import torch
from diffusers import StableDiffusionXLPipeline
import numpy as np
import gradio as gr
import random
from compel import Compel, ReturnedEmbeddingsType

device = "cuda" if torch.cuda.is_available() else "cpu"

if torch.cuda.is_available():
  torch.cuda.max_memory_allocated(device=device)
  pipe = StableDiffusionXLPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True)
  pipe = pipe.to(device)
else:
  pipe = StableDiffusionXLPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True)
  pipe = pipe.to(device)

pipe.safety_checker = None

pipe.load_lora_weights("artificialguybr/ps1redmond-ps1-game-graphics-lora-for-sdxl", weight_name="PS1Redmond-PS1Game-Playstation1Graphics.safetensors")
lora_activation_words = "playstation 1 graphics, PS1 Game, "

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1024

def infer(conditioning, pooled, neg_conditioning, neg_pooled, height, width, num_inference_steps, guidance_scale, seed, randomize_seed, lora_weight):
  if randomize_seed:
    seed = random.randint(0, MAX_SEED)

  generator = torch.Generator().manual_seed(seed)

  image = pipe(
    prompt_embeds=conditioning, 
    pooled_prompt_embeds=pooled,
    negative_prompt_embeds=neg_conditioning, 
    negative_pooled_prompt_embeds=neg_pooled,
    height=height,
    width=width,
    num_inference_steps=num_inference_steps,
    guidance_scale=guidance_scale,
    generator=generator,
    cross_attention_kwargs={"scale": lora_weight}
  ).images[0]

  return image

def get_embeds(prompt, negative_prompt, height, width, num_inference_steps, guidance_scale, seed, randomize_seed, lora_weight):

  compel = Compel(
    tokenizer=[pipe.tokenizer, pipe.tokenizer_2] ,
    text_encoder=[pipe.text_encoder, pipe.text_encoder_2],
    returned_embeddings_type=ReturnedEmbeddingsType.PENULTIMATE_HIDDEN_STATES_NON_NORMALIZED,
    requires_pooled=[False, True]
  )

  prompt = lora_activation_words + prompt

  conditioning, pooled = compel(prompt)
  neg_conditioning, neg_pooled = compel(negative_prompt)

  image = infer(conditioning, pooled, neg_conditioning, neg_pooled, height, width, num_inference_steps, guidance_scale, seed, randomize_seed, lora_weight)

  return image

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(f"""
        # Text-to-Image Gradio Template
        Currently running on {device.upper()}.
        """)
        
        with gr.Row():
            
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )
            
            run_button = gr.Button("Run", scale=0)
        
        result = gr.Image(label="Result", show_label=False)

        with gr.Accordion("Advanced Settings", open=False):
            
            negative_prompt = gr.Text(
                label="Negative prompt",
                max_lines=1,
                placeholder="Enter a negative prompt",
                visible=True,
            )
            
            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=1024,
                )
                
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1024,
                )
            
            with gr.Row():
                
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=10.0,
                    step=0.1,
                    value=7.5,
                )
                
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=100,
                    step=1,
                    value=30,
                )
              
            with gr.Row():

                lora_weight = gr.Slider(
                    label="LoRA weight",
                    minimum=0.0,
                    maximum=5.0,
                    step=0.01,
                    value=1,
                )

    run_button.click(
        fn = get_embeds,
        inputs = [prompt, negative_prompt, height, width, num_inference_steps, guidance_scale, seed, randomize_seed, lora_weight],
        outputs = [result]
    )

demo.launch(debug=True)