import argparse
import re
from pathlib import Path
from tqdm import tqdm

import torch
from PIL import Image
from torch import nn
from transformers import (
    AutoModel,
    AutoModelForCausalLM,
    AutoProcessor,
    AutoTokenizer,
    PreTrainedTokenizer,
    PreTrainedTokenizerFast,
)
import torchvision.transforms.functional as TVF

# Constants
CLIP_PATH = "google/siglip-so400m-patch14-384"
CHECKPOINT_PATH = Path("joy-caption-alpha-two/cgrkzexw-599808")
CAPTION_TYPE_MAP = {
    "Descriptive": [
        "Write a descriptive caption for this image in a formal tone.",
        "Write a descriptive caption for this image in a formal tone within {word_count} words.",
        "Write a {length} descriptive caption for this image in a formal tone.",
    ],
    "Descriptive (Informal)": [
        "Write a descriptive caption for this image in a casual tone.",
        "Write a descriptive caption for this image in a casual tone within {word_count} words.",
        "Write a {length} descriptive caption for this image in a casual tone.",
    ],
    "Training Prompt": [
        "Write a stable diffusion prompt for this image.",
        "Write a stable diffusion prompt for this image within {word_count} words.",
        "Write a {length} stable diffusion prompt for this image.",
    ],
    "MidJourney": [
        "Write a MidJourney prompt for this image.",
        "Write a MidJourney prompt for this image within {word_count} words.",
        "Write a {length} MidJourney prompt for this image.",
    ],
    "Booru tag list": [
        "Write a list of Booru tags for this image.",
        "Write a list of Booru tags for this image within {word_count} words.",
        "Write a {length} list of Booru tags for this image.",
    ],
    "Booru-like tag list": [
        "Write a list of Booru-like tags for this image.",
        "Write a list of Booru-like tags for this image within {word_count} words.",
        "Write a {length} list of Booru-like tags for this image.",
    ],
    "Art Critic": [
        "Analyze this image like an art critic would with information about its composition, style, symbolism, the use of color, light, any artistic movement it might belong to, etc.",
        "Analyze this image like an art critic would with information about its composition, style, symbolism, the use of color, light, any artistic movement it might belong to, etc. Keep it within {word_count} words.",
        "Analyze this image like an art critic would with information about its composition, style, symbolism, the use of color, light, any artistic movement it might belong to, etc. Keep it {length}.",
    ],
    "Product Listing": [
        "Write a caption for this image as though it were a product listing.",
        "Write a caption for this image as though it were a product listing. Keep it under {word_count} words.",
        "Write a {length} caption for this image as though it were a product listing.",
    ],
    "Social Media Post": [
        "Write a caption for this image as if it were being used for a social media post.",
        "Write a caption for this image as if it were being used for a social media post. Limit the caption to {word_count} words.",
        "Write a {length} caption for this image as if it were being used for a social media post.",
    ],
}

class ImageAdapter(nn.Module):
    def __init__(self, input_features: int, output_features: int, ln1: bool, pos_emb: bool, num_image_tokens: int, deep_extract: bool):
        super().__init__()
        self.deep_extract = deep_extract

        if self.deep_extract:
            input_features = input_features * 5

        self.linear1 = nn.Linear(input_features, output_features)
        self.activation = nn.GELU()
        self.linear2 = nn.Linear(output_features, output_features)
        self.ln1 = nn.Identity() if not ln1 else nn.LayerNorm(input_features)
        self.pos_emb = None if not pos_emb else nn.Parameter(torch.zeros(num_image_tokens, input_features))

        # Other tokens (<|image_start|>, <|image_end|>, <|eot_id|>)
        self.other_tokens = nn.Embedding(3, output_features)
        self.other_tokens.weight.data.normal_(mean=0.0, std=0.02)

    def forward(self, vision_outputs: torch.Tensor):
        if self.deep_extract:
            x = torch.concat((
                vision_outputs[-2],
                vision_outputs[3],
                vision_outputs[7],
                vision_outputs[13],
                vision_outputs[20],
            ), dim=-1)
        else:
            x = vision_outputs[-2]

        x = self.ln1(x)

        if self.pos_emb is not None:
            x = x + self.pos_emb

        x = self.linear1(x)
        x = self.activation(x)
        x = self.linear2(x)

