app_lora_cpu.py

import spaces
import torch
from diffusers.pipelines.wan.pipeline_wan_i2v import WanImageToVideoPipeline
from diffusers.models.transformers.transformer_wan import WanTransformer3DModel
from diffusers.utils.export_utils import export_to_video
import gradio as gr
import tempfile
import numpy as np
from PIL import Image
import random
import gc
import os

from torchao.quantization import quantize_
from torchao.quantization import Float8DynamicActivationFloat8WeightConfig
from torchao.quantization import Int8WeightOnlyConfig

import aoti


MODEL_ID = "Wan-AI/Wan2.2-I2V-A14B-Diffusers"

MAX_DIM = 832
MIN_DIM = 480
SQUARE_DIM = 640
MULTIPLE_OF = 16

MAX_SEED = np.iinfo(np.int32).max

FIXED_FPS = 16
MIN_FRAMES_MODEL = 8
MAX_FRAMES_MODEL = 80

MIN_DURATION = round(MIN_FRAMES_MODEL/FIXED_FPS,1)
MAX_DURATION = round(MAX_FRAMES_MODEL/FIXED_FPS,1)


pipe = WanImageToVideoPipeline.from_pretrained(MODEL_ID,
    transformer=WanTransformer3DModel.from_pretrained('cbensimon/Wan2.2-I2V-A14B-bf16-Diffusers',
        subfolder='transformer',
        torch_dtype=torch.bfloat16,
        device_map='cuda',
    ),
    transformer_2=WanTransformer3DModel.from_pretrained('cbensimon/Wan2.2-I2V-A14B-bf16-Diffusers',
        subfolder='transformer_2',
        torch_dtype=torch.bfloat16,
        device_map='cuda',
    ),
    torch_dtype=torch.bfloat16,
).to('cuda')

# 加载并融合你的LoRA模型
#weight_name="Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16.safetensors",
pipe.load_lora_weights(
    "Kijai/WanVideo_comfy",
    weight_name="LoRAs/Wan22_Lightx2v/Wan_2_2_I2V_A14B_HIGH_lightx2v_4step_lora_v1030_rank_64_bf16.safetensors",
    adapter_name="lightx2v"
)
kwargs_lora = {}
kwargs_lora["load_into_transformer_2"] = True

#weight_name="Lightx2v/lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16.safetensors",
pipe.load_lora_weights(
    "Kijai/WanVideo_comfy",
    weight_name="LoRAs/Wan22-Lightning/old/Wan2.2-Lightning_I2V-A14B-4steps-lora_LOW_fp16.safetensors",
    adapter_name="lightx2v_2", **kwargs_lora
)



# 新增：加载你提供的high noise LoRA

pipe.load_lora_weights(
    "rahul7star/wan2.2Lora",
    weight_name="DR34ML4Y_I2V_14B_HIGH.safetensors",
    adapter_name="high_noise_lora",
    token=os.environ.get("HF_TOKEN")
)
# 新增：加载你提供的low noise LoRA
pipe.load_lora_weights(
    "rahul7star/wan2.2Lora",
    weight_name="DR34ML4Y_I2V_14B_LOW.safetensors",
    adapter_name="low_noise_lora",
    token=os.environ.get("HF_TOKEN"),
    load_into_transformer_2=True
)


## 2 attempt
pipe.load_lora_weights(
    "rahul7star/wan2.2Lora",
    weight_name="wan2.2_i2v_highnoise_pov_missionary_v1.0.safetensors",
    adapter_name="high_noise_lora1",
    token=os.environ.get("HF_TOKEN")
)
# 新增：加载你提供的low noise LoRA
pipe.load_lora_weights(
    "rahul7star/wan2.2Lora",
    weight_name="wan2.2_i2v_lownoise_pov_missionary_v1.0.safetensors",
    adapter_name="low_noise_lora1",
    token=os.environ.get("HF_TOKEN"),
    load_into_transformer_2=True
)



# pipe.set_adapters(["lightx2v", "lightx2v_2", "high_noise_lora", "low_noise_lora","high_noise_lora1", "low_noise_lora1","high_noise_lora2", "low_noise_lora2"], adapter_weights=[1., 1., 1., 1.,1.,1.,1.,1.])
# # 修改了lora_scale
# pipe.fuse_lora(adapter_names=["lightx2v", "high_noise_lora","high_noise_lora1","high_noise_lora2"], lora_scales=[3.0, 3.0,3.0,1.0], components=["transformer"])
# # 修改了lora_scale
# pipe.fuse_lora(adapter_names=["lightx2v_2", "low_noise_lora","low_noise_lora1","low_noise_lora2"], lora_scales=[1.0, 1.0,1.0,1.0], components=["transformer_2"])


