.device("cuda"):
midas_model = midas_model.to(memory_format=torch.channels_last)
midas_model = midas_model.half()
midas_model.to(device)
return midas_model, midas_transform
def load_img(path, shape):
if path.startswith('http://') or path.startswith('https://'):
image = Image.open(requests.get(path, stream=True).raw).convert('RGB')
else:
image = Image.open(path).convert('RGB')
image = image.resize(shape, resample=Image.LANCZOS)
image = np.array(image).astype(np.float16) / 255.0
image = image[None].transpose(0, 3, 1, 2)
image = torch.from_numpy(image)
return 2.*image - 1.
def load_mask_img(path, shape):
# path (str): Path to the mask image
# shape (list-like len(4)): shape of the image to match, usually latent_image.shape
mask_w_h = (shape[-1], shape[-2])
if path.startswith('http://') or path.startswith('https://'):
mask_image = Image.open(requests.get(path, stream=True).raw).convert('RGBA')
else:
mask_image = Image.open(path).convert('RGBA')
mask = mask_image.resize(mask_w_h, resample=Image.LANCZOS)
mask = mask.convert("L")
return mask
def maintain_colors(prev_img, color_match_sample, mode):
if mode == 'Match Frame 0 RGB':
return match_histograms(prev_img, color_match_sample, multichannel=True)
elif mode == 'Match Frame 0 HSV':
prev_img_hsv = cv2.cvtColor(prev_img, cv2.COLOR_RGB2HSV)
color_match_hsv = cv2.cvtColor(color_match_sample, cv2.COLOR_RGB2HSV)
matched_hsv = match_histograms(prev_img_hsv, color_match_hsv, multichannel=True)
return cv2.cvtColor(matched_hsv, cv2.COLOR_HSV2RGB)
else: # Match Frame 0 LAB
prev_img_lab = cv2.cvtColor(prev_img, cv2.COLOR_RGB2LAB)
color_match_lab = cv2.cvtColor(color_match_sample, cv2.COLOR_RGB2LAB)
matched_lab = match_histograms(prev_img_lab, color_match_lab, multichannel=True)
return cv2.cvtColor(matched_lab, cv2.COLOR_LAB2RGB)
def make_callback(sampler_name, dynamic_threshold=None, static_threshold=None, mask=None, init_latent=None, sigmas=None, sampler=None, masked_noise_modifier=1.0):
# Creates the callback function to be passed into the samplers
# The callback function is applied to the image at each step
def dynamic_thresholding_(img, threshold):
# Dynamic thresholding from Imagen paper (May 2022)
s = np.percentile(np.abs(img.cpu()), threshold, axis=tuple(range(1,img.ndim)))
s = np.max(np.append(s,1.0))
torch.clamp_(img, -1*s, s)
torch.FloatTensor.div_(img, s)
# Callback for samplers in the k-diffusion repo, called thus:
# callback({'x': x, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i], 'denoised': denoised})
def k_callback_(args_dict):
if dynamic_threshold is not None:
dynamic_thresholding_(args_dict['x'], dynamic_threshold)
if static_threshold is not None:
torch.clamp_(args_dict['x'], -1*static_threshold, static_threshold)
if mask is not None:
init_noise = init_latent + noise * args_dict['sigma']
is_masked = torch.logical_and(mask >= mask_schedule[args_dict['i']], mask != 0 )
new_img = init_noise * torch.where(is_masked,1,0) + args_dict['x'] * torch.where(is_masked,0,1)
args_dict['x'].copy_(new_img)
# Function that is called on the image (img) and step (i) at each step
def img_callback_(img, i):
# Thresholding functions
if dynamic_threshold is not None:
