-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_ti_inference.py
130 lines (104 loc) · 4.94 KB
/
run_ti_inference.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
import cv2
import datetime
import torch
import pathlib
import transformers
import diffusers
import numpy as np
from accelerate import Accelerator
from accelerate.utils import DistributedDataParallelKwargs, ProjectConfiguration, set_seed
from diffusers import AutoencoderKL, StableDiffusionXLPipeline
from diffusers.utils.torch_utils import randn_tensor
from safetensors.torch import load_file
from src.utils.logger import get_logger
from src.methods.textual_inversion.arguments import parse_inference_args
def generate_and_save(args, accelerator, pipeline, latents, path_to_save_images):
image = pipeline(
latents=latents,
num_inference_steps=args.num_inference_steps,
guidance_scale=args.guidance_scale,
prompt=args.prompt,
prompt_2=args.prompt_2,
negative_prompt=args.negative_prompt,
negative_prompt_2=args.negative_prompt_2
).images[0]
if accelerator.is_main_process:
ts_now = datetime.datetime.now().isoformat(timespec="seconds")
path_to_save_image = pathlib.Path(path_to_save_images,
"_".join(args.prompt.lower().replace(",","").replace(".", "").split(" "))+f"_{ts_now}.png")
image = cv2.cvtColor(np.asarray(image), cv2.COLOR_BGR2RGB)
cv2.imwrite(str(path_to_save_image), image)
def main(args):
logging_dir = pathlib.Path(args.output_dir, args.logging_dir)
path_to_save_images = pathlib.Path(args.output_dir, "inference_results")
path_to_save_images.mkdir(parents=True, exist_ok=True)
accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)
kwargs = DistributedDataParallelKwargs(find_unused_parameters=True)
accelerator = Accelerator(
mixed_precision=args.mixed_precision,
project_config=accelerator_project_config,
kwargs_handlers=[kwargs],
)
logger.info(accelerator.state, main_process_only=False)
if accelerator.is_local_main_process:
transformers.utils.logging.set_verbosity_warning()
diffusers.utils.logging.set_verbosity_info()
else:
transformers.utils.logging.set_verbosity_error()
diffusers.utils.logging.set_verbosity_error()
if args.seed is not None:
set_seed(args.seed)
vae_path = (
args.pretrained_model_name_or_path
if args.pretrained_vae_model_name_or_path is None
else args.pretrained_vae_model_name_or_path
)
vae = AutoencoderKL.from_pretrained(
vae_path,
subfolder="vae" if args.pretrained_vae_model_name_or_path is None else None,
revision=args.revision,
variant=args.variant,
)
# For mixed precision training we cast all non-trainable weights (vae, non-lora text_encoder and non-lora unet) to half-precision
# as these weights are only used for inference, keeping weights in full precision is not required.
weight_dtype = torch.float32
if accelerator.mixed_precision == "fp16":
weight_dtype = torch.float16
elif accelerator.mixed_precision == "bf16":
weight_dtype = torch.bfloat16
if args.mixed_precision == "fp16":
vae.to(weight_dtype)
logger.info(f"Loading pipeline in {weight_dtype} precision", main_process_only=True)
pipeline = StableDiffusionXLPipeline.from_pretrained(
args.pretrained_model_name_or_path,
vae=vae,
revision=args.revision,
variant=args.variant,
torch_dtype=weight_dtype,
).to(accelerator.device)
embeddings_state_dict = load_file(args.path_to_embeddings)
generator = torch.Generator(device=accelerator.device).manual_seed(args.seed) if args.seed else None
latents_shape = (1,
pipeline.unet.config.in_channels,
args.resolution // pipeline.vae_scale_factor,
args.resolution // pipeline.vae_scale_factor)
latents = randn_tensor(latents_shape, generator, device=accelerator.device, dtype=weight_dtype)
# Load token embeddings for text encoder one and text encoder two
pipeline.load_textual_inversion(
embeddings_state_dict["text_encoder_one"],
token=args.placeholder_token,
text_encoder=pipeline.text_encoder,
tokenizer=pipeline.tokenizer)
pipeline.load_textual_inversion(
embeddings_state_dict["text_encoder_two"],
token=args.placeholder_token,
text_encoder=pipeline.text_encoder_2,
tokenizer=pipeline.tokenizer_2)
for _ in range(args.num_images_to_generate):
# latents = randn_tensor(latents_shape, generator, device=accelerator.device, dtype=weight_dtype)
generate_and_save(args, accelerator, pipeline, latents, path_to_save_images)
if __name__ == "__main__":
args = parse_inference_args()
path_to_log_file = pathlib.Path(args.output_dir, args.logging_dir, "ti_inference.log")
logger = get_logger(__name__, log_level="INFO", path_to_log_file=path_to_log_file)
main(args)