full change
Browse files
main.py
CHANGED
@@ -34,25 +34,77 @@ class Upscale_CaseCade:
|
|
34 |
)
|
35 |
self.models_b.generator.eval().requires_grad_(False)
|
36 |
print("STAGE B READY")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
|
38 |
|
39 |
-
def upscale_image(self,image_pil,scale_fator):
|
40 |
batch_size = 1
|
41 |
cnet_override = None
|
42 |
images = resize_image(image_pil).unsqueeze(0).expand(batch_size, -1, -1, -1)
|
43 |
|
44 |
batch = {'images': images}
|
45 |
|
|
|
46 |
with torch.no_grad(), torch.cuda.amp.autocast(dtype=torch.bfloat16):
|
47 |
effnet_latents = self.core.encode_latents(batch, self.models, self.extras)
|
48 |
effnet_latents_up = torch.nn.functional.interpolate(effnet_latents, scale_factor=scale_fator, mode="nearest")
|
49 |
cnet = self.models.controlnet(effnet_latents_up)
|
50 |
cnet_uncond = cnet
|
51 |
cnet_input = torch.nn.functional.interpolate(images, scale_factor=scale_fator, mode="nearest")
|
52 |
-
# cnet, cnet_input = core.get_cnet(batch, models, extras)
|
53 |
# cnet_uncond = cnet
|
54 |
-
|
55 |
-
|
56 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
|
58 |
|
|
|
34 |
)
|
35 |
self.models_b.generator.eval().requires_grad_(False)
|
36 |
print("STAGE B READY")
|
37 |
+
self.caption = "a photo of image"
|
38 |
+
self.cnet_multiplier = 1.0 # 0.8 # 0.3
|
39 |
+
# Stage C Parameters
|
40 |
+
self.extras.sampling_configs['cfg'] = 1
|
41 |
+
self.extras.sampling_configs['shift'] = 2
|
42 |
+
self.extras.sampling_configs['timesteps'] = 20
|
43 |
+
self.extras.sampling_configs['t_start'] = 1.0
|
44 |
+
# Stage B Parameters
|
45 |
+
self.extras_b.sampling_configs['cfg'] = 1.1
|
46 |
+
self.extras_b.sampling_configs['shift'] = 1
|
47 |
+
self.extras_b.sampling_configs['timesteps'] = 10
|
48 |
+
self.extras_b.sampling_configs['t_start'] = 1.0
|
49 |
+
self.models = ControlNetCore.Models(
|
50 |
+
**{**self.models.to_dict(), 'generator': torch.compile(self.models.generator, mode="reduce-overhead", fullgraph=True)}
|
51 |
+
)
|
52 |
+
|
53 |
+
self.models_b = WurstCoreB.Models(
|
54 |
+
**{**self.models_b.to_dict(), 'generator': torch.compile(self.models_b.generator, mode="reduce-overhead", fullgraph=True)}
|
55 |
+
)
|
56 |
+
|
57 |
|
58 |
|
59 |
+
def upscale_image(self,caption,image_pil,scale_fator):
|
60 |
batch_size = 1
|
61 |
cnet_override = None
|
62 |
images = resize_image(image_pil).unsqueeze(0).expand(batch_size, -1, -1, -1)
|
63 |
|
64 |
batch = {'images': images}
|
65 |
|
66 |
+
|
67 |
with torch.no_grad(), torch.cuda.amp.autocast(dtype=torch.bfloat16):
|
68 |
effnet_latents = self.core.encode_latents(batch, self.models, self.extras)
|
69 |
effnet_latents_up = torch.nn.functional.interpolate(effnet_latents, scale_factor=scale_fator, mode="nearest")
|
70 |
cnet = self.models.controlnet(effnet_latents_up)
|
71 |
cnet_uncond = cnet
|
72 |
cnet_input = torch.nn.functional.interpolate(images, scale_factor=scale_fator, mode="nearest")
|
73 |
+
# cnet, cnet_input = self.core.get_cnet(batch, self.models, self.extras)
|
74 |
# cnet_uncond = cnet
|
75 |
+
height, width = int(cnet[0].size(-2)*32*4/3), int(cnet[0].size(-1)*32*4/3)
|
76 |
+
stage_c_latent_shape, stage_b_latent_shape = calculate_latent_sizes(height, width, batch_size=batch_size)
|
77 |
+
# PREPARE CONDITIONS
|
78 |
+
batch['captions'] = [caption] * batch_size
|
79 |
+
conditions = self.core.get_conditions(batch, self.models, self.extras, is_eval=True, is_unconditional=False, eval_image_embeds=False)
|
80 |
+
unconditions = self.core.get_conditions(batch, self.models, self.extras, is_eval=True, is_unconditional=True, eval_image_embeds=False)
|
81 |
+
conditions['cnet'] = [c.clone() * self.cnet_multiplier if c is not None else c for c in cnet]
|
82 |
+
unconditions['cnet'] = [c.clone() * self.cnet_multiplier if c is not None else c for c in cnet_uncond]
|
83 |
+
conditions_b = self.core_b.get_conditions(batch, self.models_b, self.extras_b, is_eval=True, is_unconditional=False)
|
84 |
+
unconditions_b = self.core_b.get_conditions(batch, self.models_b, self.extras_b, is_eval=True, is_unconditional=True)
|
85 |
+
# torch.manual_seed(42)
|
86 |
+
sampling_c = self.extras.gdf.sample(
|
87 |
+
self.models.generator, conditions, stage_c_latent_shape,
|
88 |
+
unconditions, device=device, **self.extras.sampling_configs,
|
89 |
+
)
|
90 |
+
for (sampled_c, _, _) in tqdm(sampling_c, total=self.extras.sampling_configs['timesteps']):
|
91 |
+
sampled_c = sampled_c
|
92 |
+
|
93 |
+
# preview_c = models.previewer(sampled_c).float()
|
94 |
+
# show_images(preview_c)
|
95 |
+
|
96 |
+
conditions_b['effnet'] = sampled_c
|
97 |
+
unconditions_b['effnet'] = torch.zeros_like(sampled_c)
|
98 |
+
|
99 |
+
sampling_b = self.extras_b.gdf.sample(
|
100 |
+
self.models_b.generator, conditions_b, stage_b_latent_shape,
|
101 |
+
unconditions_b, device=device, **self.extras_b.sampling_configs
|
102 |
+
)
|
103 |
+
for (sampled_b, _, _) in tqdm(sampling_b, total=self.extras_b.sampling_configs['timesteps']):
|
104 |
+
sampled_b = sampled_b
|
105 |
+
sampled = self.models_b.stage_a.decode(sampled_b).float()
|
106 |
+
# og=show_images(batch['images'],return_images=True)
|
107 |
+
upscale=show_images(sampled,return_images=True)
|
108 |
+
return upscale
|
109 |
|
110 |
|