dynamic_thresholding_(img, dynamic_threshold)
if static_threshold is not None:
torch.clamp_(img, -1*static_threshold, static_threshold)
if mask is not None:
i_inv = len(sigmas) - i - 1
init_noise = sampler.stochastic_encode(init_latent, torch.tensor([i_inv]*batch_size).to(device), noise=noise)
is_masked = torch.logical_and(mask >= mask_schedule[i], mask != 0 )
new_img = init_noise * torch.where(is_masked,1,0) + img * torch.where(is_masked,0,1)
img.copy_(new_img)
if init_latent is not None:
noise = torch.randn_like(init_latent, device=device) * masked_noise_modifier
if sigmas is not None and len(sigmas) > 0:
mask_schedule, _ = torch.sort(sigmas/torch.max(sigmas))
elif len(sigmas) == 0:
mask = None # no mask needed if no steps (usually happens because strength==1.0)
if sampler_name in ["plms","ddim"]:
# Callback function formated for compvis latent diffusion samplers
if mask is not None:
assert sampler is not None, "Callback function for stable-diffusion samplers requires sampler variable"
batch_size = init_latent.shape[0]
callback = img_callback_
else:
# Default callback function uses k-diffusion sampler variables
callback = k_callback_
return callback
def prepare_mask(mask_file, mask_shape, mask_brightness_adjust=1.0, mask_contrast_adjust=1.0):
# path (str): Path to the mask image
# shape (list-like len(4)): shape of the image to match, usually latent_image.shape
# mask_brightness_adjust (non-negative float): amount to adjust brightness of the iamge,
# 0 is black, 1 is no adjustment, >1 is brighter
# mask_contrast_adjust (non-negative float): amount to adjust contrast of the image,
# 0 is a flat grey image, 1 is no adjustment, >1 is more contrast
mask = load_mask_img(mask_file, mask_shape)
# Mask brightness/contrast adjustments
if mask_brightness_adjust != 1:
mask = TF.adjust_brightness(mask, mask_brightness_adjust)
if mask_contrast_adjust != 1:
mask = TF.adjust_contrast(mask, mask_contrast_adjust)
# Mask image to array
mask = np.array(mask).astype(np.float32) / 255.0
mask = np.tile(mask,(4,1,1))
mask = np.expand_dims(mask,axis=0)
mask = torch.from_numpy(mask)
if args.invert_mask:
mask = ( (mask - 0.5) * -1) + 0.5
mask = np.clip(mask,0,1)
return mask
def sample_from_cv2(sample: np.ndarray) -> torch.Tensor:
sample = ((sample.astype(float) / 255.0) * 2) - 1
sample = sample[None].transpose(0, 3, 1, 2).astype(np.float16)
sample = torch.from_numpy(sample)
return sample
def sample_to_cv2(sample: torch.Tensor) -> np.ndarray:
sample_f32 = rearrange(sample.squeeze().cpu().numpy(), "c h w -> h w c").astype(np.float32)
sample_f32 = ((sample_f32 * 0.5) + 0.5).clip(0, 1)
sample_int8 = (sample_f32 * 255).astype(np.uint8)
return sample_int8
@torch.no_grad()
def transform_image_3d(prev_img_cv2, adabins_helper, midas_model, midas_transform, rot_mat, translate, anim_args):
# adapted and optimized version of transform_image_3d from Disco Diffusion https://github.com/alembics/disco-diffusion
w, h = prev_img_cv2.shape[1], prev_img_cv2.shape[0]
# predict depth with AdaBins
use_adabins = anim_args.midas_weight < 1.0 and adabins_helper is not None
if use_adabins:
print(f"Estimating depth of {w}x{h} image with AdaBins...")
