import os, json, requests, runpod

import torch, random
from kolors.pipelines.pipeline_stable_diffusion_xl_chatglm_256 import StableDiffusionXLPipeline
from kolors.models.modeling_chatglm import ChatGLMModel
from kolors.models.tokenization_chatglm import ChatGLMTokenizer
from diffusers import UNet2DConditionModel, AutoencoderKL
from diffusers import EulerDiscreteScheduler

with torch.inference_mode():
    ckpt_dir = f'/content/Kolors/weights/Kolors'
    text_encoder = ChatGLMModel.from_pretrained(
        f'{ckpt_dir}/text_encoder',
        torch_dtype=torch.float16).half()
    tokenizer = ChatGLMTokenizer.from_pretrained(f'{ckpt_dir}/text_encoder')
    vae = AutoencoderKL.from_pretrained(f"{ckpt_dir}/vae", revision=None).half()
    scheduler = EulerDiscreteScheduler.from_pretrained(f"{ckpt_dir}/scheduler")
    unet = UNet2DConditionModel.from_pretrained(f"{ckpt_dir}/unet", revision=None).half()
    pipe = StableDiffusionXLPipeline(
            vae=vae,
            text_encoder=text_encoder,
            tokenizer=tokenizer,
            unet=unet,
            scheduler=scheduler,
            force_zeros_for_empty_prompt=False)
    pipe = pipe.to("cuda")
    pipe.enable_model_cpu_offload()

def closestNumber(n, m):
    q = int(n / m)
    n1 = m * q
    if (n * m) > 0:
        n2 = m * (q + 1)
    else:
        n2 = m * (q - 1)
    if abs(n - n1) < abs(n - n2):
        return n1
    return n2

@torch.inference_mode()
def generate(input):
    values = input["input"]

    prompt = values['prompt']
    width = values['width']
    height = values['height']
    num_inference_steps = values['num_inference_steps']
    guidance_scale = values['guidance_scale']
    num_images_per_prompt = values['num_images_per_prompt']
    seed = values['seed']

    if seed == 0:
        seed = random.randint(0, 18446744073709551615)

    image = pipe(
        prompt=prompt,
        width=closestNumber(width, 8),
        height=closestNumber(height, 8),
        num_inference_steps=num_inference_steps,
        guidance_scale=guidance_scale,
        num_images_per_prompt=num_images_per_prompt,
        generator=torch.Generator(pipe.device).manual_seed(seed)).images[0]
    image.save(f'/content/Kolors/scripts/outputs/kolors.jpg')

    result = "/content/Kolors/scripts/outputs/kolors.jpg"
    try:
        notify_uri = values['notify_uri']
        del values['notify_uri']
        notify_token = values['notify_token']
        del values['notify_token']
        discord_id = values['discord_id']
        del values['discord_id']
        if(discord_id == "discord_id"):
            discord_id = os.getenv('com_camenduru_discord_id')
        discord_channel = values['discord_channel']
        del values['discord_channel']
        if(discord_channel == "discord_channel"):
            discord_channel = os.getenv('com_camenduru_discord_channel')
        discord_token = values['discord_token']
        del values['discord_token']
        if(discord_token == "discord_token"):
            discord_token = os.getenv('com_camenduru_discord_token')
        job_id = values['job_id']
        del values['job_id']
        default_filename = os.path.basename(result)
        with open(result, "rb") as file:
            files = {default_filename: file.read()}
        payload = {"content": f"{json.dumps(values)} <@{discord_id}>"}
        response = requests.post(
            f"https://discord.com/api/v9/channels/{discord_channel}/messages",
            data=payload,
            headers={"Authorization": f"Bot {discord_token}"},
            files=files
        )
        response.raise_for_status()
        result_url = response.json()['attachments'][0]['url']
        notify_payload = {"jobId": job_id, "result": result_url, "status": "DONE"}
        web_notify_uri = os.getenv('com_camenduru_web_notify_uri')
        web_notify_token = os.getenv('com_camenduru_web_notify_token')
        if(notify_uri == "notify_uri"):
            requests.post(web_notify_uri, data=json.dumps(notify_payload), headers={'Content-Type': 'application/json', "Authorization": web_notify_token})
        else:
            requests.post(web_notify_uri, data=json.dumps(notify_payload), headers={'Content-Type': 'application/json', "Authorization": web_notify_token})
            requests.post(notify_uri, data=json.dumps(notify_payload), headers={'Content-Type': 'application/json', "Authorization": notify_token})
        return {"jobId": job_id, "result": result_url, "status": "DONE"}
    except Exception as e:
        error_payload = {"jobId": job_id, "status": "FAILED"}
        try:
            if(notify_uri == "notify_uri"):
                requests.post(web_notify_uri, data=json.dumps(error_payload), headers={'Content-Type': 'application/json', "Authorization": web_notify_token})
            else:
                requests.post(web_notify_uri, data=json.dumps(error_payload), headers={'Content-Type': 'application/json', "Authorization": web_notify_token})
                requests.post(notify_uri, data=json.dumps(error_payload), headers={'Content-Type': 'application/json', "Authorization": notify_token})
        except:
            pass
        return {"jobId": job_id, "result": f"FAILED: {str(e)}", "status": "FAILED"}
    finally:
        if os.path.exists(result):
            os.remove(result)

runpod.serverless.start({"handler": generate})