Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,63 +1,17 @@
|
|
1 |
import json
|
2 |
import random
|
3 |
-
import requests
|
4 |
-
import os
|
5 |
-
from PIL import Image
|
6 |
|
7 |
import gradio as gr
|
8 |
import numpy as np
|
9 |
import spaces
|
10 |
import torch
|
11 |
from diffusers import DiffusionPipeline, LCMScheduler
|
12 |
-
from peft import PeftModel
|
13 |
-
|
14 |
-
# Custom LCMScheduler to ignore unexpected attributes
|
15 |
-
class CustomLCMScheduler(LCMScheduler):
|
16 |
-
@property
|
17 |
-
def config(self):
|
18 |
-
return {k: v for k, v in super().config.items() if k != "skip_prk_steps"}
|
19 |
-
|
20 |
-
def get_image(image_data):
|
21 |
-
if isinstance(image_data, str):
|
22 |
-
return image_data
|
23 |
-
|
24 |
-
if isinstance(image_data, dict):
|
25 |
-
local_path = image_data.get('local_path')
|
26 |
-
hf_url = image_data.get('hf_url')
|
27 |
-
else:
|
28 |
-
print(f"Unexpected image_data format: {type(image_data)}")
|
29 |
-
return None
|
30 |
-
|
31 |
-
# Try loading from local path first
|
32 |
-
if local_path and os.path.exists(local_path):
|
33 |
-
try:
|
34 |
-
Image.open(local_path).verify() # Verify that it's a valid image
|
35 |
-
return local_path
|
36 |
-
except Exception as e:
|
37 |
-
print(f"Error loading local image {local_path}: {e}")
|
38 |
-
|
39 |
-
# If local path fails or doesn't exist, try URL
|
40 |
-
if hf_url:
|
41 |
-
try:
|
42 |
-
response = requests.get(hf_url)
|
43 |
-
if response.status_code == 200:
|
44 |
-
img = Image.open(requests.get(hf_url, stream=True).raw)
|
45 |
-
img.verify() # Verify that it's a valid image
|
46 |
-
img.save(local_path) # Save for future use
|
47 |
-
return local_path
|
48 |
-
else:
|
49 |
-
print(f"Failed to fetch image from URL {hf_url}. Status code: {response.status_code}")
|
50 |
-
except Exception as e:
|
51 |
-
print(f"Error loading image from URL {hf_url}: {e}")
|
52 |
-
|
53 |
-
print(f"Failed to load image for {image_data}")
|
54 |
-
return None
|
55 |
|
56 |
with open("sdxl_lora.json", "r") as file:
|
57 |
data = json.load(file)
|
58 |
sdxl_loras_raw = [
|
59 |
{
|
60 |
-
"image":
|
61 |
"title": item["title"],
|
62 |
"repo": item["repo"],
|
63 |
"trigger_word": item["trigger_word"],
|
@@ -69,27 +23,33 @@ with open("sdxl_lora.json", "r") as file:
|
|
69 |
for item in data
|
70 |
]
|
71 |
|
|
|
72 |
sdxl_loras_raw = sorted(sdxl_loras_raw, key=lambda x: x["likes"], reverse=True)
|
73 |
|
74 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
75 |
model_id = "stabilityai/stable-diffusion-xl-base-1.0"
|
76 |
|
77 |
pipe = DiffusionPipeline.from_pretrained(model_id, variant="fp16")
|
78 |
-
pipe.scheduler =
|
|
|
79 |
pipe.to(device=DEVICE, dtype=torch.float16)
|
80 |
|
81 |
-
# Load Flash SDXL LoRA
|
82 |
-
flash_sdxl_id = "jasperai/flash-sdxl"
|
83 |
-
pipe.load_lora_weights(flash_sdxl_id, adapter_name="flash_lora")
|
84 |
|
85 |
MAX_SEED = np.iinfo(np.int32).max
|
86 |
MAX_IMAGE_SIZE = 1024
|
87 |
|
88 |
-
|
|
|
|
|
|
|
|
|
|
|
89 |
lora_id = gr_sdxl_loras[selected_state.index]["repo"]
|
90 |
trigger_word = gr_sdxl_loras[selected_state.index]["trigger_word"]
|
|
|
91 |
return lora_id, trigger_word
|
92 |
|
|
|
93 |
@spaces.GPU
|
94 |
def infer(
|
95 |
pre_prompt,
|
@@ -103,51 +63,19 @@ def infer(
|
|
103 |
user_lora_weight,
|
104 |
progress=gr.Progress(track_tqdm=True),
|
105 |
):
|
106 |
-
|
107 |
-
# Load the user-selected LoRA
|
108 |
-
new_adapter_id = user_lora_selector.replace("/", "_")
|
109 |
-
pipe.load_lora_weights(user_lora_selector, adapter_name=new_adapter_id)