        # <|image_start|>, IMAGE, <|image_end|>
        other_tokens = self.other_tokens(torch.tensor([0, 1], device=self.other_tokens.weight.device).expand(x.shape[0], -1))
        x = torch.cat((other_tokens[:, 0:1], x, other_tokens[:, 1:2]), dim=1)

        return x

    def get_eot_embedding(self):
        return self.other_tokens(torch.tensor([2], device=self.other_tokens.weight.device)).squeeze(0)


# Load CLIP
print("Loading CLIP")
clip_processor = AutoProcessor.from_pretrained(CLIP_PATH)
clip_model = AutoModel.from_pretrained(CLIP_PATH)
clip_model = clip_model.vision_model

assert (CHECKPOINT_PATH / "clip_model.pt").exists()
print("Loading VLM's custom vision model")
checkpoint = torch.load(CHECKPOINT_PATH / "clip_model.pt", map_location='cpu')
checkpoint = {k.replace("_orig_mod.module.", ""): v for k, v in checkpoint.items()}
clip_model.load_state_dict(checkpoint)
del checkpoint

clip_model.eval()
clip_model.requires_grad_(False)
clip_model.to("cuda")

# Tokenizer
print("Loading tokenizer")
tokenizer = AutoTokenizer.from_pretrained(CHECKPOINT_PATH / "text_model", use_fast=True)
assert isinstance(tokenizer, PreTrainedTokenizer) or isinstance(tokenizer, PreTrainedTokenizerFast), f"Tokenizer is of type {type(tokenizer)}"

# LLM
print("Loading LLM")
print("Loading VLM's custom text model")
text_model = AutoModelForCausalLM.from_pretrained(CHECKPOINT_PATH / "text_model", device_map=0, torch_dtype=torch.bfloat16)
text_model.eval()

# Image Adapter
print("Loading image adapter")
image_adapter = ImageAdapter(clip_model.config.hidden_size, text_model.config.hidden_size, False, False, 38, False)
image_adapter.load_state_dict(torch.load(CHECKPOINT_PATH / "image_adapter.pt", map_location="cpu"))
image_adapter.eval()
image_adapter.to("cuda")


def filter_caption_start(caption):
    # Remove any leading and trailing whitespace
    caption = caption.strip()
    # Split caption into lines
    lines = caption.splitlines()
    # Find the longest line
    if not lines:
        return caption
    longest_line = max(lines, key=lambda line: len(line.strip()))
    # Return the longest line
    return longest_line.strip()


@torch.no_grad()
def stream_chat(folder_path: str, caption_type: str, caption_length: str | int, extra_options: list[str], name_input: str, custom_prompt: str):
    folder_path = Path(folder_path)
    if not folder_path.is_dir():
        return "Invalid folder path."

    torch.cuda.empty_cache()
    length = None if caption_length == "any" else caption_length

    if isinstance(length, str):
        try:
            length = int(length)
        except ValueError:
            pass

    if length is None:
        map_idx = 0
    elif isinstance(length, int):
        map_idx = 1
    elif isinstance(length, str):
        map_idx = 2
    else:
        raise ValueError(f"Invalid caption length: {length}")

    prompt_str = CAPTION_TYPE_MAP[caption_type][map_idx]

    if len(extra_options) > 0:
        prompt_str += " " + " ".join(extra_options)

    prompt_str = prompt_str.format(name=name_input, length=caption_length, word_count=caption_length)

    if custom_prompt.strip() != "":
        prompt_str = custom_prompt.strip()

    for image_file in tqdm(folder_path.iterdir(), desc="Processing images"):
        if image_file.suffix.lower() in ['.png', '.jpg', '.jpeg', '.bmp']:
            input_image = Image.open(image_file)

            image = input_image.resize((384, 384), Image.LANCZOS)
            pixel_values = TVF.pil_to_tensor(image).unsqueeze(0) / 255.0
            pixel_values = TVF.normalize(pixel_values, [0.5], [0.5])
            pixel_values = pixel_values.to('cuda')

