import spaces # type: ignore import os import uuid from PIL import Image import gradio as gr import numpy as np import random import torch from diffusers import FluxPipeline, DiffusionPipeline, FluxTransformer2DModel # noqa: F401 from torchao.quantization import quantize_, int8_weight_only from sd_embed.embedding_funcs import get_weighted_text_embeddings_flux1 dtype = torch.bfloat16 device = "cuda" if torch.cuda.is_available() else "cpu" transformer = FluxTransformer2DModel.from_pretrained( "black-forest-labs/FLUX.1-dev", subfolder="transformer", torch_dtype=torch.bfloat16 ) quantize_(transformer, int8_weight_only()) pipe = DiffusionPipeline.from_pretrained( "black-forest-labs/FLUX.1-dev", transformer=transformer, torch_dtype=torch.bfloat16 ) # pipe = FluxPipeline.from_pretrained( # "black-forest-labs/FLUX.1-dev", # torch_dtype=dtype, # ) pipe.to(device) MAX_SEED = np.iinfo(np.int32).max MAX_IMAGE_SIZE = 2048 @spaces.GPU(duration=90) def infer( prompt: str, seed=42, randomize_seed=False, width=1024, height=1024, guidance_scale=5.0, num_inference_steps=28, progress=gr.Progress(track_tqdm=True), ): if randomize_seed: seed = random.randint(0, MAX_SEED) generator = torch.Generator().manual_seed(seed) prompt_embeds, pooled_prompt_embeds = get_weighted_text_embeddings_flux1( pipe=pipe, prompt=prompt ) image = pipe( prompt_embeds=prompt_embeds, pooled_prompt_embeds=pooled_prompt_embeds, width=width, height=height, num_inference_steps=num_inference_steps, generator=generator, guidance_scale=guidance_scale, ).images[0] assert isinstance( image, Image.Image ), "The output is not an instance of Image.Image" filepath = os.path.join("images", "{uuid}.png".format(uuid=str(uuid.uuid4().hex))) image.save(filepath) return ( image, gr.DownloadButton( label="Download PNG", value=filepath, size="sm", visible=True ), seed, ) examples = [ "a cat holding a sign that says flux.1 is great", "an old man holding a sign that says Increase Zero-GPU Limit", ] 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 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(equal_height=False): with gr.Column(): prompt = gr.TextArea( label="Prompt", show_label=False, lines=3, placeholder="Enter your prompt", container=False, ) run_button = gr.Button("Run", variant="primary", scale=0) result = gr.Image( format="webp", type="pil", label="Result", show_label=False, show_download_button=False, show_share_button=False, ) download = gr.DownloadButton(size="sm", visible=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=832, ) height = gr.Slider( label="Height", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1216, ) with gr.Row(): guidance_scale = gr.Slider( label="Guidance Scale", minimum=0, 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, download, seed], cache_examples="lazy", ) gr.on( triggers=[run_button.click], fn=lambda: gr.update(visible=False), outputs=download, api_name=False, ) gr.on( triggers=[run_button.click, prompt.submit], fn=infer, inputs=[ prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, ], outputs=[result, download, seed], ) if __name__ == "__main__": os.makedirs("images", exist_ok=True) demo.queue(api_open=True).launch()