-
Notifications
You must be signed in to change notification settings - Fork 4
/
train_en_model.py
717 lines (630 loc) · 26.7 KB
/
train_en_model.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
import pandas as pd
from collections import namedtuple
import os
os.environ['KMP_DUPLICATE_LIB_OK']='True'
import argparse
import logging
import math
import os
import random
from pathlib import Path
from typing import Iterable, Optional
import numpy as np
import torch
import torch.nn.functional as F
import torch.utils.checkpoint
from accelerate import Accelerator
from accelerate.logging import get_logger
from accelerate.utils import set_seed
from datasets import load_dataset
from diffusers import AutoencoderKL, DDPMScheduler, PNDMScheduler, StableDiffusionPipeline, UNet2DConditionModel
from diffusers.optimization import get_scheduler
from diffusers.pipelines.stable_diffusion import StableDiffusionSafetyChecker
from huggingface_hub import HfFolder, Repository, whoami
from torchvision import transforms
from tqdm.auto import tqdm
from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer
logger = get_logger(__name__)
def parse_args():
parser = argparse.ArgumentParser(description="Simple example of a training script.")
parser.add_argument(
"--pretrained_model_name_or_path",
type=str,
default=None,
required=True,
help="Path to pretrained model or model identifier from huggingface.co/models.",
)
parser.add_argument(
"--dataset_name",
type=str,
default=None,
help=(
"The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
" dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
" or to a folder containing files that 🤗 Datasets can understand."
),
)
parser.add_argument(
"--dataset_config_name",
type=str,
default=None,
help="The config of the Dataset, leave as None if there's only one config.",
)
parser.add_argument(
"--train_data_dir",
type=str,
default=None,
help=(
"A folder containing the training data. Folder contents must follow the structure described in"
" https://huggingface.co/docs/datasets/image_dataset#imagefolder. In particular, a `metadata.jsonl` file"
" must exist to provide the captions for the images. Ignored if `dataset_name` is specified."
),
)
parser.add_argument(
"--image_column", type=str, default="image", help="The column of the dataset containing an image."
)
parser.add_argument(
"--caption_column",
type=str,
default="text",
help="The column of the dataset containing a caption or a list of captions.",
)
parser.add_argument(
"--max_train_samples",
type=int,
default=None,
help=(
"For debugging purposes or quicker training, truncate the number of training examples to this "
"value if set."
),
)
parser.add_argument(
"--output_dir",
type=str,
default="sd-model-finetuned",
help="The output directory where the model predictions and checkpoints will be written.",
)
parser.add_argument(
"--cache_dir",
type=str,
default=None,
help="The directory where the downloaded models and datasets will be stored.",
)
parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
parser.add_argument(
"--resolution",
type=int,
default=512,
help=(
"The resolution for input images, all the images in the train/validation dataset will be resized to this"
" resolution"
),
)
parser.add_argument(
"--center_crop",
action="store_true",
help="Whether to center crop images before resizing to resolution (if not set, random crop will be used)",
)
parser.add_argument(
"--random_flip",
action="store_true",
help="whether to randomly flip images horizontally",
)
parser.add_argument(
"--train_batch_size", type=int, default=16, help="Batch size (per device) for the training dataloader."
)
parser.add_argument("--num_train_epochs", type=int, default=100)
parser.add_argument(
"--max_train_steps",
type=int,
default=None,
help="Total number of training steps to perform. If provided, overrides num_train_epochs.",
)
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.",
)
parser.add_argument(
"--gradient_checkpointing",
action="store_true",
help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
)
parser.add_argument(
"--learning_rate",
type=float,
default=1e-4,
help="Initial learning rate (after the potential warmup period) to use.",
)
parser.add_argument(
"--scale_lr",
action="store_true",
default=False,
help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
)
parser.add_argument(
"--lr_scheduler",
type=str,
default="constant",
help=(
'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
' "constant", "constant_with_warmup"]'
),
)
parser.add_argument(
"--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
)
parser.add_argument(
"--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
)
parser.add_argument("--use_ema", action="store_true", help="Whether to use EMA model.")
parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
parser.add_argument(
"--hub_model_id",
type=str,
default=None,
help="The name of the repository to keep in sync with the local `output_dir`.",
)
parser.add_argument(
"--logging_dir",
type=str,
default="logs",
help=(
"[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
" *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
),
)
parser.add_argument(
"--mixed_precision",
type=str,
default="no",
choices=["no", "fp16", "bf16"],
help=(
"Whether to use mixed precision. Choose"
"between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >= 1.10."
