import gradio as gr
import jax
import numpy as np
import jax.numpy as jnp
from flax.jax_utils import replicate
from flax.training.common_utils import shard
from PIL import Image
from diffusers import FlaxStableDiffusionControlNetPipeline, FlaxControlNetModel
import cv2 

def create_key(seed=0):
    return jax.random.PRNGKey(seed) 

def canny_filter(image):
    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    blurred_image = cv2.GaussianBlur(gray_image, (5, 5), 0)
    edges_image = cv2.Canny(blurred_image, 50, 200)
    return edges_image

# load control net and stable diffusion v1-5
controlnet, controlnet_params = FlaxControlNetModel.from_pretrained(
    "tsungtao/controlnet-mlsd-202305011046", from_flax=True, dtype=jnp.bfloat16
)

#controlnet.save_pretrained("tsungtao/controlnet-mlsd-202305011046",params=controlnet_params)

pipe, params = FlaxStableDiffusionControlNetPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5", controlnet=controlnet, revision="flax", dtype=jnp.bfloat16
)


def infer(prompts, negative_prompts, image):
    params["controlnet"] = controlnet_params
    
    num_samples = 1 #jax.device_count()
    rng = create_key(0)
    rng = jax.random.split(rng, jax.device_count())
    im = canny_filter(image)
    canny_image = Image.fromarray(im)
    
    prompt_ids = pipe.prepare_text_inputs([prompts] * num_samples)
    negative_prompt_ids = pipe.prepare_text_inputs([negative_prompts] * num_samples)
    processed_image = pipe.prepare_image_inputs([canny_image] * num_samples)
    
    p_params = replicate(params)
    prompt_ids = shard(prompt_ids)
    negative_prompt_ids = shard(negative_prompt_ids)
    processed_image = shard(processed_image)
    
    output = pipe(
        prompt_ids=prompt_ids,
        image=processed_image,
        params=p_params,
        prng_seed=rng,
        num_inference_steps=50,
        neg_prompt_ids=negative_prompt_ids,
        jit=True,
    ).images
    
    output_images = pipe.numpy_to_pil(np.asarray(output.reshape((num_samples,) + output.shape[-3:])))
    return output_images

title = "ControlNet MLSD"
description = "This is a demo on ControlNet MLSD."
examples = [["living room with TV", "fan", "image_01.jpg"],
            ["a living room with hardwood floors and a flat screen tv", "sea", "image_02.jpg"],
            ["a living room with a fireplace and a view of the ocean", "pendant", "image_03.jpg"]
           ]

with gr.Blocks() as demo:
    gr.Interface(infer, inputs=["text", "text", "image"], outputs="gallery", title = title, description = description, examples = examples, theme='gradio/soft')

    gr.Markdown(
    """ 
    * * *
    * [Dataset](https://huggingface.co/datasets/tsungtao/diffusers-testing)
    * [Diffusers model](https://huggingface.co/runwayml/stable-diffusion-v1-5)
    * [Training Report](https://wandb.ai/tsungtao0311/controlnet-mlsd-202305011046/runs/ezfn6bkz?workspace=user-tsungtao0311)
    """)

#    with gr.Accordion("Open for More!"): 
#        gr.Markdown("Team:https://huggingface.co/ellljoy, https://huggingface.co/zenkig, https://huggingface.co/aze555, https://huggingface.co/tsungtao, https://huggingface.co/Mayyu")
    
    gr.Markdown("* * *")
#    gr.Markdown(""" <img src='https://huggingface.co/spaces/tsungtao/tsungtao-controlnet-mlsd-202305011046/blob/main/test.png' /> """)


demo.launch()