Update app.py
Browse files
app.py
CHANGED
@@ -8,39 +8,57 @@ from diffusers import StableDiffusionXLPipeline
|
|
8 |
from diffusers import EulerAncestralDiscreteScheduler
|
9 |
import torch
|
10 |
from compel import Compel, ReturnedEmbeddingsType
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
|
12 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
pipe.
|
29 |
-
pipe.
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
|
40 |
MAX_SEED = np.iinfo(np.int32).max
|
41 |
MAX_IMAGE_SIZE = 1216
|
42 |
|
43 |
-
#
|
44 |
def process_long_prompt(prompt, negative_prompt=""):
|
45 |
"""Simple long prompt processing using Compel"""
|
46 |
try:
|
@@ -52,7 +70,11 @@ def process_long_prompt(prompt, negative_prompt=""):
|
|
52 |
|
53 |
@spaces.GPU
|
54 |
def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
|
55 |
-
|
|
|
|
|
|
|
|
|
56 |
use_long_prompt = len(prompt.split()) > 60 or len(prompt) > 300
|
57 |
|
58 |
if randomize_seed:
|
@@ -61,7 +83,7 @@ def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance
|
|
61 |
generator = torch.Generator(device=device).manual_seed(seed)
|
62 |
|
63 |
try:
|
64 |
-
#
|
65 |
if use_long_prompt:
|
66 |
print("Using long prompt processing...")
|
67 |
conditioning, pooled = process_long_prompt(prompt, negative_prompt)
|
@@ -109,13 +131,15 @@ css = """
|
|
109 |
with gr.Blocks(css=css) as demo:
|
110 |
|
111 |
with gr.Column(elem_id="col-container"):
|
|
|
|
|
112 |
|
113 |
with gr.Row():
|
114 |
prompt = gr.Text(
|
115 |
label="Prompt",
|
116 |
show_label=False,
|
117 |
max_lines=1,
|
118 |
-
placeholder="Enter your prompt (long prompts are automatically supported)",
|
119 |
container=False,
|
120 |
)
|
121 |
|
@@ -182,4 +206,4 @@ with gr.Blocks(css=css) as demo:
|
|
182 |
outputs=[result]
|
183 |
)
|
184 |
|
185 |
-
demo.queue().launch()
|
|
|
8 |
from diffusers import EulerAncestralDiscreteScheduler
|
9 |
import torch
|
10 |
from compel import Compel, ReturnedEmbeddingsType
|
11 |
+
from huggingface_hub import login
|
12 |
+
import os
|
13 |
+
|
14 |
+
# Add your Hugging Face token here or set it as an environment variable
|
15 |
+
HF_TOKEN = os.getenv("HF_TOKEN") # Get from environment variable
|
16 |
+
# Or directly: HF_TOKEN = "hf_your_token_here"
|
17 |
+
|
18 |
+
if HF_TOKEN:
|
19 |
+
login(token=HF_TOKEN)
|
20 |
|
21 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
22 |
|
23 |
+
try:
|
24 |
+
# Make sure to use torch.float16 consistently throughout the pipeline
|
25 |
+
pipe = StableDiffusionXLPipeline.from_pretrained(
|
26 |
+
"votepurchase/waiREALCN_v14",
|
27 |
+
torch_dtype=torch.float16,
|
28 |
+
variant="fp16", # Explicitly use fp16 variant
|
29 |
+
use_safetensors=True, # Use safetensors if available
|
30 |
+
use_auth_token=HF_TOKEN # Pass token to download
|
31 |
+
)
|
32 |
+
|
33 |
+
pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
|
34 |
+
pipe.to(device)
|
35 |
+
|
36 |
+
# Force all components to use the same dtype
|
37 |
+
pipe.text_encoder.to(torch.float16)
|
38 |
+
pipe.text_encoder_2.to(torch.float16)
|
39 |
+
pipe.vae.to(torch.float16)
|
40 |
+
pipe.unet.to(torch.float16)
|
41 |
+
|
42 |
+
# Initialize Compel for long prompt processing
|
43 |
+
compel = Compel(
|
44 |
+
tokenizer=[pipe.tokenizer, pipe.tokenizer_2],
|
45 |
+
text_encoder=[pipe.text_encoder, pipe.text_encoder_2],
|
46 |
+
returned_embeddings_type=ReturnedEmbeddingsType.PENULTIMATE_HIDDEN_STATES_NON_NORMALIZED,
|
47 |
+
requires_pooled=[False, True],
|
48 |
+
truncate_long_prompts=False
|
49 |
+
)
|
50 |
+
|
51 |
+
model_loaded = True
|
52 |
+
except Exception as e:
|
53 |
+
print(f"Failed to load model: {e}")
|
54 |
+
model_loaded = False
|
55 |
+
pipe = None
|
56 |
+
compel = None
|
57 |
|
58 |
MAX_SEED = np.iinfo(np.int32).max
|
59 |
MAX_IMAGE_SIZE = 1216
|
60 |
|
61 |
+
# Simple long prompt processing function
|
62 |
def process_long_prompt(prompt, negative_prompt=""):
|
63 |
"""Simple long prompt processing using Compel"""
|
64 |
try:
|
|
|
70 |
|
71 |
@spaces.GPU
|
72 |
def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
|
73 |
+
if not model_loaded:
|
74 |
+
error_img = Image.new('RGB', (width, height), color=(50, 50, 50))
|
75 |
+
return error_img
|
76 |
+
|
77 |
+
# Remove the 60-word limit warning and add long prompt check
|
78 |
use_long_prompt = len(prompt.split()) > 60 or len(prompt) > 300
|
79 |
|
80 |
if randomize_seed:
|
|
|
83 |
generator = torch.Generator(device=device).manual_seed(seed)
|
84 |
|
85 |
try:
|
86 |
+
# Try long prompt processing first if prompt is long
|
87 |
if use_long_prompt:
|
88 |
print("Using long prompt processing...")
|
89 |
conditioning, pooled = process_long_prompt(prompt, negative_prompt)
|
|
|
131 |
with gr.Blocks(css=css) as demo:
|
132 |
|
133 |
with gr.Column(elem_id="col-container"):
|
134 |
+
if not model_loaded:
|
135 |
+
gr.Markdown("⚠️ **Model failed to load. Please check your Hugging Face token.**")
|
136 |
|
137 |
with gr.Row():
|
138 |
prompt = gr.Text(
|
139 |
label="Prompt",
|
140 |
show_label=False,
|
141 |
max_lines=1,
|
142 |
+
placeholder="Enter your prompt (long prompts are automatically supported)",
|
143 |
container=False,
|
144 |
)
|
145 |
|
|
|
206 |
outputs=[result]
|
207 |
)
|
208 |
|
209 |
+
demo.queue().launch()
|