style(method): rename variables
Browse files
svgdreamer.py
CHANGED
@@ -20,8 +20,8 @@ from svgdreamer.pipelines.SVGDreamer_pipeline import SVGDreamerPipeline
|
|
20 |
@hydra.main(version_base=None, config_path="conf", config_name='config')
|
21 |
def main(cfg: omegaconf.DictConfig):
|
22 |
"""
|
23 |
-
The project configuration is stored in './conf/config.yaml
|
24 |
-
And style configurations are stored in './conf/x/iconographic.yaml
|
25 |
"""
|
26 |
|
27 |
# set seed
|
|
|
20 |
@hydra.main(version_base=None, config_path="conf", config_name='config')
|
21 |
def main(cfg: omegaconf.DictConfig):
|
22 |
"""
|
23 |
+
The project configuration is stored in './conf/config.yaml'
|
24 |
+
And style configurations are stored in './conf/x/iconographic.yaml'
|
25 |
"""
|
26 |
|
27 |
# set seed
|
svgdreamer/painter/VPSD_pipeline.py
CHANGED
@@ -3,12 +3,9 @@
|
|
3 |
# Author: XiMing Xing
|
4 |
# Description:
|
5 |
import re
|
6 |
-
import PIL
|
7 |
-
from PIL import Image
|
8 |
from typing import Any, List, Optional, Union, Dict
|
9 |
from omegaconf import DictConfig
|
10 |
|
11 |
-
import numpy as np
|
12 |
import torch
|
13 |
import torch.nn.functional as F
|
14 |
from torchvision import transforms
|
|
|
3 |
# Author: XiMing Xing
|
4 |
# Description:
|
5 |
import re
|
|
|
|
|
6 |
from typing import Any, List, Optional, Union, Dict
|
7 |
from omegaconf import DictConfig
|
8 |
|
|
|
9 |
import torch
|
10 |
import torch.nn.functional as F
|
11 |
from torchvision import transforms
|
svgdreamer/pipelines/SVGDreamer_pipeline.py
CHANGED
@@ -439,14 +439,14 @@ class SVGDreamerPipeline(ModelState):
|
|
439 |
|
440 |
# for convenience
|
441 |
guidance_cfg = self.x_cfg.vpsd
|
442 |
-
|
443 |
n_particle = guidance_cfg.n_particle
|
444 |
total_step = guidance_cfg.num_iter
|
445 |
path_reinit = self.x_cfg.path_reinit
|
446 |
|
447 |
# init VPSD
|
448 |
-
pipeline = VectorizedParticleSDSPipeline(
|
449 |
-
|
450 |
# init reward model
|
451 |
reward_model = None
|
452 |
if guidance_cfg.phi_ReFL:
|
@@ -529,7 +529,7 @@ class SVGDreamerPipeline(ModelState):
|
|
529 |
negative_prompt=self.args.neg_prompt,
|
530 |
grad_scale=guidance_cfg.grad_scale,
|
531 |
enhance_particle=guidance_cfg.particle_aug,
|
532 |
-
im_size=model2res(
|
533 |
)
|
534 |
|
535 |
# Xing Loss for Self-Interaction Problem
|
|
|
439 |
|
440 |
# for convenience
|
441 |
guidance_cfg = self.x_cfg.vpsd
|
442 |
+
sd_model_cfg = self.x_cfg.vpsd_model_cfg
|
443 |
n_particle = guidance_cfg.n_particle
|
444 |
total_step = guidance_cfg.num_iter
|
445 |
path_reinit = self.x_cfg.path_reinit
|
446 |
|
447 |
# init VPSD
|
448 |
+
pipeline = VectorizedParticleSDSPipeline(
|
449 |
+
sd_model_cfg, self.args.diffuser, guidance_cfg, self.device, self.weight_dtype)
|
450 |
# init reward model
|
451 |
reward_model = None
|
452 |
if guidance_cfg.phi_ReFL:
|
|
|
529 |
negative_prompt=self.args.neg_prompt,
|
530 |
grad_scale=guidance_cfg.grad_scale,
|
531 |
enhance_particle=guidance_cfg.particle_aug,
|
532 |
+
im_size=model2res(sd_model_cfg.model_id)
|
533 |
)
|
534 |
|
535 |
# Xing Loss for Self-Interaction Problem
|