Spaces:
Running
Running
cocktailpeanut
commited on
Commit
•
824b2b5
1
Parent(s):
da30812
update
Browse files- app.py +30 -22
- requirements.txt +1 -1
app.py
CHANGED
@@ -21,26 +21,26 @@ DESCRIPTION += "\n<p style=\"text-align: center\">Unofficial demo for <a href='h
|
|
21 |
|
22 |
MAX_SEED = np.iinfo(np.int32).max
|
23 |
CACHE_EXAMPLES = False
|
24 |
-
#CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES") != "0"
|
25 |
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "1536"))
|
26 |
USE_TORCH_COMPILE = False
|
27 |
ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD") == "1"
|
|
|
28 |
PREVIEW_IMAGES = True
|
29 |
|
30 |
dtype = torch.bfloat16
|
31 |
-
#device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
32 |
if torch.cuda.is_available():
|
33 |
-
|
34 |
elif torch.backends.mps.is_available():
|
35 |
-
|
36 |
-
|
37 |
else:
|
38 |
-
|
|
|
39 |
print(f"device={device}")
|
40 |
-
|
41 |
if device != "cpu":
|
42 |
-
prior_pipeline = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", torch_dtype=dtype
|
43 |
-
decoder_pipeline = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade", torch_dtype=dtype
|
44 |
|
45 |
if ENABLE_CPU_OFFLOAD:
|
46 |
prior_pipeline.enable_model_cpu_offload()
|
@@ -57,10 +57,12 @@ if device != "cpu":
|
|
57 |
previewer = Previewer()
|
58 |
previewer_state_dict = torch.load("previewer/previewer_v1_100k.pt", map_location=torch.device('cpu'))["state_dict"]
|
59 |
previewer.load_state_dict(previewer_state_dict)
|
60 |
-
def callback_prior(
|
|
|
61 |
output = previewer(latents)
|
62 |
output = numpy_to_pil(output.clamp(0, 1).permute(0, 2, 3, 1).float().cpu().numpy())
|
63 |
-
|
|
|
64 |
callback_steps = 1
|
65 |
else:
|
66 |
previewer = None
|
@@ -92,14 +94,19 @@ def generate(
|
|
92 |
num_images_per_prompt: int = 2,
|
93 |
# profile: gr.OAuthProfile | None = None,
|
94 |
) -> PIL.Image.Image:
|
|
|
95 |
try:
|
96 |
previewer.eval().requires_grad_(False).to(device).to(dtype)
|
97 |
except:
|
98 |
print("")
|
99 |
-
prior_pipeline.to(device)
|
100 |
-
decoder_pipeline.to(device)
|
101 |
|
|
|
|
|
|
|
|
|
|
|
102 |
generator = torch.Generator().manual_seed(seed)
|
|
|
103 |
prior_output = prior_pipeline(
|
104 |
prompt=prompt,
|
105 |
height=height,
|
@@ -110,15 +117,17 @@ def generate(
|
|
110 |
guidance_scale=prior_guidance_scale,
|
111 |
num_images_per_prompt=num_images_per_prompt,
|
112 |
generator=generator,
|
113 |
-
|
114 |
-
|
115 |
)
|
116 |
-
|
117 |
if PREVIEW_IMAGES:
|
118 |
for _ in range(len(DEFAULT_STAGE_C_TIMESTEPS)):
|
119 |
-
|
120 |
-
|
121 |
-
|
|
|
|
|
|
|
122 |
prior_output = r
|
123 |
|
124 |
decoder_output = decoder_pipeline(
|
@@ -131,7 +140,6 @@ def generate(
|
|
131 |
generator=generator,
|
132 |
output_type="pil",
|
133 |
).images
|
134 |
-
|
135 |
# #Save images
|
136 |
# for image in decoder_output:
|
137 |
# user_history.save_image(
|
@@ -156,7 +164,7 @@ def generate(
|
|
156 |
examples = [
|
157 |
"An astronaut riding a green horse",
|
158 |
"A mecha robot in a favela by Tarsila do Amaral",
|
159 |
-
"The
|
160 |
"A delicious feijoada ramen dish"
|
161 |
]
|
162 |
|
@@ -289,4 +297,4 @@ with gr.Blocks(css="style.css") as demo_with_history:
|
|
289 |
# user_history.render()
|
290 |
|
291 |
if __name__ == "__main__":
|
292 |
-
demo_with_history.queue(max_size=20).launch()
|
|
|
21 |
|
22 |
MAX_SEED = np.iinfo(np.int32).max
|
23 |
CACHE_EXAMPLES = False
|
|
|
24 |
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "1536"))
|
25 |
USE_TORCH_COMPILE = False
|
26 |
ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD") == "1"
|
27 |
+
#PREVIEW_IMAGES = False
|
28 |
PREVIEW_IMAGES = True
|
29 |