###### use this for 3rd Lora

# # ## 3rd
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora",
#     weight_name="Wan2.2-Doggy_high_noise.safetensors",
#     adapter_name="high_noise_lora2",
#     token=os.environ.get("HF_TOKEN")
# )
# # 新增：加载你提供的low noise LoRA
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora",
#     weight_name="Wan2.2-Doggy_low_noise.safetensors",
#     adapter_name="low_noise_lora2",
#     token=os.environ.get("HF_TOKEN"),
#     load_into_transformer_2=True
# )


# pipe.set_adapters(["lightx2v", "lightx2v_2", "high_noise_lora", "low_noise_lora","high_noise_lora1", "low_noise_lora1","high_noise_lora2", "low_noise_lora2"], adapter_weights=[1., 1., 1., 1.,1.,1.,1.,1.])
# # 修改了lora_scale
# pipe.fuse_lora(adapter_names=["lightx2v", "high_noise_lora","high_noise_lora1","high_noise_lora2"], lora_scales=[3.0, 3.0,3.0,3.0], components=["transformer"])
# # 修改了lora_scale
# pipe.fuse_lora(adapter_names=["lightx2v_2", "low_noise_lora","low_noise_lora1","low_noise_lora2"], lora_scales=[1.0, 1.0,1.0,1.0], components=["transformer_2"])

# #### 3rd lora ends @######



pipe.set_adapters(["lightx2v", "lightx2v_2", "high_noise_lora", "low_noise_lora","high_noise_lora1", "low_noise_lora1"], adapter_weights=[1.5, 1., 1., 1.,1.,1.])
# 修改了lora_scale
pipe.fuse_lora(adapter_names=["lightx2v", "high_noise_lora","high_noise_lora1"], lora_scales=[3.0, 3.0,3.0], components=["transformer"])
# 修改了lora_scale
pipe.fuse_lora(adapter_names=["lightx2v_2", "low_noise_lora","low_noise_lora1"], lora_scales=[1.0, 1.0,1.0], components=["transformer_2"])



########testing all. 4 together


# 原始 v8normal LoRA

# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="DR34ML4Y_I2V_14B_HIGH.safetensors", adapter_name="high_noise_lora", token=os.environ.get("HF_TOKEN")
# )
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="DR34ML4Y_I2V_14B_LOW.safetensors", adapter_name="low_noise_lora", token=os.environ.get("HF_TOKEN"), load_into_transformer_2=True
# )
# # dremal LoRA
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="wan2.2_i2v_highnoise_pov_missionary_v1.0.safetensors", adapter_name="high_dremal_lora", token=os.environ.get("HF_TOKEN")
# )
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="wan2.2_i2v_lownoise_pov_missionary_v1.0.safetensors", adapter_name="low_dremal_lora", token=os.environ.get("HF_TOKEN"), load_into_transformer_2=True
# )
# # missimd LoRA
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="Wan2.2-Doggy_high_noise.safetensors", adapter_name="high_missimd_lora", token=os.environ.get("HF_TOKEN")
# )
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="Wan2.2-Doggy_low_noise.safetensors", adapter_name="low_missimd_lora", token=os.environ.get("HF_TOKEN"), load_into_transformer_2=True
# )
# # ultrade LoRA
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="WAN-2.2-I2V-POV-Titfuck-Paizuri-HIGH-v1.0.safetensors", adapter_name="high_ultrade_lora", token=os.environ.get("HF_TOKEN")
# )
# pipe.load_lora_weights(
#     "rahul7star/wan2.2Lora", weight_name="WAN-2.2-I2V-POV-Titfuck-Paizuri-LOW-v1.0.safetensors", adapter_name="low_ultrade_lora", token=os.environ.get("HF_TOKEN"), load_into_transformer_2=True
# )