MAX_ADABINS_AREA = 500000
MIN_ADABINS_AREA = 448*448
# resize image if too large or too small
img_pil = Image.fromarray(cv2.cvtColor(prev_img_cv2, cv2.COLOR_RGB2BGR))
image_pil_area = w*h
resized = True
if image_pil_area > MAX_ADABINS_AREA:
scale = math.sqrt(MAX_ADABINS_AREA) / math.sqrt(image_pil_area)
depth_input = img_pil.resize((int(w*scale), int(h*scale)), Image.LANCZOS) # LANCZOS is good for downsampling
print(f" resized to {depth_input.width}x{depth_input.height}")
elif image_pil_area < MIN_ADABINS_AREA:
scale = math.sqrt(MIN_ADABINS_AREA) / math.sqrt(image_pil_area)
depth_input = img_pil.resize((int(w*scale), int(h*scale)), Image.BICUBIC)
print(f" resized to {depth_input.width}x{depth_input.height}")
else:
depth_input = img_pil
resized = False
# predict depth and resize back to original dimensions
try:
_, adabins_depth = adabins_helper.predict_pil(depth_input)
if resized:
adabins_depth = torchvision.transforms.functional.resize(
torch.from_numpy(adabins_depth),
torch.Size([h, w]),
interpolation=torchvision.transforms.functional.InterpolationMode.BICUBIC
)
adabins_depth = adabins_depth.squeeze()
except:
print(f" exception encountered, falling back to pure MiDaS")
use_adabins = False
torch.cuda.empty_cache()
if midas_model is not None:
# convert image from 0->255 uint8 to 0->1 float for feeding to MiDaS
img_midas = prev_img_cv2.astype(np.float32) / 255.0
img_midas_input = midas_transform({"image": img_midas})["image"]
# MiDaS depth estimation implementation
print(f"Estimating depth of {w}x{h} image with MiDaS...")
sample = torch.from_numpy(img_midas_input).float().to(device).unsqueeze(0)
if device == torch.device("cuda"):
sample = sample.to(memory_format=torch.channels_last)
sample = sample.half()
midas_depth = midas_model.forward(sample)
midas_depth = torch.nn.functional.interpolate(
midas_depth.unsqueeze(1),
size=img_midas.shape[:2],
mode="bicubic",
align_corners=False,
).squeeze()
midas_depth = midas_depth.cpu().numpy()
torch.cuda.empty_cache()
# MiDaS makes the near values greater, and the far values lesser. Let's reverse that and try to align with AdaBins a bit better.
midas_depth = np.subtract(50.0, midas_depth)
midas_depth = midas_depth / 19.0
# blend between MiDaS and AdaBins predictions
if use_adabins:
depth_map = midas_depth*anim_args.midas_weight + adabins_depth*(1.0-anim_args.midas_weight)
else:
depth_map = midas_depth
depth_map = np.expand_dims(depth_map, axis=0)
depth_tensor = torch.from_numpy(depth_map).squeeze().to(device)
else:
depth_tensor = torch.ones((h, w), device=device)
pixel_aspect = 1.0 # aspect of an individual pixel (so usually 1.0)
near, far, fov_deg = anim_args.near_plane, anim_args.far_plane, anim_args.fov
persp_cam_old = p3d.FoVPerspectiveCameras(near, far, pixel_aspect, fov=fov_deg, degrees=True, device=device)
persp_cam_new = p3d.FoVPerspectiveCameras(near, far, pixel_aspect, fov=fov_deg, degrees=True, R=rot_mat, T=torch.tensor([translate]), device=device)
# range of [-1,1] is important to torch grid_sample's padding handling
y,x = torch.meshgrid(torch.linspace(-1.,1.,h,dtype=torch.float32,device=device),torch.linspace(-1.,1.,w,dtype=torch.float32,device=device))
z = torch.as_tensor(depth_tensor, dtype=torch.float32, device=device)
xyz_old_world = torch.stack((x.flatten(), y.flatten(), z.flatten()), dim=1)
xyz_old_cam_xy = persp_cam_old.get_full_projection_transform().transform_points(xyz_old_world)[:,0:2]
xyz_new_cam_xy = persp_cam_new.get_full_projection_transform().transform_points(xyz_old_world)[:,0:2]
offset_xy = xyz_new_cam_xy - xyz_old_cam_xy
# affine_grid theta param expects a batch of 2D mats. Each is 2x3 to do rotation+translation.
identity_2d_batch = torch.tensor([[1.,0.,0.],[0.,1.,0.]], device=device).unsqueeze(0)