"and an Nvidia Ampere GPU."
),
)
parser.add_argument(
"--report_to",
type=str,
default="tensorboard",
help=(
'The integration to report the results and logs to. Supported platforms are `"tensorboard"`,'
' `"wandb"` and `"comet_ml"`. Use `"all"` (default) to report to all integrations.'
"Only applicable when `--with_tracking` is passed."
),
)
parser.add_argument("--local_rank", type=int, default=-1, help="For distributed training: local_rank")
#args = parser.parse_args()
return parser
env_local_rank = int(os.environ.get("LOCAL_RANK", -1))
if env_local_rank != -1 and env_local_rank != args.local_rank:
args.local_rank = env_local_rank
# Sanity checks
if args.dataset_name is None and args.train_data_dir is None:
raise ValueError("Need either a dataset name or a training folder.")
return args
def parse_parser_add_arg(parser, as_named_tuple = False):
args_df = pd.DataFrame(
pd.Series(parser.__dict__["_actions"]).map(
lambda x:x.__dict__
).values.tolist())
args_df = args_df.explode("option_strings")
args_df["option_strings"] = args_df["option_strings"].map(
lambda x: x[2:] if x.startswith("--") else x
).map(
lambda x: x[1:] if x.startswith("-") else x
)
args_df = args_df[["option_strings", "default"]]
args = dict(args_df.values.tolist())
if as_named_tuple:
args_parser_namedtuple = namedtuple("args_config", args)
return args_parser_namedtuple(**args)
return args_df
def transform_named_tuple_to_dict(N_tuple):
return dict(map(
lambda x: (x, getattr(N_tuple, x))
,filter(lambda x: not x.startswith("_") ,dir(N_tuple))
))
def transform_dict_to_named_tuple(dict_, name = "NamedTuple"):
args_parser_namedtuple = namedtuple(name, dict_)
return args_parser_namedtuple(**dict_)
def setattr_gen_option_cls(src_obj):
assert isinstance(src_obj, tuple) or isinstance(src_obj, dict)
if isinstance(src_obj, tuple):
src_obj_ = transform_named_tuple_to_dict(src_obj)
else:
src_obj_ = src_obj
assert isinstance(src_obj_, dict)
class Option(object):
pass
option = Option()
for k, v in src_obj_.items():
setattr(option, k, v)
return option
args = parse_args()
args = parse_parser_add_arg(args, as_named_tuple = True)
'''
export MODEL_NAME="stable-diffusion-v1-4/"
export dataset_name="lambdalabs/pokemon-blip-captions"
accelerate launch train_text_to_image_ori.py \
--pretrained_model_name_or_path=$MODEL_NAME \
--dataset_name=$dataset_name \
--use_ema \
--resolution=32 --center_crop --random_flip \
--train_batch_size=1 \
--gradient_accumulation_steps=4 \
--gradient_checkpointing \
--max_train_steps=15000 \
--learning_rate=1e-05 \
--max_grad_norm=1 \
--lr_scheduler="constant" --lr_warmup_steps=0 \
--output_dir="sd-pokemon-model"
'''
args_dict = transform_named_tuple_to_dict(args)
args_dict["pretrained_model_name_or_path"] = "stable-diffusion-v1-4/"
#args_dict["pretrained_model_name_or_path"] = "japanese-stable-diffusion/"
args_dict["dataset_name"] = "svjack/pokemon-blip-captions-en-ja"
args_dict["use_ema"] = True
###args_dict["use_ema"] = False
args_dict["resolution"] = 256
args_dict["center_crop"] = True
args_dict["random_flip"] = True
args_dict["train_batch_size"] = 1
args_dict["gradient_accumulation_steps"] = 4
args_dict["train_batch_size"] = 4
args_dict["gradient_checkpointing"] = True
#### to 15000
args_dict["max_train_steps"] = 50000
args_dict["learning_rate"] = 1e-05
args_dict["max_grad_norm"] = 1
args_dict["lr_scheduler"] = "constant"
args_dict["lr_warmup_steps"] = 0
args_dict["output_dir"] = "sd-pokemon-model"
args_dict["caption_column"] = "en_text"
args_dict["mixed_precision"] = "no"
args = transform_dict_to_named_tuple(args_dict)
def get_full_repo_name(model_id: str, organization: Optional[str] = None, token: Optional[str] = None):
if token is None:
token = HfFolder.get_token()
if organization is None:
username = whoami(token)["name"]
return f"{username}/{model_id}"
else:
return f"{organization}/{model_id}"
dataset_name_mapping = {
"svjack/pokemon-blip-captions-en-ja": ("image", "en_text"),
}
class EMAModel:
"""
Exponential Moving Average of models weights
"""
def __init__(self, parameters: Iterable[torch.nn.Parameter], decay=0.9999):
parameters = list(parameters)
self.shadow_params = [p.clone().detach() for p in parameters]
self.decay = decay
self.optimization_step = 0
def get_decay(self, optimization_step):
"""
Compute the decay factor for the exponential moving average.