            with torch.amp.autocast_mode.autocast('cuda', enabled=True):
                vision_outputs = clip_model(pixel_values=pixel_values, output_hidden_states=True)
                embedded_images = image_adapter(vision_outputs.hidden_states)
                embedded_images = embedded_images.to('cuda')

            convo = [
                {
                    "role": "system",
                    "content": "You are a helpful image captioner. Do not include any preamble or assistant's name in your response.",
                },
                {
                    "role": "user",
                    "content": prompt_str,
                },
            ]

            convo_string = tokenizer.apply_chat_template(convo, tokenize=False, add_generation_prompt=True)
            assert isinstance(convo_string, str)
            convo_tokens = tokenizer.encode(convo_string, return_tensors="pt", add_special_tokens=False, truncation=False).squeeze(0)
            prompt_tokens = tokenizer.encode(prompt_str, return_tensors="pt", add_special_tokens=False, truncation=False).squeeze(0)

            eot_id_indices = (convo_tokens == tokenizer.convert_tokens_to_ids("<|eot_id|>")).nonzero(as_tuple=True)[0].tolist()
            assert len(eot_id_indices) == 2, f"Expected 2 <|eot_id|> tokens, got {len(eot_id_indices)}"
            preamble_len = eot_id_indices[1] - prompt_tokens.shape[0]

            convo_embeds = text_model.model.embed_tokens(convo_tokens.unsqueeze(0).to('cuda'))

            input_embeds = torch.cat([
                convo_embeds[:, :preamble_len],
                embedded_images.to(dtype=convo_embeds.dtype),
                convo_embeds[:, preamble_len:],
            ], dim=1).to('cuda')

            input_ids = torch.cat([
                convo_tokens[:preamble_len].unsqueeze(0),
                torch.zeros((1, embedded_images.shape[1]), dtype=torch.long),
                convo_tokens[preamble_len:].unsqueeze(0),
            ], dim=1).to('cuda')

            attention_mask = torch.ones_like(input_ids)

            generate_ids = text_model.generate(
                input_ids,
                inputs_embeds=input_embeds,
                attention_mask=attention_mask,
                max_new_tokens=300,
                do_sample=True
            )

            generate_ids = generate_ids[:, input_ids.shape[1]:]
            if generate_ids[0][-1] == tokenizer.eos_token_id or generate_ids[0][-1] == tokenizer.convert_tokens_to_ids("<|eot_id|>"):
                generate_ids = generate_ids[:, :-1]

            caption = tokenizer.batch_decode(generate_ids, skip_special_tokens=True)[0]

            caption = filter_caption_start(caption)

            base_name = image_file.stem
            text_file_path = folder_path / f"{base_name}.txt"
            with open(text_file_path, 'w', encoding='utf-8') as f:
                f.write(caption)

            print(f"Saved caption to {text_file_path}")

    return "Processing complete."

def main():
    parser = argparse.ArgumentParser(description="Image Captioning Script")
    parser.add_argument('--input', '--folder_path', dest='folder_path', type=str, default='/content/images', help='Folder Path containing images')
    parser.add_argument('--caption_type', type=str, default='Descriptive', choices=list(CAPTION_TYPE_MAP.keys()), help='Caption Type')
    parser.add_argument('--length', '--caption_length', dest='caption_length', type=str, default='short', help='Caption Length')
    parser.add_argument('--extra_options', nargs='*', default=[], help='Extra Options')
    parser.add_argument('--name_input', type=str, default='', help='Person/Character Name (if applicable)')
    parser.add_argument('--custom_prompt', type=str, default='', help='Custom Prompt (optional)')

    args = parser.parse_args()

    result = stream_chat(
        folder_path=args.folder_path,
        caption_type=args.caption_type,
        caption_length=args.caption_length,
        extra_options=args.extra_options,
        name_input=args.name_input,
        custom_prompt=args.custom_prompt,
    )
    print(result)


if __name__ == '__main__':
    main()