# # 设置所有 LoRA 权重
# pipe.set_adapters(
#     [
#         "lightx2v", "lightx2v_2", 
#         "high_noise_lora", "low_noise_lora", 
#         "high_dremal_lora", "low_dremal_lora",
#         "high_missimd_lora", "low_missimd_lora", 
#         "high_ultrade_lora", "low_ultrade_lora" 
#     ],
#     adapter_weights=[1.7, 1.5, 0.4, 0.4, 0, 0, 0.7, 0.7, 0.4, 0.4] 
# )
# # 融合 LoRA 到 transformer
# pipe.fuse_lora(
#     adapter_names=[
#         "lightx2v", 
#         "high_noise_lora", 
#         "high_dremal_lora", 
#         "high_missimd_lora", 
#         "high_ultrade_lora" 
#     ], 
#     lora_scales=[4.0, 3.0, 2.0, 2.0, 2.0], 
#     components=["transformer"]
# )
# # 融合 LoRA 到 transformer_2
# pipe.fuse_lora(
#     adapter_names=[
#         "lightx2v_2", 
#         "low_noise_lora", 
#         "low_dremal_lora", 
#         "low_missimd_lora", 
#         "low_ultrade_lora" 
#     ], 
#     lora_scales=[2.0, 1.5, 1.0, 1.0, 1.0], 
#     components=["transformer_2"]
# )


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






pipe.unload_lora_weights()

quantize_(pipe.text_encoder, Int8WeightOnlyConfig())
quantize_(pipe.transformer, Float8DynamicActivationFloat8WeightConfig())
quantize_(pipe.transformer_2, Float8DynamicActivationFloat8WeightConfig())

aoti.aoti_blocks_load(pipe.transformer, 'rahul7star/WanAot', variant='fp8da')
aoti.aoti_blocks_load(pipe.transformer_2, 'rahul7star/WanAot', variant='fp8da')


default_prompt_i2v = "make this image come alive, cinematic motion, smooth animation"
default_negative_prompt = "色调艳丽, 过曝, 静态, 细节模糊不清, 字幕, 风格, 作品, 画作, 画面, 静止, 整体发灰, 最差质量, 低质量, JPEG压缩残留, 丑陋的, 残缺的, 多余的手指, 画得不好的手部, 画得不好的脸部, 畸形的, 毁容的, 形态畸形的肢体, 手指融合, 静止不动的画面, 杂乱的背景, 三条腿, 背景人很多, 倒着走"

def resize_image(image: Image.Image) -> Image.Image:
    """
    Resizes an image to fit within the model's constraints, preserving aspect ratio as much as possible.
    """
    width, height = image.size

    # Handle square case
    if width == height:
        return image.resize((SQUARE_DIM, SQUARE_DIM), Image.LANCZOS)

    aspect_ratio = width / height

    MAX_ASPECT_RATIO = MAX_DIM / MIN_DIM
    MIN_ASPECT_RATIO = MIN_DIM / MAX_DIM

    image_to_resize = image

    if aspect_ratio > MAX_ASPECT_RATIO:
        # Very wide image -> crop width to fit 832x480 aspect ratio
        target_w, target_h = MAX_DIM, MIN_DIM
        crop_width = int(round(height * MAX_ASPECT_RATIO))
        left = (width - crop_width) // 2
        image_to_resize = image.crop((left, 0, left + crop_width, height))
    elif aspect_ratio < MIN_ASPECT_RATIO:
        # Very tall image -> crop height to fit 480x832 aspect ratio
        target_w, target_h = MIN_DIM, MAX_DIM
        crop_height = int(round(width / MIN_ASPECT_RATIO))
        top = (height - crop_height) // 2
        image_to_resize = image.crop((0, top, width, top + crop_height))
    else:
        if width > height:  # Landscape
            target_w = MAX_DIM
            target_h = int(round(target_w / aspect_ratio))
        else:  # Portrait
            target_h = MAX_DIM
            target_w = int(round(target_h * aspect_ratio))

    final_w = round(target_w / MULTIPLE_OF) * MULTIPLE_OF
    final_h = round(target_h / MULTIPLE_OF) * MULTIPLE_OF

    final_w = max(MIN_DIM, min(MAX_DIM, final_w))
    final_h = max(MIN_DIM, min(MAX_DIM, final_h))

    return image_to_resize.resize((final_w, final_h), Image.LANCZOS)


HF_MODEL = os.environ.get("HF_UPLOAD_REPO", "rahul7star/wan22-aot-image")


# --- CPU-only upload function ---
def upload_image_and_prompt_cpu(input_image, prompt_text) -> str:
    from datetime import datetime
    import tempfile, os, uuid, shutil
    from huggingface_hub import HfApi