|
110 |
|
111 |
-
|
112 |
-
|
113 |
-
gr.Info("LoRA setup complete")
|
114 |
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
if pre_prompt != "":
|
121 |
-
prompt = f"{pre_prompt} {prompt}"
|
122 |
-
|
123 |
-
# Use Flash Diffusion settings
|
124 |
-
image = pipe(
|
125 |
-
prompt=prompt,
|
126 |
-
negative_prompt=negative_prompt,
|
127 |
-
guidance_scale=1.0, # Flash Diffusion typically uses guidance_scale=1
|
128 |
-
num_inference_steps=4, # Flash Diffusion uses fewer steps
|
129 |
-
generator=generator,
|
130 |
-
).images[0]
|
131 |
-
|
132 |
-
return image
|
133 |
-
except Exception as e:
|
134 |
-
gr.Error(f"An error occurred: {str(e)}")
|
135 |
-
return None
|
136 |
|
137 |
-
|
138 |
-
|
139 |
-
pre_prompt,
|
140 |
-
prompt,
|
141 |
-
seed,
|
142 |
-
randomize_seed,
|
143 |
-
num_inference_steps,
|
144 |
-
negative_prompt,
|
145 |
-
guidance_scale,
|
146 |
-
user_lora_selector,
|
147 |
-
user_lora_weight,
|
148 |
-
progress=gr.Progress(track_tqdm=True),
|
149 |
-
):
|
150 |
-
load_lora_for_style(user_lora_selector)
|
151 |
|
152 |
if randomize_seed:
|
153 |
seed = random.randint(0, MAX_SEED)
|
|
|
1 |
import json
|
2 |
import random
|
|
|
|
|
|
|
3 |
|
4 |
import gradio as gr
|
5 |
import numpy as np
|
6 |
import spaces
|
7 |
import torch
|
8 |
from diffusers import DiffusionPipeline, LCMScheduler
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
with open("sdxl_lora.json", "r") as file:
|
11 |
data = json.load(file)
|
12 |
sdxl_loras_raw = [
|
13 |
{
|
14 |
+
"image": item["image"],
|
15 |
"title": item["title"],
|
16 |
"repo": item["repo"],
|
17 |
"trigger_word": item["trigger_word"],
|
|
|
23 |
for item in data
|
24 |
]
|
25 |
|
26 |
+
# Sort the loras by likes
|
27 |
sdxl_loras_raw = sorted(sdxl_loras_raw, key=lambda x: x["likes"], reverse=True)
|
28 |
|
29 |
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
30 |
model_id = "stabilityai/stable-diffusion-xl-base-1.0"
|
31 |
|
32 |
pipe = DiffusionPipeline.from_pretrained(model_id, variant="fp16")
|
33 |
+
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
|
34 |
+
pipe.load_lora_weights("jasperai/flash-sdxl", adapter_name="lora")
|
35 |
pipe.to(device=DEVICE, dtype=torch.float16)
|
36 |
|
|
|
|
|
|
|
37 |
|
38 |
MAX_SEED = np.iinfo(np.int32).max
|
39 |
MAX_IMAGE_SIZE = 1024
|
40 |
|
41 |
+
|
42 |
+
def update_selection(
|
43 |
+
selected_state: gr.SelectData,
|
44 |
+
gr_sdxl_loras,
|
45 |
+
):
|
46 |
+
|
47 |
lora_id = gr_sdxl_loras[selected_state.index]["repo"]
|
48 |
trigger_word = gr_sdxl_loras[selected_state.index]["trigger_word"]
|
49 |
+
|
50 |
return lora_id, trigger_word
|
51 |
|
52 |
+
|
53 |
@spaces.GPU
|
54 |
def infer(
|
55 |
pre_prompt,
|
|
|
63 |
user_lora_weight,
|
64 |
progress=gr.Progress(track_tqdm=True),
|
65 |
):
|
66 |
+
flash_sdxl_id = "jasperai/flash-sdxl"
|
|
|
|
|
|
|
67 |
|
68 |
+
new_adapter_id = user_lora_selector.replace("/", "_")
|
69 |
+
loaded_adapters = pipe.get_list_adapters()
|
|
|
70 |
|
71 |
+
if new_adapter_id not in loaded_adapters["unet"]:
|
72 |
+
gr.Info("Swapping LoRA")
|
73 |
+
pipe.unload_lora_weights()
|
74 |
+
pipe.load_lora_weights(flash_sdxl_id, adapter_name="lora")
|
75 |
+
pipe.load_lora_weights(user_lora_selector, adapter_name=new_adapter_id)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
|
77 |
+
pipe.set_adapters(["lora", new_adapter_id], adapter_weights=[1.0, user_lora_weight])
|
78 |
+
gr.Info("LoRA setup done")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
|
80 |
if randomize_seed:
|
81 |
seed = random.randint(0, MAX_SEED)
|