# coords_2d will have shape (N,H,W,2).. which is also what grid_sample needs.
coords_2d = torch.nn.functional.affine_grid(identity_2d_batch, [1,1,h,w], align_corners=False)
offset_coords_2d = coords_2d - torch.reshape(offset_xy, (h,w,2)).unsqueeze(0)
image_tensor = torchvision.transforms.functional.to_tensor(Image.fromarray(prev_img_cv2)).to(device)
new_image = torch.nn.functional.grid_sample(
image_tensor.add(1/512 - 0.0001).unsqueeze(0),
offset_coords_2d,
mode=anim_args.sampling_mode,
padding_mode=anim_args.padding_mode,
align_corners=False
)
# convert back to cv2 style numpy array 0->255 uint8
result = rearrange(
new_image.squeeze().clamp(0,1) * 255.0,
'c h w -> h w c'
).cpu().numpy().astype(np.uint8)
return result
def generate(args, return_latent=False, return_sample=False, return_c=False):
seed_everything(args.seed)
os.makedirs(args.outdir, exist_ok=True)
if args.sampler == 'plms':
sampler = PLMSSampler(model)
else:
sampler = DDIMSampler(model)
model_wrap = CompVisDenoiser(model)
batch_size = args.n_samples
prompt = args.prompt
assert prompt is not None
data = [batch_size * [prompt]]
init_latent = None
if args.init_latent is not None:
init_latent = args.init_latent
elif args.init_sample is not None:
init_latent = model.get_first_stage_encoding(model.encode_first_stage(args.init_sample))
elif args.use_init and args.init_image != None and args.init_image != '':
init_image = load_img(args.init_image, shape=(args.W, args.H)).to(device)
init_image = repeat(init_image, '1 ... -> b ...', b=batch_size)
init_latent = model.get_first_stage_encoding(model.encode_first_stage(init_image)) # move to latent space
if not args.use_init and args.strength > 0:
print("\nNo init image, but strength > 0. This may give you some strange results.\n")
# Mask functions
mask = None
if args.use_mask:
assert args.mask_file is not None, "use_mask==True: An mask image is required for a mask"
assert args.use_init, "use_mask==True: use_init is required for a mask"
assert init_latent is not None, "use_mask==True: An latent init image is required for a mask"
mask = prepare_mask(args.mask_file,
init_latent.shape,
args.mask_contrast_adjust,
args.mask_brightness_adjust)
mask = mask.to(device)
mask = repeat(mask, '1 ... -> b ...', b=batch_size)
t_enc = int((1.0-args.strength) * args.steps)
# Noise schedule for the k-diffusion samplers (used for masking)
k_sigmas = model_wrap.get_sigmas(args.steps)
k_sigmas = k_sigmas[len(k_sigmas)-t_enc-1:]
if args.sampler in ['plms','ddim']:
sampler.make_schedule(ddim_num_steps=args.steps, ddim_eta=args.ddim_eta, ddim_discretize='fill', verbose=False)
callback = make_callback(sampler_name=args.sampler,
dynamic_threshold=args.dynamic_threshold,
static_threshold=args.static_threshold,
mask=mask,
init_latent=init_latent,
sigmas=k_sigmas,
sampler=sampler)
results = []
precision_scope = autocast if args.precision == "autocast" else nullcontext
with torch.no_grad():
with precision_scope("cuda"):
with model.ema_scope():
for prompts in data:
uc = None
if args.scale != 1.0:
uc = model.get_learned_conditioning(batch_size * [""])
if isinstance(prompts, tuple):
prompts = list(prompts)
c = model.get_learned_conditioning(prompts)
if args.init_c != None:
c = args.init_c
if args.sampler in ["klms","dpm2","dpm2_ancestral","heun","euler","euler_ancestral"]:
samples = sampler_fn(
c=c,
uc=uc,
args=args,
model_wrap=model_wrap,
init_latent=init_latent,
t_enc=t_enc,
device=device,
cb=callback)
else:
# args.sampler == 'plms' or args.sampler == 'ddim':
if init_latent is not None and args.strength > 0:
z_enc = sampler.stochastic_encode(init_latent, torch.tensor([t_enc]*batch_size).to(device))
else:
z_enc = torch.randn([args.n_samples, args.C, args.H // args.f, args.W // args.f], device=device)
if args.sampler == 'ddim':
samples = sampler.decode(z_enc,
c,
t_enc,
unconditional_guidance_scale=args.scale,
unconditional_conditioning=uc,
img_callback=callback)
elif args.sampler == 'plms': # no "decode" function in plms, so use "sample"
shape = [args.C, args.H // args.f, args.W // args.f]
samples, _ = sampler.sample(S=args.steps,
conditioning=c,
batch_size=args.n_samples,
shape=shape,
verbose=False,
unconditional_guidance_scale=args.scale,
unconditional_conditioning=uc,
eta=args.ddim_eta,
x_T=z_enc,
img_callback=callback)
else:
raise Exception(f"Sampler {args.sampler} not recognised.")