"""
value = (1 + optimization_step) / (10 + optimization_step)
return 1 - min(self.decay, value)
@torch.no_grad()
def step(self, parameters):
parameters = list(parameters)
self.optimization_step += 1
self.decay = self.get_decay(self.optimization_step)
for s_param, param in zip(self.shadow_params, parameters):
if param.requires_grad:
tmp = self.decay * (s_param - param)
s_param.sub_(tmp)
else:
s_param.copy_(param)
torch.cuda.empty_cache()
def copy_to(self, parameters: Iterable[torch.nn.Parameter]) -> None:
"""
Copy current averaged parameters into given collection of parameters.
Args:
parameters: Iterable of `torch.nn.Parameter`; the parameters to be
updated with the stored moving averages. If `None`, the
parameters with which this `ExponentialMovingAverage` was
initialized will be used.
"""
parameters = list(parameters)
for s_param, param in zip(self.shadow_params, parameters):
param.data.copy_(s_param.data)
def to(self, device=None, dtype=None) -> None:
r"""Move internal buffers of the ExponentialMovingAverage to `device`.
Args:
device: like `device` argument to `torch.Tensor.to`
"""
# .to() on the tensors handles None correctly
self.shadow_params = [
p.to(device=device, dtype=dtype) if p.is_floating_point() else p.to(device=device)
for p in self.shadow_params
]
#### model prepare
'''
from japanese_stable_diffusion import JapaneseStableDiffusionPipeline
import torch
import pandas as pd
from torch import autocast
from diffusers import LMSDiscreteScheduler
scheduler = LMSDiscreteScheduler(beta_start=0.00085, beta_end=0.012,
beta_schedule="scaled_linear", num_train_timesteps=1000)
# pretrained_model_name_or_path = "jap_to_zh_35000"
pretrained_model_name_or_path = "japanese-stable-diffusion/"
pipe = JapaneseStableDiffusionPipeline.from_pretrained(pretrained_model_name_or_path,
scheduler=scheduler, use_auth_token=True)
tokenizer, text_encoder, vae, unet = pipe.tokenizer, pipe.text_encoder, pipe.vae, pipe.unet
'''
tokenizer = CLIPTokenizer.from_pretrained(args.pretrained_model_name_or_path, subfolder="tokenizer")
text_encoder = CLIPTextModel.from_pretrained(args.pretrained_model_name_or_path, subfolder="text_encoder")
vae = AutoencoderKL.from_pretrained(args.pretrained_model_name_or_path, subfolder="vae")
unet = UNet2DConditionModel.from_pretrained(args.pretrained_model_name_or_path, subfolder="unet")
vae.requires_grad_(False)
text_encoder.requires_grad_(False)
if args.gradient_checkpointing:
unet.enable_gradient_checkpointing()
if args.scale_lr:
args.learning_rate = (
args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
)
# Initialize the optimizer
if args.use_8bit_adam:
try:
import bitsandbytes as bnb
except ImportError:
raise ImportError(
"Please install bitsandbytes to use 8-bit Adam. You can do so by running `pip install bitsandbytes`"
)
optimizer_cls = bnb.optim.AdamW8bit
else:
optimizer_cls = torch.optim.AdamW
optimizer = optimizer_cls(
unet.parameters(),
lr=args.learning_rate,
betas=(args.adam_beta1, args.adam_beta2),
weight_decay=args.adam_weight_decay,
eps=args.adam_epsilon,
)
noise_scheduler = DDPMScheduler(
beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear", num_train_timesteps=1000,
#tensor_format="pt"
)
dataset = load_dataset(
args.dataset_name,
args.dataset_config_name,
cache_dir=args.cache_dir,
)
column_names = dataset["train"].column_names
dataset_columns = dataset_name_mapping.get(args.dataset_name, None)
if args.image_column is None:
image_column = dataset_columns[0] if dataset_columns is not None else column_names[0]
else:
image_column = args.image_column
if image_column not in column_names:
raise ValueError(
f"--image_column' value '{args.image_column}' needs to be one of: {', '.join(column_names)}"
)
if args.caption_column is None:
caption_column = dataset_columns[1] if dataset_columns is not None else column_names[1]
else:
caption_column = args.caption_column
if caption_column not in column_names:
raise ValueError(
f"--caption_column' value '{args.caption_column}' needs to be one of: {', '.join(column_names)}"
)
