# Copyright (c) 2015-present, Facebook, Inc.
# All rights reserved.
import torch
import torch.nn as nn
from functools import partial

from timm.models.vision_transformer import VisionTransformer, _cfg
from timm.models.registry import register_model
from timm.models.layers import trunc_normal_

from architectures.vit import TimmViT

__all__ = [
    "deit_tiny_patch16_224",
    "deit_small_patch16_224",
    "deit_base_patch16_224",
    "deit_tiny_distilled_patch16_224",
    "deit_small_distilled_patch16_224",
    "deit_base_distilled_patch16_224",
    "deit_base_patch16_384",
    "deit_base_distilled_patch16_384",
]


class DistilledVisionTransformer(TimmViT):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.dist_token = nn.Parameter(torch.zeros(1, 1, self.embed_dim))
        num_patches = self.patch_embed.num_patches
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 2, self.embed_dim))
        self.head_dist = nn.Linear(self.embed_dim, self.num_classes) if self.num_classes > 0 else nn.Identity()

        trunc_normal_(self.dist_token, std=0.02)
        trunc_normal_(self.pos_embed, std=0.02)
        self.head_dist.apply(self._init_weights)

    def forward_features(self, x):
        # taken from https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py
        # with slight modifications to add the dist_token
        B = x.shape[0]
        x = self.patch_embed(x)

        cls_tokens = self.cls_token.expand(B, -1, -1)  # stole cls_tokens impl from Phil Wang, thanks
        dist_token = self.dist_token.expand(B, -1, -1)
        x = torch.cat((cls_tokens, dist_token, x), dim=1)

        x = x + self.pos_embed
        x = self.pos_drop(x)

        for blk in self.blocks:
            x = blk(x)

        x = self.norm(x)
        return x[:, 0], x[:, 1]

    def forward(self, x):
        x, x_dist = self.forward_features(x)
        x = self.head(x)
        x_dist = self.head_dist(x_dist)
        if self.training:
            return x, x_dist
        else:
            # during inference, return the average of both classifier predictions
            return (x + x_dist) / 2


def _clean_kwargs(kwargs):
    allowed_keys = {key for key in kwargs.keys() if not key.startswith("pretrain")}
    allowed_keys = {key for key in allowed_keys if not key.startswith("cache")}
    return {key: kwargs[key] for key in allowed_keys}


@register_model
def deit_tiny_patch16(pretrained=False, img_size=224, drop_path_rate=0.1, num_classes=1000, drop_rate=0.0, **kwargs):
    kwargs = _clean_kwargs(kwargs)
    model = TimmViT(
        patch_size=16,
        embed_dim=192,
        depth=12,
        num_heads=3,
        mlp_ratio=4,
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        img_size=img_size,
        drop_path_rate=drop_path_rate,
        num_classes=num_classes,
        drop_rate=drop_rate,
    )
    model.default_cfg = _cfg()
    if pretrained:
        checkpoint = torch.hub.load_state_dict_from_url(
            url="https://dl.fbaipublicfiles.com/deit/deit_tiny_patch16_224-a1311bcf.pth",
            map_location="cpu",
            check_hash=True,
        )
        model.load_state_dict(checkpoint["model"])
    return model


@register_model
def deit_small_patch16(pretrained=False, img_size=224, drop_path_rate=0.1, num_classes=1000, drop_rate=0.0, **kwargs):
    kwargs = _clean_kwargs(kwargs)
    model = TimmViT(
        patch_size=16,
        embed_dim=384,
        depth=12,
        num_heads=6,
        mlp_ratio=4,
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        img_size=img_size,
        drop_path_rate=drop_path_rate,
        num_classes=num_classes,
        drop_rate=drop_rate,
    )
    model.default_cfg = _cfg()
    if pretrained:
        checkpoint = torch.hub.load_state_dict_from_url(
            url="https://dl.fbaipublicfiles.com/deit/deit_small_patch16_224-cd65a155.pth",
            map_location="cpu",
            check_hash=True,
        )
        model.load_state_dict(checkpoint["model"])
    return model


@register_model
def deit_base_patch16(pretrained=False, img_size=224, drop_path_rate=0.1, num_classes=1000, drop_rate=0.0, **kwargs):
    kwargs = _clean_kwargs(kwargs)
    model = TimmViT(
        patch_size=16,
        embed_dim=768,
        depth=12,
        num_heads=12,
        mlp_ratio=4,
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        img_size=img_size,
        drop_path_rate=drop_path_rate,
        num_classes=num_classes,
        drop_rate=drop_rate,
    )
    model.default_cfg = _cfg()
    if pretrained:
        checkpoint = torch.hub.load_state_dict_from_url(
            url="https://dl.fbaipublicfiles.com/deit/deit_base_patch16_224-b5f2ef4d.pth",
            map_location="cpu",
            check_hash=True,
        )
        model.load_state_dict(checkpoint["model"])
    return model


@register_model
def deit_tiny_distilled_patch16(
    pretrained=False, img_size=224, drop_path_rate=0.1, num_classes=1000, drop_rate=0.0, **kwargs
):
    kwargs = _clean_kwargs(kwargs)
    model = DistilledVisionTransformer(
        patch_size=16,
        embed_dim=192,
        depth=12,
        num_heads=3,
        mlp_ratio=4,
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        img_size=img_size,
        drop_path_rate=drop_path_rate,
        num_classes=num_classes,
        drop_rate=drop_rate,
    )
    model.default_cfg = _cfg()
    if pretrained:
        checkpoint = torch.hub.load_state_dict_from_url(
            url="https://dl.fbaipublicfiles.com/deit/deit_tiny_distilled_patch16_224-b40b3cf7.pth",
            map_location="cpu",
            check_hash=True,
        )
        model.load_state_dict(checkpoint["model"])
    return model


@register_model
def deit_small_distilled_patch16(
    pretrained=False, img_size=224, drop_path_rate=0.1, num_classes=1000, drop_rate=0.0, **kwargs
):
    kwargs = _clean_kwargs(kwargs)
    model = DistilledVisionTransformer(
        patch_size=16,
        embed_dim=384,
        depth=12,
        num_heads=6,
        mlp_ratio=4,
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        img_size=img_size,
        drop_path_rate=drop_path_rate,
        num_classes=num_classes,
        drop_rate=drop_rate,
    )
    model.default_cfg = _cfg()
    if pretrained:
        checkpoint = torch.hub.load_state_dict_from_url(
            url="https://dl.fbaipublicfiles.com/deit/deit_small_distilled_patch16_224-649709d9.pth",
            map_location="cpu",
            check_hash=True,
        )
        model.load_state_dict(checkpoint["model"])
    return model


@register_model
def deit_base_distilled_patch16(
    pretrained=False, img_size=224, drop_path_rate=0.1, num_classes=1000, drop_rate=0.0, **kwargs
):
    kwargs = _clean_kwargs(kwargs)
    model = DistilledVisionTransformer(
        patch_size=16,
        embed_dim=768,
        depth=12,
        num_heads=12,
        mlp_ratio=4,
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        img_size=img_size,
        drop_path_rate=drop_path_rate,
        num_classes=num_classes,
        drop_rate=drop_rate,
    )
    model.default_cfg = _cfg()
    if pretrained:
        checkpoint = torch.hub.load_state_dict_from_url(
            url="https://dl.fbaipublicfiles.com/deit/deit_base_distilled_patch16_224-df68dfff.pth",
            map_location="cpu",
            check_hash=True,
        )
        model.load_state_dict(checkpoint["model"])
    return model