if return_latent:
results.append(samples.clone())
x_samples = model.decode_first_stage(samples)
if return_sample:
results.append(x_samples.clone())
x_samples = torch.clamp((x_samples + 1.0) / 2.0, min=0.0, max=1.0)
if return_c:
results.append(c.clone())
for x_sample in x_samples:
x_sample = 255. * rearrange(x_sample.cpu().numpy(), 'c h w -> h w c')
image = Image.fromarray(x_sample.astype(np.uint8))
results.append(image)
return results
# %%
# !! {"metadata":{
# !! "cellView": "form",
# !! "id": "CIUJ7lWI4v53"
# !! }}
#@markdown **Select and Load Model**
model_config = "v1-inference.yaml" #@param ["custom","v1-inference.yaml"]
model_checkpoint = "sd-v1-4.ckpt" #@param ["custom","sd-v1-4-full-ema.ckpt","sd-v1-4.ckpt","sd-v1-3-full-ema.ckpt","sd-v1-3.ckpt","sd-v1-2-full-ema.ckpt","sd-v1-2.ckpt","sd-v1-1-full-ema.ckpt","sd-v1-1.ckpt"]
custom_config_path = "" #@param {type:"string"}
custom_checkpoint_path = "" #@param {type:"string"}
load_on_run_all = True #@param {type: 'boolean'}
half_precision = True # check
check_sha256 = True #@param {type:"boolean"}
model_map = {
"sd-v1-4-full-ema.ckpt": {'sha256': '14749efc0ae8ef0329391ad4436feb781b402f4fece4883c7ad8d10556d8a36a'},
"sd-v1-4.ckpt": {'sha256': 'fe4efff1e174c627256e44ec2991ba279b3816e364b49f9be2abc0b3ff3f8556'},
"sd-v1-3-full-ema.ckpt": {'sha256': '54632c6e8a36eecae65e36cb0595fab314e1a1545a65209f24fde221a8d4b2ca'},
"sd-v1-3.ckpt": {'sha256': '2cff93af4dcc07c3e03110205988ff98481e86539c51a8098d4f2236e41f7f2f'},
"sd-v1-2-full-ema.ckpt": {'sha256': 'bc5086a904d7b9d13d2a7bccf38f089824755be7261c7399d92e555e1e9ac69a'},
"sd-v1-2.ckpt": {'sha256': '3b87d30facd5bafca1cbed71cfb86648aad75d1c264663c0cc78c7aea8daec0d'},
"sd-v1-1-full-ema.ckpt": {'sha256': 'efdeb5dc418a025d9a8cc0a8617e106c69044bc2925abecc8a254b2910d69829'},
"sd-v1-1.ckpt": {'sha256': '86cd1d3ccb044d7ba8db743d717c9bac603c4043508ad2571383f954390f3cea'}
}
# config path
ckpt_config_path = custom_config_path if model_config == "custom" else os.path.join(models_path, model_config)
if os.path.exists(ckpt_config_path):
print(f"{ckpt_config_path} exists")
else:
ckpt_config_path = "./stable-diffusion/configs/stable-diffusion/v1-inference.yaml"
print(f"Using config: {ckpt_config_path}")
# checkpoint path or download
ckpt_path = custom_checkpoint_path if model_checkpoint == "custom" else os.path.join(models_path, model_checkpoint)
ckpt_valid = True
if os.path.exists(ckpt_path):
print(f"{ckpt_path} exists")
else:
print(f"Please download model checkpoint and place in {os.path.join(models_path, model_checkpoint)}")
ckpt_valid = False
if check_sha256 and model_checkpoint != "custom" and ckpt_valid:
import hashlib
print("\n...checking sha256")
with open(ckpt_path, "rb") as f:
bytes = f.read()
hash = hashlib.sha256(bytes).hexdigest()
del bytes
if model_map[model_checkpoint]["sha256"] == hash:
print("hash is correct\n")
else:
print("hash in not correct\n")
ckpt_valid = False
if ckpt_valid:
print(f"Using ckpt: {ckpt_path}")
def load_model_from_config(config, ckpt, verbose=False, device='cuda', half_precision=True):
map_location = "cuda" #@param ["cpu", "cuda"]
print(f"Loading model from {ckpt}")
pl_sd = torch.load(ckpt, map_location=map_location)
if "global_step" in pl_sd:
print(f"Global Step: {pl_sd['global_step']}")