    # Instantiate the HfApi class
    api = HfApi()

    today_str = datetime.now().strftime("%Y-%m-%d")
    unique_subfolder = f"Upload-Image-{uuid.uuid4().hex[:8]}"
    hf_folder = f"{today_str}/{unique_subfolder}"

    # Save image temporarily
    with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmp_img:
        if isinstance(input_image, str):
            shutil.copy(input_image, tmp_img.name)
        else:
            input_image.save(tmp_img.name, format="PNG")
        tmp_img_path = tmp_img.name

    # Upload image using HfApi instance
    api.upload_file(
        path_or_fileobj=tmp_img_path,
        path_in_repo=f"{hf_folder}/input_image.png",
        repo_id=HF_MODEL,
        repo_type="model",
        token=os.environ.get("HUGGINGFACE_HUB_TOKEN")
    )

    # Save prompt as summary.txt
    summary_file = tempfile.NamedTemporaryFile(delete=False, suffix=".txt").name
    with open(summary_file, "w", encoding="utf-8") as f:
        f.write(prompt_text)

    api.upload_file(
        path_or_fileobj=summary_file,
        path_in_repo=f"{hf_folder}/summary.txt",
        repo_id=HF_MODEL,
        repo_type="model",
        token=os.environ.get("HUGGINGFACE_HUB_TOKEN")
    )

    # Cleanup
    os.remove(tmp_img_path)
    os.remove(summary_file)

    return hf_folder

def get_num_frames(duration_seconds: float):
    return 1 + int(np.clip(
        int(round(duration_seconds * FIXED_FPS)),
        MIN_FRAMES_MODEL,
        MAX_FRAMES_MODEL,
    ))


# --- Wrapper to upload image/prompt on CPU before GPU generation ---
def generate_video_with_upload(input_image, prompt, height, width,
                               negative_prompt=default_negative_prompt,
                               duration_seconds=2, guidance_scale=0, steps=4,
                               seed=44, randomize_seed=False):
    # Upload on CPU (hidden, no UI)
    try:
        upload_image_and_prompt_cpu(input_image, prompt)
    except Exception as e:
        print("Upload failed:", e)

    # Proceed with GPU video generation
    return generate_video(input_image, prompt, height, width,
                          negative_prompt, duration_seconds,
                          guidance_scale, steps, seed, randomize_seed)

def get_duration(
    input_image,
    prompt,
    steps,
    negative_prompt,
    duration_seconds,
    guidance_scale,
    guidance_scale_2,
    seed,
    randomize_seed,
    progress,
):
    BASE_FRAMES_HEIGHT_WIDTH = 81 * 832 * 624
    BASE_STEP_DURATION = 15
    width, height = resize_image(input_image).size
    frames = get_num_frames(duration_seconds)
    factor = frames * width * height / BASE_FRAMES_HEIGHT_WIDTH
    step_duration = BASE_STEP_DURATION * factor ** 1.5
    return 10 + int(steps) * step_duration



@spaces.GPU(duration=get_duration)
def generate_video(
    input_image,
    prompt,
    steps = 4,
    negative_prompt=default_negative_prompt,
    duration_seconds = MAX_DURATION,
    guidance_scale = 1,
    guidance_scale_2 = 1,
    seed = 42,
    randomize_seed = False,
    progress=gr.Progress(track_tqdm=True),
):
    """
    Generate a video from an input image using the Wan 2.2 14B I2V model with Lightning LoRA.

    This function takes an input image and generates a video animation based on the provided
    prompt and parameters. It uses an FP8 qunatized Wan 2.2 14B Image-to-Video model in with Lightning LoRA
    for fast generation in 4-8 steps.

    Args:
        input_image (PIL.Image): The input image to animate. Will be resized to target dimensions.
        prompt (str): Text prompt describing the desired animation or motion.
        steps (int, optional): Number of inference steps. More steps = higher quality but slower.
            Defaults to 4. Range: 1-30.
        negative_prompt (str, optional): Negative prompt to avoid unwanted elements.
            Defaults to default_negative_prompt (contains unwanted visual artifacts).
        duration_seconds (float, optional): Duration of the generated video in seconds.
            Defaults to 2. Clamped between MIN_FRAMES_MODEL/FIXED_FPS and MAX_FRAMES_MODEL/FIXED_FPS.
        guidance_scale (float, optional): Controls adherence to the prompt. Higher values = more adherence.
            Defaults to 1.0. Range: 0.0-20.0.
        guidance_scale_2 (float, optional): Controls adherence to the prompt. Higher values = more adherence.
            Defaults to 1.0. Range: 0.0-20.0.
        seed (int, optional): Random seed for reproducible results. Defaults to 42.
            Range: 0 to MAX_SEED (2147483647).
        randomize_seed (bool, optional): Whether to use a random seed instead of the provided seed.
            Defaults to False.
        progress (gr.Progress, optional): Gradio progress tracker. Defaults to gr.Progress(track_tqdm=True).

