Bernard Maltais commited on
Commit
2002a4d
1 Parent(s): c7bc6d2

1st commit

Browse files
README.md ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: unknown
3
+ ---
4
+ # Silvery Trait finetuned style Model
5
+
6
+ Produced from publicly available pictures in landscape, portrait and square format.
7
+
8
+ Using words found in `prompt_words.md` within your prompt will produce better results. Other words can be used also but will tend to produce "weaker" results. Combining the use of the Aesthetic Gradient file provided in the `easthetic_embeddings` folder can greatly enhance the results.
9
+
10
+ ## Model info
11
+
12
+ The models included was trained on "multi-resolution" images.
13
+
14
+ ## Using the model
15
+
16
+ * common subject prompt tokens: `<wathever>, by asd artstyle`
17
+
18
+ ## Example prompts
19
+
20
+ `a sheep, symmetry, by asd artstyle`:
21
+
22
+ * without easthetic_embeddings
23
+
24
+ <img src="https://huggingface.co/cyburn/silvery_trait/resolve/main/1.png" alt="Picture." width="500"/>
25
+
26
+ * with easthetic_embeddings
27
+
28
+ <img src="https://huggingface.co/cyburn/silvery_trait/resolve/main/2.png" alt="Picture." width="500"/>
29
+
30
+ `crow, skull, symmetry, flower, feather, circle, by asd artstyle`
31
+
32
+ * without easthetic_embeddings
33
+
34
+ <img src="https://huggingface.co/cyburn/silvery_trait/resolve/main/3.png" alt="Picture." width="500"/>
35
+
36
+ * with easthetic_embeddings
37
+
38
+ <img src="https://huggingface.co/cyburn/silvery_trait/resolve/main/4.png" alt="Picture." width="500"/>
aesthetic_embeddings/silvery_trait.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:070550d3462c89b59ca0a93fb387a9b0bc8dd7ba157e87416957acd145e6879c
3
+ size 3819
kohya_train_db_v9.ps1 ADDED
@@ -0,0 +1,52 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # This powershell script will create a model using the fine tuning dreambooth method. It will require landscape,
2
+ # portrait and square images.
3
+ #
4
+ # Adjust the script to your own needs
5
+
6
+ # variable values
7
+ $pretrained_model_name_or_path = "D:\models\v1-5-pruned-mse-vae.ckpt"
8
+ $data_dir = "D:\dreambooth\silvery_trait\raw_data\all-images-v3"
9
+ $train_dir = "D:\dreambooth\silvery_trait\"
10
+ $resolution = "768,576"
11
+ $logging_dir = "D:\dreambooth\silvery_trait\training_logs"
12
+
13
+ $image_num = Get-ChildItem $data_dir -Recurse -File -Include *.png, *.jpg, *.webp | Measure-Object | %{$_.Count}
14
+
15
+ Write-Output "image_num: $image_num"
16
+
17
+ $learning_rate = 1e-6
18
+ $dataset_repeats = 40
19
