#Author: Lart Pang (https://github.com/lartpang)  

import math
from functools import partial

import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
from timm.models.vision_transformer import _cfg
from torch.backends import cuda
from torch.hub import load_state_dict_from_url
from torch.utils.checkpoint import checkpoint


class Mlp(nn.Module):
    def __init__(
        self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.0, linear=False
    ):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features)
        self.dwconv = DWConv(hidden_features)
        self.act = act_layer()
        self.fc2 = nn.Linear(hidden_features, out_features)
        self.drop = nn.Dropout(drop)
        self.linear = linear
        if self.linear:
            self.relu = nn.ReLU(inplace=True)
        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=0.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)
        elif isinstance(m, nn.Conv2d):
            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
            fan_out //= m.groups
            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
            if m.bias is not None:
                m.bias.data.zero_()

    def forward(self, x, H, W):
        x = self.fc1(x)
        if self.linear:
            x = self.relu(x)
        x = self.dwconv(x, H, W)
        x = self.act(x)
        x = self.drop(x)
        x = self.fc2(x)
        x = self.drop(x)
        return x


class Attention(nn.Module):
    def __init__(
        self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0.0, proj_drop=0.0, sr_ratio=1, linear=False
    ):
        super().__init__()
        assert dim % num_heads == 0, f"dim {dim} should be divided by num_heads {num_heads}."

        self.dim = dim
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim**-0.5

        self.q = nn.Linear(dim, dim, bias=qkv_bias)
        self.kv = nn.Linear(dim, dim * 2, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        self.linear = linear
        self.sr_ratio = sr_ratio
        if not linear:
            if sr_ratio > 1:
                self.sr = nn.Conv2d(dim, dim, kernel_size=sr_ratio, stride=sr_ratio)
                self.norm = nn.LayerNorm(dim)
        else:
            self.pool = nn.AdaptiveAvgPool2d(7)
            self.sr = nn.Conv2d(dim, dim, kernel_size=1, stride=1)
            self.norm = nn.LayerNorm(dim)
            self.act = nn.GELU()
        self.apply(self._init_weights)

        device_properties = torch.cuda.get_device_properties(torch.device("cuda"))
        if device_properties.major == 8 and device_properties.minor == 0:
            # print("A100 GPU detected, using flash attention if input tensor is on cuda")
            self.cuda_config = {"enable_flash": True, "enable_math": False, "enable_mem_efficient": False}
        else:
            # print("Non-A100 GPU detected, using math or mem efficient attention if input tensor is on cuda")
            self.cuda_config = {"enable_flash": False, "enable_math": True, "enable_mem_efficient": True}

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=0.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)
        elif isinstance(m, nn.Conv2d):
            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
            fan_out //= m.groups
            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
            if m.bias is not None:
                m.bias.data.zero_()

    def forward(self, x, H, W):
        B, N, C = x.shape
        q = self.q(x).reshape(B, N, self.num_heads, C // self.num_heads).permute(0, 2, 1, 3)

        if not self.linear:
            if self.sr_ratio > 1:
                x_ = x.permute(0, 2, 1).reshape(B, C, H, W)
                x_ = self.sr(x_).reshape(B, C, -1).permute(0, 2, 1)
                x_ = self.norm(x_)
                kv = self.kv(x_).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
            else:
                kv = self.kv(x).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        else:
            x_ = x.permute(0, 2, 1).reshape(B, C, H, W)
            x_ = self.sr(self.pool(x_)).reshape(B, C, -1).permute(0, 2, 1)
            x_ = self.norm(x_)
            x_ = self.act(x_)
            kv = self.kv(x_).reshape(B, -1, 2, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        k, v = kv[0], kv[1]