sd = pl_sd["state_dict"]
model = instantiate_from_config(config.model)
m, u = model.load_state_dict(sd, strict=False)
if len(m) > 0 and verbose:
print("missing keys:")
print(m)
if len(u) > 0 and verbose:
print("unexpected keys:")
print(u)
if half_precision:
model = model.half().to(device)
else:
model = model.to(device)
model.eval()
return model
if load_on_run_all and ckpt_valid:
local_config = OmegaConf.load(f"{ckpt_config_path}")
model = load_model_from_config(local_config, f"{ckpt_path}",half_precision=half_precision)
device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
model = model.to(device)
# %%
# !! {"metadata":{
# !! "id": "ov3r4RD1tzsT"
# !! }}
"""
# Settings
"""
# %%
# !! {"metadata":{
# !! "id": "0j7rgxvLvfay"
# !! }}
"""
### Animation Settings
"""
# %%
# !! {"metadata":{
# !! "cellView": "form",
# !! "id": "8HJN2TE3vh-J"
# !! }}
def DeforumAnimArgs():
#@markdown ####**Animation:**
animation_mode = 'None' #@param ['None', '2D', '3D', 'Video Input', 'Interpolation'] {type:'string'}
max_frames = 1000 #@param {type:"number"}
border = 'wrap' #@param ['wrap', 'replicate'] {type:'string'}
#@markdown ####**Motion Parameters:**
angle = "0:(0)"#@param {type:"string"}
zoom = "0:(1.04)"#@param {type:"string"}
translation_x = "0:(0)"#@param {type:"string"}
translation_y = "0:(0)"#@param {type:"string"}
translation_z = "0:(10)"#@param {type:"string"}
rotation_3d_x = "0:(0)"#@param {type:"string"}
rotation_3d_y = "0:(0)"#@param {type:"string"}
rotation_3d_z = "0:(0)"#@param {type:"string"}
noise_schedule = "0: (0.02)"#@param {type:"string"}
strength_schedule = "0: (0.65)"#@param {type:"string"}
contrast_schedule = "0: (1.0)"#@param {type:"string"}
#@markdown ####**Coherence:**
color_coherence = 'Match Frame 0 LAB' #@param ['None', 'Match Frame 0 HSV', 'Match Frame 0 LAB', 'Match Frame 0 RGB'] {type:'string'}
#@markdown #### Depth Warping
use_depth_warping = True #@param {type:"boolean"}
midas_weight = 0.3#@param {type:"number"}
near_plane = 200
far_plane = 10000
fov = 40#@param {type:"number"}
padding_mode = 'border'#@param ['border', 'reflection', 'zeros'] {type:'string'}
sampling_mode = 'bicubic'#@param ['bicubic', 'bilinear', 'nearest'] {type:'string'}
#@markdown ####**Video Input:**
video_init_path ='/content/video_in.mp4'#@param {type:"string"}
extract_nth_frame = 1#@param {type:"number"}
#@markdown ####**Interpolation:**
interpolate_key_frames = False #@param {type:"boolean"}
interpolate_x_frames = 4 #@param {type:"number"}
#@markdown ####**Resume Animation:**
resume_from_timestring = False #@param {type:"boolean"}
resume_timestring = "20220829210106" #@param {type:"string"}
return locals()
class DeformAnimKeys():
def __init__(self, anim_args):
self.angle_series = get_inbetweens(parse_key_frames(anim_args.angle))
self.zoom_series = get_inbetweens(parse_key_frames(anim_args.zoom))
self.translation_x_series = get_inbetweens(parse_key_frames(anim_args.translation_x))
self.translation_y_series = get_inbetweens(parse_key_frames(anim_args.translation_y))
self.translation_z_series = get_inbetweens(parse_key_frames(anim_args.translation_z))
self.rotation_3d_x_series = get_inbetweens(parse_key_frames(anim_args.rotation_3d_x))
