Commit
•
2caf84c
1
Parent(s):
8458096
Add custom LoRA loading
Browse files
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
import json
|
3 |
import logging
|
@@ -5,6 +6,7 @@ import torch
|
|
5 |
from PIL import Image
|
6 |
import spaces
|
7 |
from diffusers import DiffusionPipeline
|
|
|
8 |
import copy
|
9 |
import random
|
10 |
import time
|
@@ -35,7 +37,6 @@ class calculateDuration:
|
|
35 |
else:
|
36 |
print(f"Elapsed time: {self.elapsed_time:.6f} seconds")
|
37 |
|
38 |
-
|
39 |
def update_selection(evt: gr.SelectData, width, height):
|
40 |
selected_lora = loras[evt.index]
|
41 |
new_placeholder = f"Type a prompt for {selected_lora['title']}"
|
@@ -86,9 +87,10 @@ def run_lora(prompt, cfg_scale, steps, selected_index, randomize_seed, seed, wid
|
|
86 |
with calculateDuration(f"Loading LoRA weights for {selected_lora['title']}"):
|
87 |
if "weights" in selected_lora:
|
88 |
pipe.load_lora_weights(lora_path, weight_name=selected_lora["weights"])
|
|
|
89 |
else:
|
90 |
pipe.load_lora_weights(lora_path)
|
91 |
-
|
92 |
# Set random seed for reproducibility
|
93 |
with calculateDuration("Randomizing seed"):
|
94 |
if randomize_seed:
|
@@ -96,9 +98,80 @@ def run_lora(prompt, cfg_scale, steps, selected_index, randomize_seed, seed, wid
|
|
96 |
|
97 |
image = generate_image(prompt, trigger_word, steps, seed, cfg_scale, width, height, lora_scale, progress)
|
98 |
pipe.to("cpu")
|
|
|
99 |
pipe.unload_lora_weights()
|
100 |
return image, seed
|
101 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
102 |
run_lora.zerogpu = True
|
103 |
|
104 |
css = '''
|
@@ -107,6 +180,10 @@ css = '''
|
|
107 |
#title h1{font-size: 3em; display:inline-flex; align-items:center}
|
108 |
#title img{width: 100px; margin-right: 0.5em}
|
109 |
#gallery .grid-wrap{height: 10vh}
|
|
|
|
|
|
|
|
|
110 |
'''
|
111 |
with gr.Blocks(theme=gr.themes.Soft(), css=css) as app:
|
112 |
title = gr.HTML(
|
@@ -129,7 +206,11 @@ with gr.Blocks(theme=gr.themes.Soft(), css=css) as app:
|
|
129 |
columns=3,
|
130 |
elem_id="gallery"
|
131 |
)
|
132 |
-
|
|
|
|
|
|
|
|
|
133 |
with gr.Column(scale=4):
|
134 |
result = gr.Image(label="Generated Image")
|
135 |
|
@@ -154,7 +235,15 @@ with gr.Blocks(theme=gr.themes.Soft(), css=css) as app:
|
|
154 |
inputs=[width, height],
|
155 |
outputs=[prompt, selected_info, selected_index, width, height]
|
156 |
)
|
157 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
158 |
gr.on(
|
159 |
triggers=[generate_button.click, prompt.submit],
|
160 |
fn=run_lora,
|
|
|
1 |
+
import os
|
2 |
import gradio as gr
|
3 |
import json
|
4 |
import logging
|
|
|
6 |
from PIL import Image
|
7 |
import spaces
|
8 |
from diffusers import DiffusionPipeline
|
9 |
+
from huggingface_hub import hf_hub_download, HfFileSystem, ModelCard, snapshot_download
|
10 |
import copy
|
11 |
import random
|
12 |
import time
|
|
|
37 |
else:
|
38 |
print(f"Elapsed time: {self.elapsed_time:.6f} seconds")
|
39 |
|
|
|
40 |
def update_selection(evt: gr.SelectData, width, height):
|
41 |
