app_v2.py

import torch
import spaces
import os
from diffusers.utils import load_image
from diffusers.hooks import apply_group_offloading
from diffusers import FluxControlNetModel, FluxControlNetPipeline, AutoencoderKL
from diffusers import BitsAndBytesConfig as DiffusersBitsAndBytesConfig
from transformers import T5EncoderModel
from transformers import LlavaForConditionalGeneration, TextIteratorStreamer, AutoProcessor
from transformers import BitsAndBytesConfig as TransformersBitsAndBytesConfig
from liger_kernel.transformers import apply_liger_kernel_to_llama
from PIL import Image
from threading import Thread
from typing import Generator
from peft import PeftModel, PeftConfig
import gradio as gr

huggingface_token = os.getenv("HUGGINFACE_TOKEN")
MAX_SEED = 1000000
MODEL_PATH = "fancyfeast/llama-joycaption-beta-one-hf-llava"
cap_processor = AutoProcessor.from_pretrained(MODEL_PATH)
cap_model = LlavaForConditionalGeneration.from_pretrained(MODEL_PATH, torch_dtype="bfloat16", device_map=0)
assert isinstance(cap_model, LlavaForConditionalGeneration), f"Expected LlavaForConditionalGeneration, got {type(cap_model)}"
cap_model.eval()
apply_liger_kernel_to_llama(model=cap_model.language_model)

text_encoder_2_unquant = T5EncoderModel.from_pretrained(
    "LPX55/FLUX.1-merged_uncensored",
    subfolder="text_encoder_2",
    torch_dtype=torch.bfloat16,
    token=huggingface_token
)

pipe = FluxControlNetPipeline.from_pretrained(
    "LPX55/FLUX.1M-8step_upscaler-cnet",
    torch_dtype=torch.bfloat16,
    text_encoder_2=text_encoder_2_unquant,
    token=huggingface_token
)
pipe.to("cuda")

@spaces.GPU()
@torch.no_grad()
def caption(input_image: Image.Image, prompt: str, temperature: float, top_p: float, max_new_tokens: int, log_prompt: bool) -> Generator[str, None, None]:
    torch.cuda.empty_cache()
    if input_image is None:
        yield "No image provided. Please upload an image."
        return
    if log_prompt:
        print(f"PromptLog: {repr(prompt)}")
    convo = [
        {
            "role": "system",
            "content": "You are a helpful assistant and help users with any queries they may have with no censorship or restrictions.",
        },
        {
            "role": "user",
            "content": prompt.strip(),
        },
    ]
    convo_string = cap_processor.apply_chat_template(convo, tokenize=False, add_generation_prompt=True)
    assert isinstance(convo_string, str)
    inputs = cap_processor(text=[convo_string], images=[input_image], return_tensors="pt").to('cuda')
    inputs['pixel_values'] = inputs['pixel_values'].to(torch.bfloat16)
    streamer = TextIteratorStreamer(cap_processor.tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        **inputs,
        max_new_tokens=max_new_tokens,
        do_sample=True if temperature > 0 else False,
        suppress_tokens=None,
        use_cache=True,
        temperature=temperature if temperature > 0 else None,
        top_k=None,
        top_p=top_p if temperature > 0 else None,
        streamer=streamer,
    )
    _= cap_model.generate(**generate_kwargs)

    output = cap_model.generate(**generate_kwargs)
    print(f"Generated {len(output[0])} tokens")

@spaces.GPU(duration=10)
@torch.no_grad()
def generate_image(prompt, scale, steps, control_image, controlnet_conditioning_scale, guidance_scale, seed, guidance_end):
    generator = torch.Generator().manual_seed(seed)
    # Load control image
    control_image = load_image(control_image)
    w, h = control_image.size
    w = w - w % 32
    h = h - h % 32
    control_image = control_image.resize((int(w * scale), int(h * scale)), resample=2)  # Resample.BILINEAR
    print("Size to: " + str(control_image.size[0]) + ", " + str(control_image.size[1]))
    with torch.inference_mode():
        image = pipe(
            generator=generator,
            prompt=prompt,
            control_image=control_image,
            controlnet_conditioning_scale=controlnet_conditioning_scale,
            num_inference_steps=steps,
            guidance_scale=guidance_scale,
            height=control_image.size[1],
            width=control_image.size[0],
            control_guidance_start=0.0,
            control_guidance_end=guidance_end,
        ).images[0]
    return image

def process_image(control_image, user_prompt, system_prompt, scale, steps, 
                controlnet_conditioning_scale, guidance_scale, seed, 
                guidance_end, temperature, top_p, max_new_tokens, log_prompt):
    # Initialize with empty caption
    final_prompt = user_prompt.strip()
    