|
30 |
dtype = torch.bfloat16
|
|
|
31 |
if torch.cuda.is_available():
|
32 |
+
device = "cuda"
|
33 |
elif torch.backends.mps.is_available():
|
34 |
+
device = "mps"
|
35 |
+
dtype = torch.float32
|
36 |
else:
|
37 |
+
device = "cpu"
|
38 |
+
|
39 |
print(f"device={device}")
|
40 |
+
|
41 |
if device != "cpu":
|
42 |
+
prior_pipeline = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", torch_dtype=dtype)#.to(device)
|
43 |
+
decoder_pipeline = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade", torch_dtype=dtype)#.to(device)
|
44 |
|
45 |
if ENABLE_CPU_OFFLOAD:
|
46 |
prior_pipeline.enable_model_cpu_offload()
|
|
|
57 |
previewer = Previewer()
|
58 |
previewer_state_dict = torch.load("previewer/previewer_v1_100k.pt", map_location=torch.device('cpu'))["state_dict"]
|
59 |
previewer.load_state_dict(previewer_state_dict)
|
60 |
+
def callback_prior(pipeline, step_index, t, callback_kwargs):
|
61 |
+
latents = callback_kwargs["latents"]
|
62 |
output = previewer(latents)
|
63 |
output = numpy_to_pil(output.clamp(0, 1).permute(0, 2, 3, 1).float().cpu().numpy())
|
64 |
+
callback_kwargs["preview_output"] = output
|
65 |
+
return callback_kwargs
|
66 |
callback_steps = 1
|
67 |
else:
|
68 |
previewer = None
|
|
|
94 |
num_images_per_prompt: int = 2,
|
95 |
# profile: gr.OAuthProfile | None = None,
|
96 |
) -> PIL.Image.Image:
|
97 |
+
|
98 |
try:
|
99 |
previewer.eval().requires_grad_(False).to(device).to(dtype)
|
100 |
except:
|
101 |
print("")
|
|
|
|
|
102 |
|
103 |
+
#previewer.eval().requires_grad_(False).to(device).to(dtype)
|
104 |
+
# if device != "cpu":
|
105 |
+
# prior_pipeline.to(device)
|
106 |
+
# decoder_pipeline.to(device)
|
107 |
+
#
|
108 |
generator = torch.Generator().manual_seed(seed)
|
109 |
+
print("prior_num_inference_steps: ", prior_num_inference_steps)
|
110 |
prior_output = prior_pipeline(
|
111 |
prompt=prompt,
|
112 |
height=height,
|
|
|
117 |
guidance_scale=prior_guidance_scale,
|
118 |
num_images_per_prompt=num_images_per_prompt,
|
119 |
generator=generator,
|
120 |
+
#callback_on_step_end=callback_prior,
|
121 |
+
#callback_on_step_end_tensor_inputs=['latents']
|
122 |
)
|
|
|
123 |
if PREVIEW_IMAGES:
|
124 |
for _ in range(len(DEFAULT_STAGE_C_TIMESTEPS)):
|
125 |
+
try:
|
126 |
+
r = next(prior_output)
|
127 |
+
if isinstance(r, list):
|
128 |
+
yield r[0]
|
129 |
+
except:
|
130 |
+
print("")
|
131 |
prior_output = r
|
132 |
|
133 |
decoder_output = decoder_pipeline(
|
|
|
140 |
generator=generator,
|
141 |
output_type="pil",
|
142 |
).images
|
|
|
143 |
# #Save images
|
144 |
# for image in decoder_output:
|
145 |
# user_history.save_image(
|
|
|
164 |
examples = [
|
165 |
"An astronaut riding a green horse",
|
166 |
"A mecha robot in a favela by Tarsila do Amaral",
|
167 |
+
"The spirit of a Tamagotchi wandering in the city of Los Angeles",
|
168 |
"A delicious feijoada ramen dish"
|
169 |
]
|
170 |
|
|
|
297 |
# user_history.render()
|
298 |
|
299 |
if __name__ == "__main__":
|
300 |
+
demo_with_history.queue(max_size=20).launch()
|
requirements.txt
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
#git+https://github.com/kashif/diffusers.git@diffusers-yield-callback
|
2 |
#git+https://github.com/kashif/diffusers.git@a3dc21385b7386beb3dab3a9845962ede6765887
|
3 |
-
diffusers
|
4 |
accelerate
|
5 |
safetensors
|
6 |
transformers
|
|
|
1 |
#git+https://github.com/kashif/diffusers.git@diffusers-yield-callback
|
2 |
#git+https://github.com/kashif/diffusers.git@a3dc21385b7386beb3dab3a9845962ede6765887
|
3 |
+
git+https://github.com/apolinario/diffusers.git@yield-new-pipe
|
4 |
accelerate
|
5 |
safetensors
|
6 |
transformers
|