FLUX.1-schnell / app.py
hysts's picture
hysts HF Staff
gradio==5.33.0
5a3d18f
import gradio as gr
import numpy as np
import PIL.Image
import spaces
import torch
from diffusers import DiffusionPipeline
dtype = torch.bfloat16
device = "cuda" if torch.cuda.is_available() else "cpu"
pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-schnell", torch_dtype=dtype).to(device)
MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 2048
def get_seed(randomize_seed: bool, seed: int) -> int:
"""Determine and return the random seed to use for model generation.
- MAX_SEED is the maximum value for a 32-bit integer (np.iinfo(np.int32).max).
- This function is typically used to ensure reproducibility or to introduce randomness in model generation.
Args:
randomize_seed (bool): If True, a random seed (an integer in [0, MAX_SEED)) is generated using NumPy's default random number generator. If False, the provided seed argument is returned as-is.
seed (int): The seed value to use if randomize_seed is False.
Returns:
int: The selected seed value. If randomize_seed is True, a randomly generated integer; otherwise, the value of the seed argument.
"""
rng = np.random.default_rng()
return int(rng.integers(0, MAX_SEED)) if randomize_seed else seed
@spaces.GPU
def infer(
prompt: str,
seed: int,
width: int = 1024,
height: int = 1024,
num_inference_steps: int = 4,
progress: gr.Progress = gr.Progress(track_tqdm=True), # noqa: ARG001, B008
) -> PIL.Image.Image:
"""Generate an image from a text prompt using the FLUX.1 [schnell] model.
- Prompts must be in English. Other languages are not currently supported.
- Prompts are limited to a maximum of 77 tokens, due to the CLIP tokenizer constraint.
Args:
prompt: A text prompt in English used to guide the image generation. Limited to 77 tokens.
seed: The seed used for deterministic random number generation.
width: Width of the generated image in pixels. Defaults to 1024.
height: Height of the generated image in pixels. Defaults to 1024.
num_inference_steps: Number of inference steps to perform. A higher value may improve image quality. Defaults to 4.
progress: (Internal) Used to display progress in the UI; should not be modified by the user.
Returns:
A PIL.Image.Image object representing the generated image.
"""
generator = torch.Generator().manual_seed(seed)
return pipe(
prompt=prompt,
width=width,
height=height,
num_inference_steps=num_inference_steps,
generator=generator,
guidance_scale=0.0,
).images[0]
def run_example(prompt: str) -> tuple[PIL.Image.Image, int]:
return infer(prompt, seed=42)
examples = [
"a tiny astronaut hatching from an egg on the moon",
"a cat holding a sign that says hello world",
"an anime illustration of a wiener schnitzel",
]
css = """
#col-container {
margin: 0 auto;
max-width: 520px;
}
"""
with gr.Blocks(css=css) as demo:
with gr.Column(elem_id="col-container"):
gr.Markdown("""# FLUX.1 [schnell]
12B param rectified flow transformer distilled from [FLUX.1 [pro]](https://blackforestlabs.ai/) for 4 step generation
[[blog](https://blackforestlabs.ai/announcing-black-forest-labs/)] [[model](https://huggingface.co/black-forest-labs/FLUX.1-schnell)]
""")
with gr.Row():
prompt = gr.Text(
label="Prompt",
show_label=False,
max_lines=1,
placeholder="Enter your prompt",
submit_btn=True,
)
result = gr.Image(label="Result", show_label=False)
with gr.Accordion("Advanced Settings", open=False):
seed = gr.Slider(
label="Seed",
minimum=0,
maximum=MAX_SEED,
step=1,
value=0,
)
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
with gr.Row():
width = gr.Slider(
label="Width",
minimum=256,
maximum=MAX_IMAGE_SIZE,
step=32,
value=1024,
)
height = gr.Slider(
label="Height",
minimum=256,
maximum=MAX_IMAGE_SIZE,
step=32,
value=1024,
)
with gr.Row():
num_inference_steps = gr.Slider(
label="Number of inference steps",
minimum=1,
maximum=50,
step=1,
value=4,
)
gr.Examples(
examples=examples,
fn=run_example,
inputs=prompt,
outputs=result,
)
prompt.submit(
fn=get_seed,
inputs=[randomize_seed, seed],
outputs=seed,
).then(
fn=infer,
inputs=[prompt, seed, width, height, num_inference_steps],
outputs=result,
)
if __name__ == "__main__":
demo.launch(mcp_server=True)