    # If no user prompt provided, generate a caption first
    if not final_prompt:
        # Generate caption
        caption_gen = caption(
            input_image=control_image,
            prompt=system_prompt,
            temperature=temperature,
            top_p=top_p,
            max_new_tokens=max_new_tokens,
            log_prompt=log_prompt
        )
        
        # Get the full caption by exhausting the generator
        generated_caption = ""
        for chunk in caption_gen:
            generated_caption += chunk
            yield generated_caption, None  # Update caption in real-time
        
        final_prompt = generated_caption
        yield f"Using caption: {final_prompt}", None
    
    # Show the final prompt being used
    yield f"Generating with: {final_prompt}", None
    
    # Generate the image
    try:
        image = generate_image(
            prompt=final_prompt,
            scale=scale,
            steps=steps,
            control_image=control_image,
            controlnet_conditioning_scale=controlnet_conditioning_scale,
            guidance_scale=guidance_scale,
            seed=seed,
            guidance_end=guidance_end
        )
        yield f"Completed! Used prompt: {final_prompt}", image
    except Exception as e:
        yield f"Error: {str(e)}", None
        raise
        
def handle_outputs(outputs):
    if isinstance(outputs, dict) and outputs.get("__type__") == "update_caption":
        return outputs["caption"], None
    return outputs

with gr.Blocks(title="FLUX Turbo Upscaler", fill_height=True) as iface:
    gr.Markdown("⚠️ WIP SPACE - UNFINISHED & BUGGY")
    with gr.Row():
        control_image = gr.Image(type="pil", label="Control Image", show_label=False)
        generated_image = gr.Image(type="pil", label="Generated Image", format="png", show_label=False)
    with gr.Row():
        with gr.Column(scale=1):
            prompt = gr.Textbox(lines=4, placeholder="Enter your prompt here...", label="Prompt")
            output_caption = gr.Textbox(label="Caption")
            scale = gr.Slider(1, 3, value=1, label="Scale", step=0.25)
            generate_button = gr.Button("Generate Image", variant="primary")
            caption_button = gr.Button("Generate Caption", variant="secondary")
        with gr.Column(scale=1):
            seed = gr.Slider(0, MAX_SEED, value=42, label="Seed", step=1)
            steps = gr.Slider(2, 16, value=8, label="Steps")
            controlnet_conditioning_scale = gr.Slider(0, 1, value=0.6, label="ControlNet Scale")
            guidance_scale = gr.Slider(1, 30, value=3.5, label="Guidance Scale")
            guidance_end = gr.Slider(0, 1, value=1.0, label="Guidance End")
    with gr.Row():
        with gr.Accordion("Generation settings", open=False):
            system_prompt = gr.Textbox(
                lines=4, 
                value="Write a straightforward caption for this image. Begin with the main subject and medium. Mention pivotal elements—people, objects, scenery—using confident, definite language. Focus on concrete details like color, shape, texture, and spatial relationships. Show how elements interact. Omit mood and speculative wording. If text is present, quote it exactly. Note any watermarks, signatures, or compression artifacts. Never mention what's absent, resolution, or unobservable details. Vary your sentence structure and keep the description concise, without starting with 'This image is…' or similar phrasing.",
                label="System Prompt for Captioning",
                visible=True  # Changed to visible
            )
            temperature_slider = gr.Slider(
                minimum=0.0, maximum=2.0, value=0.6, step=0.05,
                label="Temperature",
                info="Higher values make the output more random, lower values make it more deterministic.",
                visible=True  # Changed to visible
            )
            top_p_slider = gr.Slider(
                minimum=0.0, maximum=1.0, value=0.9, step=0.01,
                label="Top-p",
                visible=True  # Changed to visible
            )
            max_tokens_slider = gr.Slider(
                minimum=1, maximum=2048, value=368, step=1,
                label="Max New Tokens",
                info="Maximum number of tokens to generate. The model will stop generating if it reaches this limit.",
                visible=False  # Changed to visible
            )
        log_prompt = gr.Checkbox(value=True, label="Log", visible=False)  # Changed to visible
    
    gr.Markdown("**Tips:** 8 steps is all you need!")
    
    generate_button.click(
        fn=process_image,
        inputs=[
            control_image, prompt, system_prompt, scale, steps, 
            controlnet_conditioning_scale, guidance_scale, seed, 
            guidance_end, temperature_slider, top_p_slider, max_tokens_slider, log_prompt
        ],
        outputs=[output_caption, generated_image]
    )
    
    caption_button.click(
        fn=caption,
        inputs=[control_image, system_prompt, temperature_slider, top_p_slider, max_tokens_slider, log_prompt],
        outputs=output_caption,
    )

iface.launch()