# Preprocessing the datasets.
# We need to tokenize input captions and transform the images.
def tokenize_captions(examples, is_train=True):
captions = []
for caption in examples[caption_column]:
if isinstance(caption, str):
captions.append(caption)
elif isinstance(caption, (list, np.ndarray)):
# take a random caption if there are multiple
captions.append(random.choice(caption) if is_train else caption[0])
else:
raise ValueError(
f"Caption column `{caption_column}` should contain either strings or lists of strings."
)
inputs = tokenizer(captions, max_length=tokenizer.model_max_length, padding="do_not_pad", truncation=True)
input_ids = inputs.input_ids
return input_ids
train_transforms = transforms.Compose(
[
transforms.Resize((args.resolution, args.resolution), interpolation=transforms.InterpolationMode.BILINEAR),
transforms.CenterCrop(args.resolution) if args.center_crop else transforms.RandomCrop(args.resolution),
transforms.RandomHorizontalFlip() if args.random_flip else transforms.Lambda(lambda x: x),
transforms.ToTensor(),
transforms.Normalize([0.5], [0.5]),
]
)
def preprocess_train(examples):
images = [image.convert("RGB") for image in examples[image_column]]
examples["pixel_values"] = [train_transforms(image) for image in images]
examples["input_ids"] = tokenize_captions(examples)
return examples
if args.max_train_samples is not None:
dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
train_dataset = dataset["train"].with_transform(preprocess_train)
def collate_fn(examples):
pixel_values = torch.stack([example["pixel_values"] for example in examples])
pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
input_ids = [example["input_ids"] for example in examples]
padded_tokens = tokenizer.pad({"input_ids": input_ids}, padding=True, return_tensors="pt")
return {
"pixel_values": pixel_values,
"input_ids": padded_tokens.input_ids,
"attention_mask": padded_tokens.attention_mask,
}
train_dataloader = torch.utils.data.DataLoader(
train_dataset, shuffle=True, collate_fn=collate_fn, batch_size=args.train_batch_size
)
overrode_max_train_steps = False
num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
if args.max_train_steps is None:
args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
overrode_max_train_steps = True
lr_scheduler = get_scheduler(
args.lr_scheduler,
optimizer=optimizer,
num_warmup_steps=args.lr_warmup_steps * args.gradient_accumulation_steps,
num_training_steps=args.max_train_steps * args.gradient_accumulation_steps,
)
def train_loop(config, model, noise_scheduler, optimizer, train_dataloader, lr_scheduler):
# Initialize accelerator and tensorboard logging
args = config
unet = model
logging_dir = os.path.join(args.output_dir, args.logging_dir)
accelerator = Accelerator(
gradient_accumulation_steps=args.gradient_accumulation_steps,
mixed_precision=args.mixed_precision,
log_with=args.report_to,
logging_dir=logging_dir,
)
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
if args.seed is not None:
set_seed(args.seed)
if accelerator.is_main_process:
os.makedirs(args.output_dir, exist_ok=True)
weight_dtype = torch.float32
if args.mixed_precision == "fp16":
weight_dtype = torch.float16
elif args.mixed_precision == "bf16":
weight_dtype = torch.bfloat16
text_encoder.to(accelerator.device, dtype=weight_dtype)
vae.to(accelerator.device, dtype=weight_dtype)
if args.use_ema:
ema_unet = EMAModel(unet.parameters())
ema_unet.to(accelerator.device, dtype=weight_dtype)
num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
if overrode_max_train_steps:
args_dict = transform_named_tuple_to_dict(args)
args_dict["max_train_steps"] = args.num_train_epochs * num_update_steps_per_epoch
args = transform_dict_to_named_tuple(args_dict)
args_dict = transform_named_tuple_to_dict(args)
args_dict["num_train_epochs"] = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
args = transform_dict_to_named_tuple(args_dict)
if accelerator.is_main_process:
if config.push_to_hub:
#repo = init_git_repo(config, at_init=True)
pass
accelerator.init_trackers("train_example")
total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(train_dataset)}")
logger.info(f" Num Epochs = {args.num_train_epochs}")
logger.info(f" Instantaneous batch size per device = {args.train_batch_size}")
logger.info(f" Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
logger.info(f" Gradient Accumulation steps = {args.gradient_accumulation_steps}")
logger.info(f" Total optimization steps = {args.max_train_steps}")