+ $train_batch_size = 6
20
+ $epoch = 4
21
+ $save_every_n_epochs=1
22
+ $mixed_precision="fp16"
23
+ $num_cpu_threads_per_process=6
24
+
25
+ # You should not have to change values past this point
26
+
27
+ $output_dir = $train_dir + "\finetuned_model"
28
+ $repeats = $image_num * $dataset_repeats
29
+ $mts = [Math]::Ceiling($repeats / $train_batch_size * $epoch)
30
+
31
+ Write-Output "Repeats: $repeats"
32
+
33
+ .\venv\Scripts\activate
34
+
35
+ accelerate launch --num_cpu_threads_per_process $num_cpu_threads_per_process train_db_fixed.py `
36
+ --pretrained_model_name_or_path=$pretrained_model_name_or_path `
37
+ --train_data_dir=$data_dir `
38
+ --output_dir=$output_dir `
39
+ --resolution=$resolution `
40
+ --train_batch_size=$train_batch_size `
41
+ --learning_rate=$learning_rate `
42
+ --max_train_steps=$mts `
43
+ --use_8bit_adam `
44
+ --xformers `
45
+ --mixed_precision=$mixed_precision `
46
+ --cache_latents `
47
+ --save_every_n_epochs=$save_every_n_epochs `
48
+ --fine_tuning `
49
+ --enable_bucket `
50
+ --dataset_repeats=$dataset_repeats `
51
+ --logging_dir=$logging_dir `
52
+ --save_precision="fp16"
prompt_words.md ADDED
@@ -0,0 +1,443 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Prompt Words
2
+
3
+ List of prompt words known to the model
4
+
5
+ ## Sorted by frequency
6
+
7
+ ```txt
8
+ Count Name
9
+ ----- ----
10
+ 87 a
11
+ 85 symmetry
12
+ 17 leaves
13
+ 15 two
14
+ 13 tree
15
+ 13 moon
16
+ 13 circle
17
+ 13 with
18
+ 13 an
19
+ 12 cloud
20
+ 12 and
21
+ 11 landscape
22
+ 10 coin
23
+ 10 water
24
+ 10 bird
25
+ 9 antler
26
+ 9 holding
27
+ 9 on
28
+ 9 the
29
+ 7 mountain
30
+ 7 sun
31
+ 6 art
32
+ 6 in
33
+ 6 deer
34
+ 6 concept
35
+ 6 wings
36
+ 5 stick
37
+ 5 bug
38
+ 5 horn
39
+ 5 feather
40
+ 5 star
41
+ 4 rabbit
42
+ 4 three
43
+ 4 crescent
44
+ 4 cosmic
45
+ 4 wolf
46
+ 4 sitting
47
+ 3 salamander
48
+ 3 dog
49
+ 3 grass
50
+ 3 ball
51
+ 3 branch
52
+ 3 fish
53
+ 3 cat
54
+ 3 earth
55
+ 3 pyramid
56
+ 3 yang
57
+ 3 fox
58
+ 3 globe
59
+ 3 yin
60
+ 3 lion
61
+ 3 planet
62
+ 3 wearing
63
+ 3 over
64
+ 3 monkey
65
+ 3 person
66
+ 3 spiral
67
+ 3 river
68
+ 3 sheep
69
+ 3 skull
70
+ 2 bamboo
71
+ 2 bear
72
+ 2 raccoon
73
+ 2 crown
74
+ 2 otter
75
+ 2 bull
76
+ 2 wall
77
+ 2 under
78
+ 2 flying
79
+ 2 lantern
80
+ 2 owl
81
+ 2 standing
82
+ 2 elephant
83
+ 2 paved
84
+ 2 flag
85
+ 2 hanging
86
+ 2 donkey
87
+ 2 waterfall
88
+ 2 sunflower
89
+ 2 squid
90
+ 2 cup
91
+ 2 armor
92
+ 2 from
93
+ 2 meerkat
94
+ 2 city
95
+ 1 triangle
96
+ 1 sloth
97
+ 1 snail
98
+ 1 glass
99
+ 1 panda
100
+ 1 tiger