        # attn = (q @ k.transpose(-2, -1)) * self.scale
        # attn = attn.softmax(dim=-1)
        # attn = self.attn_drop(attn)
        # x = attn @ v
        with cuda.sdp_kernel(**self.cuda_config):
            # q: bs,nh,l,hd, k: bs,nh,s,hd, v: bs,nh,s,hd
            # same as: (((q @ k.transpose(-1, -2)) * q.shape[-1] ** -0.5).softmax(dim=-1) @ v)
            x = F.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=0, is_causal=False)  # built-in scale

        x = x.transpose(1, 2).reshape(B, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)

        return x


class Block(nn.Module):
    def __init__(
        self,
        dim,
        num_heads,
        mlp_ratio=4.0,
        qkv_bias=False,
        qk_scale=None,
        drop=0.0,
        attn_drop=0.0,
        drop_path=0.0,
        act_layer=nn.GELU,
        norm_layer=nn.LayerNorm,
        sr_ratio=1,
        linear=False,
    ):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim,
            num_heads=num_heads,
            qkv_bias=qkv_bias,
            qk_scale=qk_scale,
            attn_drop=attn_drop,
            proj_drop=drop,
            sr_ratio=sr_ratio,
            linear=linear,
        )
        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
        self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
        self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop, linear=linear)

        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=0.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)
        elif isinstance(m, nn.Conv2d):
            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
            fan_out //= m.groups
            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
            if m.bias is not None:
                m.bias.data.zero_()

    def forward(self, x, H, W):
        x = x + self.drop_path(self.attn(self.norm1(x), H, W))
        x = x + self.drop_path(self.mlp(self.norm2(x), H, W))

        return x


class OverlapPatchEmbed(nn.Module):
    """Image to Patch Embedding"""

    def __init__(self, img_size=224, patch_size=7, stride=4, in_chans=3, embed_dim=768):
        super().__init__()

        img_size = to_2tuple(img_size)
        patch_size = to_2tuple(patch_size)

        assert max(patch_size) > stride, "Set larger patch_size than stride"

        self.img_size = img_size
        self.patch_size = patch_size
        self.H, self.W = img_size[0] // stride, img_size[1] // stride
        self.num_patches = self.H * self.W
        self.proj = nn.Conv2d(
            in_chans,
            embed_dim,
            kernel_size=patch_size,
            stride=stride,
            padding=(patch_size[0] // 2, patch_size[1] // 2),
        )
        self.norm = nn.LayerNorm(embed_dim)

        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=0.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)
        elif isinstance(m, nn.Conv2d):
            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
            fan_out //= m.groups
            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
            if m.bias is not None:
                m.bias.data.zero_()

    def forward(self, x):
        x = self.proj(x)
        _, _, H, W = x.shape
        x = x.flatten(2).transpose(1, 2)
        x = self.norm(x)

        return x, H, W


class PyramidVisionTransformerV2(nn.Module):
    def __init__(
        self,
        img_size=224,
        patch_size=16,
        in_chans=3,
        num_classes=1000,
        embed_dims=[64, 128, 256, 512],
        num_heads=[1, 2, 4, 8],
        mlp_ratios=[4, 4, 4, 4],
        qkv_bias=False,
        qk_scale=None,
        drop_rate=0.0,
        attn_drop_rate=0.0,
        drop_path_rate=0.0,
        norm_layer=nn.LayerNorm,
        depths=[3, 4, 6, 3],
        sr_ratios=[8, 4, 2, 1],
        num_stages=4,
        linear=False,
        use_checkpoint=False,
    ):
        super().__init__()
        self.num_classes = num_classes
        self.depths = depths
        self.num_stages = num_stages
        self.embed_dims = embed_dims
        self.use_checkpoint = use_checkpoint