self.rotation_3d_y_series = get_inbetweens(parse_key_frames(anim_args.rotation_3d_y))
self.rotation_3d_z_series = get_inbetweens(parse_key_frames(anim_args.rotation_3d_z))
self.noise_schedule_series = get_inbetweens(parse_key_frames(anim_args.noise_schedule))
self.strength_schedule_series = get_inbetweens(parse_key_frames(anim_args.strength_schedule))
self.contrast_schedule_series = get_inbetweens(parse_key_frames(anim_args.contrast_schedule))
def get_inbetweens(key_frames, integer=False, interp_method='Linear'):
key_frame_series = pd.Series([np.nan for a in range(anim_args.max_frames)])
for i, value in key_frames.items():
key_frame_series[i] = value
key_frame_series = key_frame_series.astype(float)
if interp_method == 'Cubic' and len(key_frames.items()) <= 3:
interp_method = 'Quadratic'
if interp_method == 'Quadratic' and len(key_frames.items()) <= 2:
interp_method = 'Linear'
key_frame_series[0] = key_frame_series[key_frame_series.first_valid_index()]
key_frame_series[anim_args.max_frames-1] = key_frame_series[key_frame_series.last_valid_index()]
key_frame_series = key_frame_series.interpolate(method=interp_method.lower(),limit_direction='both')
if integer:
return key_frame_series.astype(int)
return key_frame_series
def parse_key_frames(string, prompt_parser=None):
import re
pattern = r'((?P<frame>[0-9]+):[\s]*[\(](?P<param>[\S\s]*?)[\)])'
frames = dict()
for match_object in re.finditer(pattern, string):
frame = int(match_object.groupdict()['frame'])
param = match_object.groupdict()['param']
if prompt_parser:
frames[frame] = prompt_parser(param)
else:
frames[frame] = param
if frames == {} and len(string) != 0:
raise RuntimeError('Key Frame string not correctly formatted')
return frames
# %%
# !! {"metadata":{
# !! "id": "63UOJvU3xdPS"
# !! }}
"""
### Prompts
`animation_mode: None` batches on list of *prompts*. `animation_mode: 2D` uses *animation_prompts* key frame sequence
"""
# %%
# !! {"metadata":{
# !! "id": "2ujwkGZTcGev"
# !! }}
prompts = [
"a beautiful forest by Asher Brown Durand, trending on Artstation", #the first prompt I want
"a beautiful portrait of a woman by Artgerm, trending on Artstation", #the second prompt I want
#"the third prompt I don't want it I commented it with an",
]
animation_prompts = {
0: "a beautiful apple, trending on Artstation",
20: "a beautiful banana, trending on Artstation",
30: "a beautiful coconut, trending on Artstation",
40: "a beautiful durian, trending on Artstation",
}
# %%
# !! {"metadata":{
# !! "id": "s8RAo2zI-vQm"
# !! }}
"""
# Run
"""
# %%
# !! {"metadata":{
# !! "id": "qH74gBWDd2oq",
# !! "cellView": "form"
# !! }}
def DeforumArgs():
#@markdown **Image Settings**
W = 512 #@param
H = 512 #@param
W, H = map(lambda x: x - x % 64, (W, H)) # resize to integer multiple of 64
#@markdown **Sampling Settings**
seed = -1 #@param
sampler = 'klms' #@param ["klms","dpm2","dpm2_ancestral","heun","euler","euler_ancestral","plms", "ddim"]
steps = 50 #@param
scale = 7 #@param
ddim_eta = 0.0 #@param
dynamic_threshold = None
static_threshold = None
#@markdown **Save & Display Settings**
save_samples = True #@param {type:"boolean"}
save_settings = True #@param {type:"boolean"}
display_samples = True #@param {type:"boolean"}
#@markdown **Batch Settings**
n_batch = 1 #@param