101
+ 1 fountain
102
+ 1 animal
103
+ 1 algea
104
+ 1 ray
105
+ 1 lamb
106
+ 1 orange
107
+ 1 perched
108
+ 1 pumpkin
109
+ 1 manta
110
+ 1 axolotl
111
+ 1 zapus
112
+ 1 pentastar
113
+ 1 phone
114
+ 1 yak
115
+ 1 brick
116
+ 1 candle
117
+ 1 devilwingsand
118
+ 1 giraffe
119
+ 1 zebra
120
+ 1 four
121
+ 1 bee
122
+ 1 season
123
+ 1 building
124
+ 1 five
125
+ 1 starfish
126
+ 1 warthog
127
+ 1 turtle
128
+ 1 shaped
129
+ 1 wool
130
+ 1 unicorn
131
+ 1 frog
132
+ 1 land
133
+ 1 tower
134
+ 1 middle
135
+ 1 cathedral
136
+ 1 wasp
137
+ 1 butterfly
138
+ 1 head
139
+ 1 watermellon
140
+ 1 his
141
+ 1 bag
142
+ 1 leave
143
+ 1 corn
144
+ 1 crab
145
+ 1 hermit
146
+ 1 people
147
+ 1 space
148
+ 1 floating
149
+ 1 dragon
150
+ 1 caterpillan
151
+ 1 key
152
+ 1 bowl
153
+ 1 fruit
154
+ 1 burning
155
+ 1 eagle
156
+ 1 disco
157
+ 1 scorpio
158
+ 1 awk
159
+ 1 monocle
160
+ 1 duck
161
+ 1 flower
162
+ 1 crow
163
+ 1 door
164
+ 1 house
165
+ 1 woman
166
+ 1 hippo
167
+ 1 up
168
+ 1 fossa
169
+ 1 prairie
170
+ 1 gorilla
171
+ 1 toucan
172
+ 1 card
173
+ 1 beaver
174
+ 1 wood
175
+ 1 crane
176
+ 1 bridge
177
+ 1 stump
178
+ 1 alpaca
179
+ 1 hybrid
180
+ 1 badger
181
+ 1 sky
182
+ 1 comet
183
+ 1 desert
184
+ 1 deamon
185
+ 1 bunny
186
+ 1 bubble
187
+ 1 circul
188
+ 1 python
189
+ 1 snow
190
+ 1 books
191
+ 1 quetzal
192
+ 1 tattoo
193
+ 1 red
194
+ 1 whale
195
+ 1 of
196
+ 1 tiny
197
+ 1 parrot
198
+ 1 drinking
199
+ 1 climbing
200
+ 1 six
201
+ 1 reflexion
202
+ 1 skunk
203
+ 1 mouse
204
+ 1 beetle
205
+ 1 seahorse
206
+ 1 jellyfish
207
+ 1 hat
208
+ 1 mammoth
209
+ 1 jackalope
210
+ 1 hedgehog
211
+ 1 smoking
212
+ 1 table
213
+ 1 at
214
+ 1 tusk
215
+ 1 windmill
216
+ 1 floor
217
+ 1 berret
218
+ 1 lynx
219
+ 1 koala
220
+ 1 heron
221
+ 1 peacock
222
+ 1 leaf
223
+ ```
224
+
225
+ ## Sorted alphabetically
226
+
227
+ ```txt
228
+ Count Name
229
+ ----- ----
230
+ 87 a
231
+ 1 algea
232
+ 1 alpaca
233
+ 13 an
234
+ 12 and
235
+ 1 animal
236
+ 9 antler
237
+ 2 armor
238
+ 6 art
239
+ 1 at
240
+ 1 awk
241
+ 1 axolotl
242
+ 1 badger
243
+ 1 bag
244
+ 3 ball
245
+ 2 bamboo
246
+ 2 bear
247
+ 1 beaver
248
+ 1 bee
249
+ 1 beetle
250
+ 1 berret
251
+ 10 bird
252
+ 1 books
253
+ 1 bowl
254
+ 3 branch
255
+ 1 brick
256
+ 1 bridge
257
+ 1 bubble
258
+ 5 bug
259
+ 1 building
260
+ 2 bull
261
+ 1 bunny
262
+ 1 burning
263
+ 1 butterfly
264
+ 1 candle
265
+ 1 card
266
+ 3 cat
267
+ 1 caterpillan
268
+ 1 cathedral
269
+ 13 circle
270
+ 1 circul
271
+ 2 city
272
+ 1 climbing
273
+ 12 cloud
274
+ 10 coin
275
+ 1 comet
276
+ 6 concept
277