# Prepare everything
# There is no specific order to remember, you just need to unpack the
# objects in the same order you gave them to the prepare method.
'''
model, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
model, optimizer, train_dataloader, lr_scheduler
)
'''
unet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
unet, optimizer, train_dataloader, lr_scheduler
)
progress_bar = tqdm(range(args.max_train_steps), disable=not accelerator.is_local_main_process)
progress_bar.set_description("Steps")
global_step = 0
for epoch in range(args.num_train_epochs):
#unet.train()
train_loss = 0.0
for step, batch in enumerate(train_dataloader):
with accelerator.accumulate(unet):
# Convert images to latent space
latents = vae.encode(batch["pixel_values"].to(weight_dtype)).latent_dist.sample()
latents = latents * 0.18215
# Sample noise that we'll add to the latents
noise = torch.randn_like(latents)
bsz = latents.shape[0]
# Sample a random timestep for each image
timesteps = torch.randint(0, noise_scheduler.num_train_timesteps, (bsz,), device=latents.device)
timesteps = timesteps.long()
# Add noise to the latents according to the noise magnitude at each timestep
# (this is the forward diffusion process)
noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)
# Get the text embedding for conditioning
encoder_hidden_states = text_encoder(batch["input_ids"])[0]
# Predict the noise residual and compute loss
noise_pred = unet(noisy_latents, timesteps, encoder_hidden_states).sample
loss = F.mse_loss(noise_pred.float(), noise.float(), reduction="mean")
#loss = F.mse_loss(noise.float(), noise.float(), reduction="mean")
# Gather the losses across all processes for logging (if we use distributed training).
avg_loss = accelerator.gather(loss.repeat(args.train_batch_size)).mean()
train_loss += avg_loss.item() / args.gradient_accumulation_steps
# Backpropagate
accelerator.backward(loss)
if accelerator.sync_gradients:
accelerator.clip_grad_norm_(unet.parameters(), args.max_grad_norm)
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
# Checks if the accelerator has performed an optimization step behind the scenes
if accelerator.sync_gradients:
if args.use_ema:
ema_unet.step(unet.parameters())
progress_bar.update(1)
global_step += 1
accelerator.log({"train_loss": train_loss}, step=global_step)
train_loss = 0.0
logs = {"step_loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
progress_bar.set_postfix(**logs)
if global_step >= args.max_train_steps:
break
from accelerate import notebook_launcher
#### train it.
args_ = (args, unet, noise_scheduler, optimizer, train_dataloader, lr_scheduler)
notebook_launcher(train_loop, args_, num_processes=1)
from diffusers import LMSDiscreteScheduler
scheduler = LMSDiscreteScheduler(beta_start=0.00085, beta_end=0.012,
beta_schedule="scaled_linear", num_train_timesteps=1000)
#### save to local
pipeline = StableDiffusionPipeline(
text_encoder=text_encoder,
vae=vae,
unet=unet,
tokenizer=tokenizer,
scheduler=scheduler,
safety_checker=StableDiffusionSafetyChecker.from_pretrained("CompVis/stable-diffusion-safety-checker"),
feature_extractor=CLIPFeatureExtractor.from_pretrained("openai/clip-vit-base-patch32"),
)
pipeline.save_pretrained("en_model_26000")
pipeline.safety_checker = lambda images, clip_input: (images, False)
pipeline = pipeline.to("cuda")
imgs = pipeline("A cartoon character with a potted plant on his head",
num_inference_steps = 100
)
imgs.images[0]
imgs = pipeline("cartoon bird",
num_inference_steps = 100
)
imgs.images[0]
imgs = pipeline("yellow ball",
num_inference_steps = 100
)
imgs.images[0]
imgs = pipeline("blue dragon illustration",
num_inference_steps = 100
)
imgs.images[0]