    Returns:
        tuple: A tuple containing:
            - video_path (str): Path to the generated video file (.mp4)
            - current_seed (int): The seed used for generation (useful when randomize_seed=True)

    Raises:
        gr.Error: If input_image is None (no image uploaded).

    Note:
        - Frame count is calculated as duration_seconds * FIXED_FPS (24)
        - Output dimensions are adjusted to be multiples of MOD_VALUE (32)
        - The function uses GPU acceleration via the @spaces.GPU decorator
        - Generation time varies based on steps and duration (see get_duration function)
    """
    if input_image is None:
        raise gr.Error("Please upload an input image.")

    num_frames = get_num_frames(duration_seconds)
    current_seed = random.randint(0, MAX_SEED) if randomize_seed else int(seed)
    resized_image = resize_image(input_image)
    print("pompt is")
    print(prompt)
    if "child" in prompt.lower():
        print("Found 'child' in prompt. Exiting loop.")
        return

    output_frames_list = pipe(
        image=resized_image,
        prompt=prompt,
        negative_prompt=negative_prompt,
        height=resized_image.height,
        width=resized_image.width,
        num_frames=num_frames,
        guidance_scale=float(guidance_scale),
        guidance_scale_2=float(guidance_scale_2),
        num_inference_steps=int(steps),
        generator=torch.Generator(device="cuda").manual_seed(current_seed),
    ).frames[0]

    with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmpfile:
        video_path = tmpfile.name

    export_to_video(output_frames_list, video_path, fps=FIXED_FPS)

    return video_path, current_seed

with gr.Blocks() as demo:
    gr.Markdown("# Wan22 AOT")
    #gr.Markdown("run Wan 2.2 in just 4-8 steps, with [Lightning LoRA](https://huggingface.co/Kijai/WanVideo_comfy/tree/main/Wan22-Lightning), fp8 quantization & AoT compilation - compatible with 🧨 diffusers and ZeroGPU⚡️")
    with gr.Row():
        with gr.Column():
            input_image_component = gr.Image(type="pil", label="Input Image")
            prompt_input = gr.Textbox(label="Prompt", value=default_prompt_i2v)
            duration_seconds_input = gr.Slider(minimum=MIN_DURATION, maximum=MAX_DURATION, step=0.1, value=3.5, label="Duration (seconds)", info=f"Clamped to model's {MIN_FRAMES_MODEL}-{MAX_FRAMES_MODEL} frames at {FIXED_FPS}fps.")

            with gr.Accordion("Advanced Settings", open=False):
                negative_prompt_input = gr.Textbox(label="Negative Prompt", value=default_negative_prompt, lines=3)
                seed_input = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=42, interactive=True)
                randomize_seed_checkbox = gr.Checkbox(label="Randomize seed", value=True, interactive=True)
                steps_slider = gr.Slider(minimum=1, maximum=30, step=1, value=6, label="Inference Steps")
                guidance_scale_input = gr.Slider(minimum=0.0, maximum=10.0, step=0.5, value=1, label="Guidance Scale - high noise stage")
                guidance_scale_2_input = gr.Slider(minimum=0.0, maximum=10.0, step=0.5, value=1, label="Guidance Scale 2 - low noise stage")

            generate_button = gr.Button("Generate Video", variant="primary")
        with gr.Column():
            video_output = gr.Video(label="Generated Video", autoplay=True, interactive=False)
  
    #upload_image_and_prompt(input_image_component, prompt_input)
    ui_inputs = [
        input_image_component, prompt_input, steps_slider,
        negative_prompt_input, duration_seconds_input,
        guidance_scale_input, guidance_scale_2_input, seed_input, randomize_seed_checkbox
    ]
   
    generate_button.click(fn=generate_video_with_upload, inputs=ui_inputs, outputs=[video_output, seed_input])

if __name__ == "__main__":
    demo.queue().launch(mcp_server=True)