selected_lora = loras[evt.index]
|
42 |
new_placeholder = f"Type a prompt for {selected_lora['title']}"
|
|
|
87 |
with calculateDuration(f"Loading LoRA weights for {selected_lora['title']}"):
|
88 |
if "weights" in selected_lora:
|
89 |
pipe.load_lora_weights(lora_path, weight_name=selected_lora["weights"])
|
90 |
+
#pipe.fuse_lora()
|
91 |
else:
|
92 |
pipe.load_lora_weights(lora_path)
|
93 |
+
#pipe.fuse_lora()
|
94 |
# Set random seed for reproducibility
|
95 |
with calculateDuration("Randomizing seed"):
|
96 |
if randomize_seed:
|
|
|
98 |
|
99 |
image = generate_image(prompt, trigger_word, steps, seed, cfg_scale, width, height, lora_scale, progress)
|
100 |
pipe.to("cpu")
|
101 |
+
#pipe.unfuse_lora()
|
102 |
pipe.unload_lora_weights()
|
103 |
return image, seed
|
104 |
|
105 |
+
def get_huggingface_safetensors(link):
|
106 |
+
split_link = link.split("/")
|
107 |
+
if(len(split_link) == 2):
|
108 |
+
model_card = ModelCard.load(link)
|
109 |
+
image_path = model_card.data.get("widget", [{}])[0].get("output", {}).get("url", None)
|
110 |
+
trigger_word = model_card.data.get("instance_prompt", "")
|
111 |
+
image_url = f"https://huggingface.co/{link}/resolve/main/{image_path}" if image_path else None
|
112 |
+
fs = HfFileSystem()
|
113 |
+
try:
|
114 |
+
list_of_files = fs.ls(link, detail=False)
|
115 |
+
for file in list_of_files:
|
116 |
+
if(file.endswith(".safetensors")):
|
117 |
+
safetensors_name = file.split("/")[-1]
|
118 |
+
if (not image_url and file.lower().endswith((".jpg", ".jpeg", ".png", ".webp"))):
|
119 |
+
image_elements = file.split("/")
|
120 |
+
image_url = f"https://huggingface.co/{link}/resolve/main/{image_elements[-1]}"
|
121 |
+
except Exception as e:
|
122 |
+
print(e)
|
123 |
+
gr.Warning(f"You didn't include a link neither a valid Hugging Face repository with a *.safetensors LoRA {e}")
|
124 |
+
raise Exception(f"You didn't include a link neither a valid Hugging Face repository with a *.safetensors LoRA {e}")
|
125 |
+
return split_link[1], link, safetensors_name, trigger_word, image_url
|
126 |
+
|
127 |
+
def check_custom_model(link):
|
128 |
+
if(link.startswith("https://")):
|
129 |
+
if(link.startswith("https://huggingface.co") or link.startswith("https://www.huggingface.co")):
|
130 |
+
link_split = link.split("huggingface.co/")
|
131 |
+
return get_huggingface_safetensors(link_split[1])
|
132 |
+
else:
|
133 |
+
return get_huggingface_safetensors(link)
|
134 |
+
|
135 |
+
def add_custom_lora(custom_lora):
|
136 |
+
global loras
|
137 |
+
if(custom_lora):
|
138 |
+
try:
|
139 |
+
title, repo, path, trigger_word, image = check_custom_model(custom_lora)
|
140 |
+
card = f'''
|
141 |
+
<div class="custom_lora_card">
|
142 |
+
<span>Loaded custom LoRA:</span>
|
143 |
+
<div class="card_internal">
|
144 |
+
<img src="{image}" />
|
145 |
+
<div>
|
146 |
+
<h3>{title}</h3>
|
147 |
+
<small>{"Using: <code><b>"+trigger_word+"</code></b> as the trigger word" if trigger_word else "No trigger word found. If there's a trigger word, include it in your prompt"}<br></small>
