batuto_manus / app.py
BATUTO90
Initial commit of FLUX app with FLUX.1-schnell-4bit model
415abfa
import gradio as gr
import numpy as np
import random
# import spaces
import torch
from diffusers import DiffusionPipeline, AutoencoderTiny, AutoencoderKL
from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast
# from live_preview_helpers import calculate_shift, retrieve_timesteps, flux_pipe_call_that_returns_an_iterable_of_images
# Importar os para la variable de entorno, aunque no se use directamente aquí, es buena práctica
import os
dtype = torch.bfloat16
device = "cuda" if torch.cuda.is_available() else "cpu"
# Usando el modelo más ligero de la lista blanca: FLUX.1-schnell-4bit
# Nota: La lógica de FLUX es compleja, mantendré la estructura original pero con el modelo más ligero.
try:
# Intentar cargar el modelo más ligero de la lista blanca
model_id = "black-forest-labs/FLUX.1-schnell-4bit"
taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
pipe = DiffusionPipeline.from_pretrained(model_id, torch_dtype=dtype, vae=taef1).to(device)
torch.cuda.empty_cache()
except Exception as e:
# Si falla, se usa un placeholder para que la interfaz Gradio se cargue
print(f"Error al cargar el modelo FLUX: {e}")
pipe = None
MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 512 # Ajustado a un tamaño más seguro
# La función infer debe ser modificada para ser compatible con la nueva Pipeline
# y para manejar el caso de que el modelo no se haya cargado.
def infer(prompt, seed=42, randomize_seed=False, width=512, height=512, guidance_scale=3.5, num_inference_steps=28, progress=gr.Progress(track_tqdm=True)):
if pipe is None:
return gr.Error("El modelo FLUX no pudo cargarse debido a limitaciones de recursos."), seed
if randomize_seed:
seed = random.randint(0, MAX_SEED)
generator = torch.Generator().manual_seed(seed)
# La llamada original usaba una función auxiliar que no tenemos (flux_pipe_call_that_returns_an_iterable_of_images)
# Usaremos la llamada estándar de DiffusionPipeline.
# Nota: La pipeline de FLUX puede requerir parámetros específicos.
# Si la llamada estándar falla, esto requerirá una revisión más profunda.
try:
# Usamos la llamada estándar de DiffusionPipeline
image = pipe(
prompt=prompt,
guidance_scale=guidance_scale,
num_inference_steps=num_inference_steps,
width=width,
height=height,
generator=generator,
output_type="pil",
).images[0]
return image, seed
except Exception as e:
return gr.Error(f"Error durante la inferencia: {e}"), seed
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(f"""# FLUX.1 [dev]
12B param rectified flow transformer guidance-distilled from [FLUX.1 [pro]](https://blackforestlabs.ai/)
[[non-commercial license](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md)] [[blog](https://blackforestlabs.ai/announcing-black-forest-labs/)] [[model](https://huggingface.co/black-forest-labs/FLUX.1-dev)]
""")
with gr.Row():
prompt = gr.Text(
label="Prompt",
show_label=False,
max_lines=1,
placeholder="Enter your prompt",
container=False,
)
run_button = gr.Button("Run", scale=0)
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=512,
)
height = gr.Slider(
label="Height",
minimum=256,
maximum=MAX_IMAGE_SIZE,
step=32,
value=512,
)
with gr.Row():
guidance_scale = gr.Slider(
label="Guidance Scale",
minimum=1,
maximum=15,
step=0.1,
value=3.5,
)
num_inference_steps = gr.Slider(
label="Number of inference steps",
minimum=1,
maximum=50,
step=1,
value=28,
)
gr.Examples(
examples = examples,
fn = infer,
inputs = [prompt],
outputs = [result, seed],
cache_examples="lazy"
)
gr.on(
triggers=[run_button.click, prompt.submit],
fn = infer,
inputs = [prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
outputs = [result, seed]
)
demo.launch()