batch_name = "StableFun" #@param {type:"string"}
filename_format = "{timestring}_{index}_{prompt}.png" #@param ["{timestring}_{index}_{seed}.png","{timestring}_{index}_{prompt}.png"]
seed_behavior = "iter" #@param ["iter","fixed","random"]
make_grid = False #@param {type:"boolean"}
grid_rows = 2 #@param
outdir = get_output_folder(output_path, batch_name)
#@markdown **Init Settings**
use_init = False #@param {type:"boolean"}
strength = 0.0 #@param {type:"number"}
init_image = "https://cdn.pixabay.com/photo/2022/07/30/13/10/green-longhorn-beetle-7353749_1280.jpg" #@param {type:"string"}
# Whiter areas of the mask are areas that change more
use_mask = False #@param {type:"boolean"}
mask_file = "https://www.filterforge.com/wiki/images/archive/b/b7/20080927223728%21Polygonal_gradient_thumb.jpg" #@param {type:"string"}
invert_mask = False #@param {type:"boolean"}
# Adjust mask image, 1.0 is no adjustment. Should be positive numbers.
mask_brightness_adjust = 1.0 #@param {type:"number"}
mask_contrast_adjust = 1.0 #@param {type:"number"}
n_samples = 1 # doesnt do anything
precision = 'autocast'
C = 4
f = 8
prompt = ""
timestring = ""
init_latent = None
init_sample = None
init_c = None
return locals()
def next_seed(args):
if args.seed_behavior == 'iter':
args.seed += 1
elif args.seed_behavior == 'fixed':
pass # always keep seed the same
else:
args.seed = random.randint(0, 2**32)
return args.seed
def render_image_batch(args):
args.prompts = {k: f"{v:05d}" for v, k in enumerate(prompts)}
# create output folder for the batch
os.makedirs(args.outdir, exist_ok=True)
if args.save_settings or args.save_samples:
print(f"Saving to {os.path.join(args.outdir, args.timestring)}_*")
# save settings for the batch
if args.save_settings:
filename = os.path.join(args.outdir, f"{args.timestring}_settings.txt")
with open(filename, "w+", encoding="utf-8") as f:
json.dump(dict(args.__dict__), f, ensure_ascii=False, indent=4)
index = 0
# function for init image batching
init_array = []
if args.use_init:
if args.init_image == "":
raise FileNotFoundError("No path was given for init_image")
if args.init_image.startswith('http://') or args.init_image.startswith('https://'):
init_array.append(args.init_image)
elif not os.path.isfile(args.init_image):
if args.init_image[-1] != "/": # avoids path error by adding / to end if not there
args.init_image += "/"
for image in sorted(os.listdir(args.init_image)): # iterates dir and appends images to init_array
if image.split(".")[-1] in ("png", "jpg", "jpeg"):
init_array.append(args.init_image + image)
else:
init_array.append(args.init_image)
else:
init_array = [""]
# when doing large batches don't flood browser with images
clear_between_batches = args.n_batch >= 32
for iprompt, prompt in enumerate(prompts):
args.prompt = prompt
print(f"Prompt {iprompt+1} of {len(prompts)}")
print(f"{args.prompt}")
all_images = []
for batch_index in range(args.n_batch):
if clear_between_batches and batch_index % 32 == 0:
display.clear_output(wait=True)
print(f"Batch {batch_index+1} of {args.n_batch}")
for image in init_array: # iterates the init images
args.init_image = image
results = generate(args)
for image in results:
if args.make_grid:
all_images.append(T.functional.pil_to_tensor(image))