Spaces:

LWZ19
/

ecodiff_flux_dev

Sleeping

File size: 9,470 Bytes

import copy
import gradio as gr
import numpy as np
import random
import pickle
import torch
import os
import sys
import spaces
from huggingface_hub import hf_hub_download, snapshot_download
from diffusers import FluxPipeline
from diffusers.models import FluxTransformer2DModel
from diffusers.utils import SAFETENSORS_WEIGHTS_NAME
from diffusers.loaders.lora_base import LORA_WEIGHT_NAME_SAFE
from safetensors.torch import load_file

# Import essential classes for unpickling pruned models
from utils import SparsityLinear, SkipConnection, AttentionSkipConnection

# Create a simple mock module for pickle imports
class MockModule:
    def __init__(self):
        # Add all the classes that pickle might need
        self.SparsityLinear = SparsityLinear
        self.SkipConnection = SkipConnection
        self.AttentionSkipConnection = AttentionSkipConnection
        # Self-reference for nested imports
        self.utils = self

# Register the mock module for all sdib import paths
mock = MockModule()
sys.modules['sdib'] = mock
sys.modules['sdib.utils'] = mock
sys.modules['sdib.utils.utils'] = mock


################################################################################
################################################################################


# Configuration
PRUNING_RATIOS = [10, 15, 20]

device = "cuda" if torch.cuda.is_available() else "cpu"
MAX_SEED = np.iinfo(np.int32).max
dtype = torch.bfloat16

print("🚀 Loading base Flux dev pipeline...")
base_pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev",
    torch_dtype=dtype
)
print("✅ Base Flux dev pipeline loaded!")

# Global storage for all models
pruned_models = {}
lora_weights = None

print("📥 Preloading all pruned models...")
for ratio in PRUNING_RATIOS:
    try:
        print(f"Loading {ratio}% pruned model...")
        model_file = hf_hub_download(
            repo_id="LWZ19/ecodiff_flux_prune",
            filename=f"dev/pruned_model_{ratio}.pkl"
        )
        
        with open(model_file, "rb") as f:
            pruned_model = pickle.load(f)
        pruned_model.to("cpu")
        pruned_model.to(dtype)
        
        pruned_models[ratio] = pruned_model
        print(f"✅ {ratio}% pruned model loaded!")
    except Exception as e:
        print(f"❌ Failed to load {ratio}% pruned model: {e}")
        pruned_models[ratio] = None

print("📥 Preloading LoRA checkpoint for 20% pruning ratio...")
try:
    lora_repo_path = snapshot_download(
        repo_id="LWZ19/ecodiff_flux_retrain_weights",
        allow_patterns=[f"dev/lora/prune_20/*"]
    )
    lora_weights = load_file(os.path.join(lora_repo_path, "dev", "lora", "prune_20", LORA_WEIGHT_NAME_SAFE))
    print("✅ LoRA checkpoint loaded!")
except Exception as e:
    print(f"❌ Failed to load LoRA checkpoint: {e}")
    lora_weights = None

    
# Model state
base_pipe.transformer = pruned_models[10].to(device)
current_ratio = 10


def load_model(ratio, use_lora=False):
    """Apply specified model to the pipeline with optional LoRA"""
    global current_ratio

    try:
        # Switch to new pruned model if different ratio
        if current_ratio != ratio:
            base_pipe.transformer = pruned_models[ratio].to(device)
            current_ratio = ratio
        
        # Handle LoRA loading for 20% ratio
        if ratio == 20 and use_lora and lora_weights is not None:
            base_pipe.load_lora_weights(lora_weights)
            return f"✅ Ready with {ratio}% pruned Flux.1 [dev] + LoRA retrained"
        elif ratio == 20 and use_lora and lora_weights is None:
            return f"❌ LoRA weights not available for {ratio}% model"
        else:
            return f"✅ Ready with {ratio}% pruned Flux.1 [dev] (no retraining)"

    except Exception as e:
        return f"❌ Failed to apply weights: {str(e)}"


