Spaces:
Runtime error
Runtime error
RamAnanth1
commited on
Commit
•
1eefa67
1
Parent(s):
a976ca4
Make sure process function accesses model
Browse files
model.py
CHANGED
@@ -76,15 +76,15 @@ class Model:
|
|
76 |
subprocess.run(shlex.split(f'wget {pidinet_file} -O models/table5_pidinet.pth'))
|
77 |
|
78 |
|
79 |
-
model = load_model_from_config(config, "models/sd-v1-4.ckpt").to(device)
|
80 |
current_base = 'sd-v1-4.ckpt'
|
81 |
-
model_ad = Adapter(channels=[320, 640, 1280, 1280][:4], nums_rb=2, ksize=1, sk=True, use_conv=False).to(device)
|
82 |
-
model_ad.load_state_dict(torch.load("models/t2iadapter_sketch_sd14v1.pth"))
|
83 |
net_G = pidinet()
|
84 |
ckp = torch.load('models/table5_pidinet.pth', map_location='cpu')['state_dict']
|
85 |
net_G.load_state_dict({k.replace('module.',''):v for k, v in ckp.items()})
|
86 |
net_G.to(device)
|
87 |
-
sampler = PLMSSampler(model)
|
88 |
save_memory=True
|
89 |
|
90 |
@torch.inference_mode()
|
@@ -121,17 +121,17 @@ class Model:
|
|
121 |
im = im.float()
|
122 |
im_edge = tensor2img(im)
|
123 |
|
124 |
-
c = model.get_learned_conditioning([prompt])
|
125 |
-
nc = model.get_learned_conditioning([neg_prompt])
|
126 |
|
127 |
with torch.no_grad():
|
128 |
# extract condition features
|
129 |
-
features_adapter = model_ad(im.to(device))
|
130 |
|
131 |
shape = [4, 64, 64]
|
132 |
|
133 |
# sampling
|
134 |
-
samples_ddim, _ = sampler.sample(S=50,
|
135 |
conditioning=c,
|
136 |
batch_size=1,
|
137 |
shape=shape,
|
@@ -144,7 +144,7 @@ class Model:
|
|
144 |
mode = 'sketch',
|
145 |
con_strength = con_strength)
|
146 |
|
147 |
-
x_samples_ddim = model.decode_first_stage(samples_ddim)
|
148 |
x_samples_ddim = torch.clamp((x_samples_ddim + 1.0) / 2.0, min=0.0, max=1.0)
|
149 |
x_samples_ddim = x_samples_ddim.permute(0, 2, 3, 1).numpy()[0]
|
150 |
x_samples_ddim = 255.*x_samples_ddim
|
|
|
76 |
subprocess.run(shlex.split(f'wget {pidinet_file} -O models/table5_pidinet.pth'))
|
77 |
|
78 |
|
79 |
+
self.model = load_model_from_config(config, "models/sd-v1-4.ckpt").to(device)
|
80 |
current_base = 'sd-v1-4.ckpt'
|
81 |
+
self.model_ad = Adapter(channels=[320, 640, 1280, 1280][:4], nums_rb=2, ksize=1, sk=True, use_conv=False).to(device)
|
82 |
+
self.model_ad.load_state_dict(torch.load("models/t2iadapter_sketch_sd14v1.pth"))
|
83 |
net_G = pidinet()
|
84 |
ckp = torch.load('models/table5_pidinet.pth', map_location='cpu')['state_dict']
|
85 |
net_G.load_state_dict({k.replace('module.',''):v for k, v in ckp.items()})
|
86 |
net_G.to(device)
|
87 |
+
self.sampler = PLMSSampler(self.model)
|
88 |
save_memory=True
|
89 |
|
90 |
@torch.inference_mode()
|
|
|
121 |
im = im.float()
|
122 |
im_edge = tensor2img(im)
|
123 |
|
124 |
+
c = self.model.get_learned_conditioning([prompt])
|
125 |
+
nc = self.model.get_learned_conditioning([neg_prompt])
|
126 |
|
127 |
with torch.no_grad():
|
128 |
# extract condition features
|
129 |
+
features_adapter = self.model_ad(im.to(device))
|
130 |
|
131 |
shape = [4, 64, 64]
|
132 |
|
133 |
# sampling
|
134 |
+
samples_ddim, _ = self.sampler.sample(S=50,
|
135 |
conditioning=c,
|
136 |
batch_size=1,
|
137 |
shape=shape,
|
|
|
144 |
mode = 'sketch',
|
145 |
con_strength = con_strength)
|
146 |
|
147 |
+
x_samples_ddim = self.model.decode_first_stage(samples_ddim)
|
148 |
x_samples_ddim = torch.clamp((x_samples_ddim + 1.0) / 2.0, min=0.0, max=1.0)
|
149 |
x_samples_ddim = x_samples_ddim.permute(0, 2, 3, 1).numpy()[0]
|
150 |
x_samples_ddim = 255.*x_samples_ddim
|