Spaces:
Runtime error
Runtime error
File size: 7,105 Bytes
966b11e 4f41a04 966b11e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 | import random
import torch
import gradio as gr
from diffusers import (
FluxPipeline,
DPMSolverMultistepScheduler,
DPMSolverSDEScheduler,
EulerDiscreteScheduler,
EulerAncestralDiscreteScheduler,
HeunDiscreteScheduler,
DDIMScheduler,
)
# -----------------------------------------------------------------------------
# Pipeline loading helpers
# -----------------------------------------------------------------------------
def _load_pipe(hf_token: str | None = None) -> FluxPipeline:
"""Load the FLUX pipeline once and keep it in memory.
Args:
hf_token: Optional Hugging Face token if the model is gated/private.
Returns:
A fully‑initialised FluxPipeline with LoRA fused and memory‑saving
features enabled.
"""
pipe = FluxPipeline.from_pretrained(
"black-forest-labs/FLUX.1-dev",
torch_dtype=torch.float16,
use_auth_token=hf_token or None,
)
# Memory optimisations ----------------------------------------------------
pipe.enable_sequential_cpu_offload()
pipe.enable_attention_slicing()
# LoRA --------------------------------------------------------------------
pipe.load_lora_weights(
"kudzueye/boreal-flux-dev-v2", weight_name="boreal-v2.safetensors"
)
pipe.fuse_lora(lora_scale=0.8)
return pipe
# Keep a single global instance to avoid re‑loading on every request
_pipe: FluxPipeline | None = None
def _get_pipe(hf_token: str | None = None) -> FluxPipeline:
global _pipe
if _pipe is None:
_pipe = _load_pipe(hf_token)
return _pipe
# -----------------------------------------------------------------------------
# Scheduler mapping
# -----------------------------------------------------------------------------
SCHED_MAP = {
"DPM++ 2M Karras": DPMSolverMultistepScheduler,
"DPM++ SDE Karras": DPMSolverSDEScheduler,
"Euler": EulerDiscreteScheduler,
"Euler a": EulerAncestralDiscreteScheduler,
"Heun": HeunDiscreteScheduler,
"DDIM": DDIMScheduler,
}
# -----------------------------------------------------------------------------
# Inference function
# -----------------------------------------------------------------------------
def query(
prompt: str,
negative_prompt: str,
steps: int,
cfg_scale: float,
sampler: str,
seed: int,
strength: float, # kept for future img2img support
hf_token: str,
):
"""Run the generation and return a PIL image + the seed actually used."""
pipe = _get_pipe(hf_token or None)
# Replace scheduler if the user selected a different sampler
SchedulerCls = SCHED_MAP.get(sampler, DPMSolverMultistepScheduler)
if not isinstance(pipe.scheduler, SchedulerCls):
pipe.scheduler = SchedulerCls.from_config(pipe.scheduler.config)
# Handle seed
if seed == -1:
seed = random.randint(0, 1_000_000_000)
generator = torch.Generator(device=pipe.device).manual_seed(seed)
# Run inference
with torch.no_grad():
result = pipe(
prompt=prompt,
negative_prompt=negative_prompt,
num_inference_steps=steps,
guidance_scale=cfg_scale,
generator=generator,
height=512,
width=512,
)
return result.images[0], str(seed)
# -----------------------------------------------------------------------------
# Gradio UI
# -----------------------------------------------------------------------------
CSS = """
#app-container {
max-width: 600px;
margin-left: auto;
margin-right: auto;
}
#title-container {
display: flex;
align-items: center;
justify-content: center;
}
#title-icon {
width: 32px;
height: auto;
margin-right: 10px;
}
#title-text {
font-size: 24px;
font-weight: bold;
}
"""
with gr.Blocks(theme="Nymbo/Nymbo_Theme", css=CSS) as app:
gr.HTML(
"""
<center>
<div id="title-container">
<h1 id="title-text">Text-to-Image Generator App</h1>
</div>
</center>
"""
)
with gr.Column(elem_id="app-container"):
with gr.Row():
with gr.Column(elem_id="prompt-container"):
with gr.Row():
txt_prompt = gr.Textbox(
label="Prompt",
placeholder="Enter a prompt here",
lines=2,
elem_id="prompt-text-input",
)
with gr.Row():
with gr.Accordion("Advanced Settings", open=False):
neg_prompt = gr.Textbox(
label="Negative Prompt",
placeholder="What should not be in the image",
value="(deformed, distorted, disfigured), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation, misspellings, typos",
lines=3,
elem_id="negative-prompt-text-input",
)
steps_in = gr.Slider(
label="Sampling steps", value=35, minimum=1, maximum=100, step=1
)
cfg_in = gr.Slider(
label="CFG Scale", value=7, minimum=1, maximum=20, step=1
)
sampler_in = gr.Radio(
label="Sampling method",
value="DPM++ 2M Karras",
choices=list(SCHED_MAP.keys()),
)
strength_in = gr.Slider(
label="Strength", value=0.7, minimum=0, maximum=1, step=0.001
)
seed_in = gr.Slider(
label="Seed", value=-1, minimum=-1, maximum=1_000_000_000, step=1
)
api_key_in = gr.Textbox(
label="Hugging Face API Key (required for private models)",
placeholder="Enter your Hugging Face API Key here",
type="password",
elem_id="api-key",
)
with gr.Row():
run_button = gr.Button("Run", variant="primary", elem_id="gen-button")
with gr.Row():
img_out = gr.Image(type="pil", label="Image Output", elem_id="gallery")
seed_out = gr.Textbox(label="Seed Used", elem_id="seed-output")
run_button.click(
fn=query,
inputs=[
txt_prompt,
neg_prompt,
steps_in,
cfg_in,
sampler_in,
seed_in,
strength_in,
api_key_in,
],
outputs=[img_out, seed_out],
)
app.launch(show_api=True, share=False)
|