@spaces.GPU(duration=99)
def generate_image(
    ratio,
    prompt,
    seed,
    randomize_seed,
    width,
    height,
    guidance_scale,
    num_inference_steps,
    use_lora=False,
    progress=gr.Progress(track_tqdm=True),
): 
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    try:
        # Apply model configuration
        status = load_model(ratio, use_lora)
        if "❌" in status:
            return None, seed, status
        
        # Move pipeline to GPU for generation
        base_pipe.to(device)
        
        generator = torch.Generator(device).manual_seed(seed)

        # Generate image using base pipeline (already configured with pruned model)
        image = base_pipe(
            prompt=prompt,
            guidance_scale=guidance_scale,
            num_inference_steps=num_inference_steps,
            width=width,
            height=height,
            generator=generator,
        ).images[0]
        
        # Clean up GPU memory
        torch.cuda.empty_cache() if torch.cuda.is_available() else None
        
        if ratio == 20 and use_lora and lora_weights is not None:
            base_pipe.unload_lora_weights()
            result_status = f"✅ Generated with {ratio}% pruned Flux.1 [dev] + LoRA retrained"
        else:
            result_status = f"✅ Generated with {ratio}% pruned Flux.1 [dev]"

        return image, seed, result_status
        
    except Exception as e:
        error_status = f"❌ Generation failed: {str(e)}\nPlease retry after a few minutes."
        return None, seed, error_status

examples = [
    "A clock tower floating in a sea of clouds",
    "A cozy library with a roaring fireplace", 
    "A cat playing football",
    "A magical forest with glowing mushrooms",
    "An astronaut riding a rainbow unicorn",
]

css = """
#col-container {
    margin: 0 auto;
    max-width: 720px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown("# EcoDiff Flux.1 [dev]: Memory-Efficient Diffusion")
        gr.Markdown("Generate images using pruned Flux.1 [dev] models with multiple pruning ratios. For 20% pruning, optional LoRA retrained weights are available.")

        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )

        with gr.Row():
            ratio = gr.Dropdown(
                choices=PRUNING_RATIOS,
                value=10,
                label="Pruning Ratio (%)",
                info="Select pruning ratio",
                scale=1
            )
            
        with gr.Row(visible=False) as lora_row:
            use_lora = gr.Checkbox(
                label="Use LoRA Retrained Model",
                value=False,
                info="Enable LoRA fine-tuned weights (only available for 20% pruning)"
            )

        generate_button = gr.Button("Generate", variant="primary")
        result = gr.Image(label="Result", show_label=False)
        status_display = gr.Textbox(label="Status", interactive=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=512,
                    maximum=2048,
                    step=32,
                    value=1024,
                )

                height = gr.Slider(
                    label="Height", 
                    minimum=512,
                    maximum=2048,
                    step=32,
                    value=1024,
                )

            with gr.Row():
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=1.0,
                    maximum=10.0,
                    step=0.1,
                    value=3.5,
                )
                                
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=50,
                    step=1,
                    value=50,
                )

        gr.Examples(examples=examples, inputs=[prompt])
        
        gr.Markdown("""
        ### About EcoDiff Flux.1 [dev] Unified
        This space showcases multiple pruned Flux.1 [dev] models using learnable pruning techniques with optional LoRA fine-tuning.
        
        - **Base Model**: Flux.1 [dev]
        - **Pruning Ratios**: 10%, 15%, 20% of parameters removed
        - **LoRA Enhancement**: Available for 20% pruning ratio with retrained weights for improved quality
        """)

    def update_lora_visibility(ratio_value):
        return gr.update(visible=(ratio_value == 20))
    
    ratio.change(
        fn=update_lora_visibility,
        inputs=[ratio],
        outputs=[lora_row]
    )
    
    generate_button.click(
        fn=generate_image,
        inputs=[
            ratio,
            prompt,
            seed,
            randomize_seed,
            width,
            height,
            guidance_scale,
            num_inference_steps,
            use_lora,
        ],
        outputs=[result, seed, status_display],
    )

if __name__ == "__main__":
    demo.launch()