-
Notifications
You must be signed in to change notification settings - Fork 15
/
inference.py
164 lines (127 loc) · 7 KB
/
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
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
import torch
from omegaconf import OmegaConf
from transformers import CLIPTextModel, CLIPTokenizer
from diffusers import AutoencoderKL, LMSDiscreteScheduler
from my_model import unet_2d_condition
import json
from PIL import Image
from utils import compute_ca_loss, Pharse2idx, draw_box, setup_logger
import hydra
import os
from tqdm import tqdm
from utils import load_text_inversion
def inference(device, unet, vae, tokenizer, text_encoder, prompt, bboxes, phrases, cfg, logger):
logger.info("Inference")
logger.info(f"Prompt: {prompt}")
logger.info(f"Phrases: {phrases}")
# Get Object Positions
logger.info("Convert Phrases to Object Positions")
object_positions = Pharse2idx(prompt, phrases)
# Encode Classifier Embeddings
uncond_input = tokenizer(
[""] * cfg.inference.batch_size, padding="max_length", max_length=tokenizer.model_max_length, return_tensors="pt"
)
uncond_embeddings = text_encoder(uncond_input.input_ids.to(device))[0]
# Encode Prompt
input_ids = tokenizer(
[prompt] * cfg.inference.batch_size,
padding="max_length",
truncation=True,
max_length=tokenizer.model_max_length,
return_tensors="pt",
)
cond_embeddings = text_encoder(input_ids.input_ids.to(device))[0]
text_embeddings = torch.cat([uncond_embeddings, cond_embeddings])
generator = torch.manual_seed(cfg.inference.rand_seed) # Seed generator to create the initial latent noise
noise_scheduler = LMSDiscreteScheduler(beta_start=cfg.noise_schedule.beta_start, beta_end=cfg.noise_schedule.beta_end,
beta_schedule=cfg.noise_schedule.beta_schedule, num_train_timesteps=cfg.noise_schedule.num_train_timesteps)
latents = torch.randn(
(cfg.inference.batch_size, 4, 64, 64),
generator=generator,
).to(device)
noise_scheduler.set_timesteps(cfg.inference.timesteps)
latents = latents * noise_scheduler.init_noise_sigma
loss = torch.tensor(10000)
for index, t in enumerate(tqdm(noise_scheduler.timesteps)):
iteration = 0
while loss.item() / cfg.inference.loss_scale > cfg.inference.loss_threshold and iteration < cfg.inference.max_iter and index < cfg.inference.max_index_step:
latents = latents.requires_grad_(True)
latent_model_input = latents
latent_model_input = noise_scheduler.scale_model_input(latent_model_input, t)
noise_pred, attn_map_integrated_up, attn_map_integrated_mid, attn_map_integrated_down = \
unet(latent_model_input, t, encoder_hidden_states=cond_embeddings)
# update latents with guidance
loss = compute_ca_loss(attn_map_integrated_mid, attn_map_integrated_up, bboxes=bboxes,
object_positions=object_positions) * cfg.inference.loss_scale
grad_cond = torch.autograd.grad(loss.requires_grad_(True), [latents])[0]
latents = latents - grad_cond * noise_scheduler.sigmas[index] ** 2
iteration += 1
torch.cuda.empty_cache()
with torch.no_grad():
latent_model_input = torch.cat([latents] * 2)
latent_model_input = noise_scheduler.scale_model_input(latent_model_input, t)
noise_pred, attn_map_integrated_up, attn_map_integrated_mid, attn_map_integrated_down = \
unet(latent_model_input, t, encoder_hidden_states=text_embeddings)
noise_pred = noise_pred.sample
# perform guidance
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
noise_pred = noise_pred_uncond + cfg.inference.classifier_free_guidance * (noise_pred_text - noise_pred_uncond)
latents = noise_scheduler.step(noise_pred, t, latents).prev_sample
torch.cuda.empty_cache()
with torch.no_grad():
logger.info("Decode Image...")
latents = 1 / 0.18215 * latents
image = vae.decode(latents).sample
image = (image / 2 + 0.5).clamp(0, 1)
image = image.detach().cpu().permute(0, 2, 3, 1).numpy()
images = (image * 255).round().astype("uint8")
pil_images = [Image.fromarray(image) for image in images]
return pil_images
@hydra.main(version_base=None, config_path="conf", config_name="base_config")
def main(cfg):
# build and load model
with open(cfg.general.unet_config) as f:
unet_config = json.load(f)
unet = unet_2d_condition.UNet2DConditionModel(**unet_config).from_pretrained(cfg.general.model_path, subfolder="unet")
tokenizer = CLIPTokenizer.from_pretrained(cfg.general.model_path, subfolder="tokenizer")
text_encoder = CLIPTextModel.from_pretrained(cfg.general.model_path, subfolder="text_encoder")
vae = AutoencoderKL.from_pretrained(cfg.general.model_path, subfolder="vae")
if cfg.general.real_image_editing:
text_encoder, tokenizer = load_text_inversion(text_encoder, tokenizer, cfg.real_image_editing.placeholder_token, cfg.real_image_editing.text_inversion_path)
unet.load_state_dict(torch.load(cfg.real_image_editing.dreambooth_path)['unet'])
text_encoder.load_state_dict(torch.load(cfg.real_image_editing.dreambooth_path)['encoder'])
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
unet.to(device)
text_encoder.to(device)
vae.to(device)
# ------------------ example input ------------------
examples = {"prompt": "A hello kitty toy is playing with a purple ball.",
"phrases": "hello kitty; ball",
"bboxes": [[[0.1, 0.2, 0.5, 0.8]], [[0.75, 0.6, 0.95, 0.8]]],
'save_path': cfg.general.save_path
}
# ------------------ real image editing example input ------------------
if cfg.general.real_image_editing:
examples = {"prompt": "A {} is standing on grass.".format(cfg.real_image_editing.placeholder_token),
"phrases": "{}".format(cfg.real_image_editing.placeholder_token),
"bboxes": [[[0.4, 0.2, 0.9, 0.9]]],
'save_path': cfg.general.save_path
}
# ---------------------------------------------------
# Prepare the save path
if not os.path.exists(cfg.general.save_path):
os.makedirs(cfg.general.save_path)
logger = setup_logger(cfg.general.save_path, __name__)
logger.info(cfg)
# Save cfg
logger.info("save config to {}".format(os.path.join(cfg.general.save_path, 'config.yaml')))
OmegaConf.save(cfg, os.path.join(cfg.general.save_path, 'config.yaml'))
# Inference
pil_images = inference(device, unet, vae, tokenizer, text_encoder, examples['prompt'], examples['bboxes'], examples['phrases'], cfg, logger)
# Save example images
for index, pil_image in enumerate(pil_images):
image_path = os.path.join(cfg.general.save_path, 'example_{}.png'.format(index))
logger.info('save example image to {}'.format(image_path))
draw_box(pil_image, examples['bboxes'], examples['phrases'], image_path)
if __name__ == "__main__":
main()