|
148 |
+
</div>
|
149 |
+
</div>
|
150 |
+
</div>
|
151 |
+
'''
|
152 |
+
existing_item_index = next((index for (index, item) in enumerate(loras) if item['repo'] == repo), None)
|
153 |
+
if(not existing_item_index):
|
154 |
+
new_item = {
|
155 |
+
"image": image,
|
156 |
+
"title": title,
|
157 |
+
"repo": repo,
|
158 |
+
"weights": path,
|
159 |
+
"trigger_word": trigger_word
|
160 |
+
}
|
161 |
+
print(new_item)
|
162 |
+
existing_item_index = len(loras)
|
163 |
+
loras.append(new_item)
|
164 |
+
|
165 |
+
return gr.update(visible=True, value=card), gr.update(visible=True), gr.Gallery(selected_index=None), f"Custom: {path}", existing_item_index
|
166 |
+
except Exception as e:
|
167 |
+
gr.Warning(f"Invalid LoRA: either you entered an invalid link, or a non-FLUX LoRA")
|
168 |
+
return gr.update(visible=True, value=f"Invalid LoRA: either you entered an invalid link, a non-FLUX LoRA"), gr.update(visible=True), gr.update(), "", None
|
169 |
+
else:
|
170 |
+
return gr.update(visible=False), gr.update(visible=False), gr.update(), "", None
|
171 |
+
|
172 |
+
def remove_custom_lora():
|
173 |
+
return gr.update(visible=False), gr.update(visible=False), gr.update(), "", None, ""
|
174 |
+
|
175 |
run_lora.zerogpu = True
|
176 |
|
177 |
css = '''
|
|
|
180 |
#title h1{font-size: 3em; display:inline-flex; align-items:center}
|
181 |
#title img{width: 100px; margin-right: 0.5em}
|
182 |
#gallery .grid-wrap{height: 10vh}
|
183 |
+
#lora_list{background: var(--block-background-fill);padding: 0 1em .3em; font-size: 90%}
|
184 |
+
.card_internal{display: flex;height: 100px;margin-top: .5em}
|
185 |
+
.card_internal img{margin-right: 1em}
|
186 |
+
.styler{--form-gap-width: 0px !important}
|
187 |
'''
|
188 |
with gr.Blocks(theme=gr.themes.Soft(), css=css) as app:
|
189 |
title = gr.HTML(
|
|
|
206 |
columns=3,
|
207 |
elem_id="gallery"
|
208 |
)
|
209 |
+
with gr.Group():
|
210 |
+
custom_lora = gr.Textbox(label="Custom LoRA", info="LoRA Hugging Face path", placeholder="multimodalart/vintage-ads-flux")
|
211 |
+
gr.Markdown("[Check the list of FLUX LoRas](https://huggingface.co/models?other=base_model:adapter:black-forest-labs/FLUX.1-dev)", elem_id="lora_list")
|
212 |
+
custom_lora_info = gr.HTML(visible=False)
|
213 |
+
custom_lora_button = gr.Button("Remove custom LoRA", visible=False)
|
214 |
with gr.Column(scale=4):
|
215 |
result = gr.Image(label="Generated Image")
|
216 |
|
|
|
235 |
inputs=[width, height],
|
236 |
outputs=[prompt, selected_info, selected_index, width, height]
|
237 |
)
|
238 |
+
custom_lora.input(
|
239 |
+
add_custom_lora,
|
240 |
+
inputs=[custom_lora],
|
241 |
+
outputs=[custom_lora_info, custom_lora_button, gallery, selected_info, selected_index]
|
242 |
+
)
|
243 |
+
custom_lora_button.click(
|
244 |
+
remove_custom_lora,
|
245 |
+
outputs=[custom_lora_info, custom_lora_button, gallery, selected_info, selected_index, custom_lora]
|
246 |
+
)
|
247 |
gr.on(
|
248 |
triggers=[generate_button.click, prompt.submit],
|
249 |
fn=run_lora,
|