+ 1 corn
278
+ 4 cosmic
279
+ 1 crab
280
+ 1 crane
281
+ 4 crescent
282
+ 1 crow
283
+ 2 crown
284
+ 2 cup
285
+ 1 deamon
286
+ 6 deer
287
+ 1 desert
288
+ 1 devilwingsand
289
+ 1 disco
290
+ 3 dog
291
+ 2 donkey
292
+ 1 door
293
+ 1 dragon
294
+ 1 drinking
295
+ 1 duck
296
+ 1 eagle
297
+ 3 earth
298
+ 2 elephant
299
+ 5 feather
300
+ 3 fish
301
+ 1 five
302
+ 2 flag
303
+ 1 floating
304
+ 1 floor
305
+ 1 flower
306
+ 2 flying
307
+ 1 fossa
308
+ 1 fountain
309
+ 1 four
310
+ 3 fox
311
+ 1 frog
312
+ 2 from
313
+ 1 fruit
314
+ 1 giraffe
315
+ 1 glass
316
+ 3 globe
317
+ 1 gorilla
318
+ 3 grass
319
+ 2 hanging
320
+ 1 hat
321
+ 1 head
322
+ 1 hedgehog
323
+ 1 hermit
324
+ 1 heron
325
+ 1 hippo
326
+ 1 his
327
+ 9 holding
328
+ 5 horn
329
+ 1 house
330
+ 1 hybrid
331
+ 6 in
332
+ 1 jackalope
333
+ 1 jellyfish
334
+ 1 key
335
+ 1 koala
336
+ 1 lamb
337
+ 1 land
338
+ 11 landscape
339
+ 2 lantern
340
+ 1 leaf
341
+ 1 leave
342
+ 17 leaves
343
+ 3 lion
344
+ 1 lynx
345
+ 1 mammoth
346
+ 1 manta
347
+ 2 meerkat
348
+ 1 middle
349
+ 3 monkey
350
+ 1 monocle
351
+ 13 moon
352
+ 7 mountain
353
+ 1 mouse
354
+ 1 of
355
+ 9 on
356
+ 1 orange
357
+ 2 otter
358
+ 3 over
359
+ 2 owl
360
+ 1 panda
361
+ 1 parrot
362
+ 2 paved
363
+ 1 peacock
364
+ 1 pentastar
365
+ 1 people
366
+ 1 perched
367
+ 3 person
368
+ 1 phone
369
+ 3 planet
370
+ 1 prairie
371
+ 1 pumpkin
372
+ 3 pyramid
373
+ 1 python
374
+ 1 quetzal
375
+ 4 rabbit
376
+ 2 raccoon
377
+ 1 ray
378
+ 1 red
379
+ 1 reflexion
380
+ 3 river
381
+ 3 salamander
382
+ 1 scorpio
383
+ 1 seahorse
384
+ 1 season
385
+ 1 shaped
386
+ 3 sheep
387
+ 4 sitting
388
+ 1 six
389
+ 3 skull
390
+ 1 skunk
391
+ 1 sky
392
+ 1 sloth
393
+ 1 smoking
394
+ 1 snail
395
+ 1 snow
396
+ 1 space
397
+ 3 spiral
398
+ 2 squid
399
+ 2 standing
400
+ 5 star
401
+ 1 starfish
402
+ 5 stick
403
+ 1 stump
404
+ 7 sun
405
+ 2 sunflower
406
+ 85 symmetry
407
+ 1 table
408
+ 1 tattoo
409
+ 9 the
410
+ 4 three
411
+ 1 tiger
412
+ 1 tiny
413
+ 1 toucan
414
+ 1 tower
415
+ 13 tree
416
+ 1 triangle
417
+ 1 turtle
418
+ 1 tusk
419
+ 15 two
420
+ 2 under
421
+ 1 unicorn
422
+ 1 up
423
+ 2 wall
424
+ 1 warthog
425
+ 1 wasp
426
+ 10 water
427
+ 2 waterfall
428
+ 1 watermellon
429
+ 3 wearing
430
+ 1 whale
431
+ 1 windmill
432
+ 6 wings
433
+ 13 with
434
+ 4 wolf
435
+ 1 woman
436
+ 1 wood
437
+ 1 wool
438
+ 1 yak
439
+ 3 yang
440
+ 3 yin
441
+ 1 zapus
442
+ 1 zebra
443
+ ```
silvery_trait_v3.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b0308cf699caf50120d43748134b9cfbf9ac0bc40fa8c29759a52e5ae25f253
3
+ size 2132856686