        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule
        cur = 0

        for i in range(num_stages):
            patch_embed = OverlapPatchEmbed(
                img_size=img_size if i == 0 else img_size // (2 ** (i + 1)),
                patch_size=7 if i == 0 else 3,
                stride=4 if i == 0 else 2,
                in_chans=in_chans if i == 0 else embed_dims[i - 1],
                embed_dim=embed_dims[i],
            )

            block = nn.ModuleList(
                [
                    Block(
                        dim=embed_dims[i],
                        num_heads=num_heads[i],
                        mlp_ratio=mlp_ratios[i],
                        qkv_bias=qkv_bias,
                        qk_scale=qk_scale,
                        drop=drop_rate,
                        attn_drop=attn_drop_rate,
                        drop_path=dpr[cur + j],
                        norm_layer=norm_layer,
                        sr_ratio=sr_ratios[i],
                        linear=linear,
                    )
                    for j in range(depths[i])
                ]
            )
            norm = norm_layer(embed_dims[i])
            cur += depths[i]

            setattr(self, f"patch_embed{i + 1}", patch_embed)
            setattr(self, f"block{i + 1}", block)
            setattr(self, f"norm{i + 1}", norm)

        # classification head
        # self.head = nn.Linear(embed_dims[3], num_classes) if num_classes > 0 else nn.Identity()

        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=0.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)
        elif isinstance(m, nn.Conv2d):
            fan_out = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
            fan_out //= m.groups
            m.weight.data.normal_(0, math.sqrt(2.0 / fan_out))
            if m.bias is not None:
                m.bias.data.zero_()

    def freeze_patch_emb(self):
        self.patch_embed1.requires_grad = False

    @torch.jit.ignore
    def no_weight_decay(self):
        return {"pos_embed1", "pos_embed2", "pos_embed3", "pos_embed4", "cls_token"}  # has pos_embed may be better

    def extract_endpoints(self, x):
        B = x.shape[0]
        endpoints = dict()
        for i in range(self.num_stages):
            patch_embed = getattr(self, f"patch_embed{i + 1}")
            block = getattr(self, f"block{i + 1}")
            norm = getattr(self, f"norm{i + 1}")
            x, H, W = patch_embed(x)
            for blk in block:
                if self.use_checkpoint:
                    x = checkpoint(blk, x, H, W)
                else:
                    x = blk(x, H, W)
            x = norm(x)
            x = x.reshape(B, H, W, -1).permute(0, 3, 1, 2).contiguous()
            # print(i + 2, x.shape)
            endpoints["reduction_{}".format(i + 2)] = x
        return endpoints

    def forward(self, x):
        endpoints = self.extract_endpoints(x)
        return endpoints


class DWConv(nn.Module):
    def __init__(self, dim=768):
        super(DWConv, self).__init__()
        self.dwconv = nn.Conv2d(dim, dim, 3, 1, 1, bias=True, groups=dim)

    def forward(self, x, H, W):
        B, N, C = x.shape
        x = x.transpose(1, 2).view(B, C, H, W)
        x = self.dwconv(x)
        x = x.flatten(2).transpose(1, 2)

        return x


def _conv_filter(state_dict, patch_size=16):
    """convert patch embedding weight from manual patchify + linear proj to conv"""
    out_dict = {}
    for k, v in state_dict.items():
        if "patch_embed.proj.weight" in k:
            v = v.reshape((v.shape[0], 3, patch_size, patch_size))
        out_dict[k] = v

    return out_dict


def pvt_v2_eff_b0(pretrained=False, **kwargs):
    model = PyramidVisionTransformerV2(
        patch_size=4,
        embed_dims=[32, 64, 160, 256],
        num_heads=[1, 2, 5, 8],
        mlp_ratios=[8, 8, 4, 4],
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        depths=[2, 2, 2, 2],
        sr_ratios=[8, 4, 2, 1],
        **kwargs,
    )
    model.default_cfg = _cfg()
    if pretrained:
        state_dict = load_state_dict_from_url(
            "https://github.com/whai362/PVT/releases/download/v2/pvt_v2_b0.pth", progress=True
        )
        state_dict.pop("head.weight")
        state_dict.pop("head.bias")
        model.load_state_dict(state_dict)
    return model


def pvt_v2_eff_b1(pretrained=False, **kwargs):
    model = PyramidVisionTransformerV2(
        patch_size=4,
        embed_dims=[64, 128, 320, 512],
        num_heads=[1, 2, 5, 8],
        mlp_ratios=[8, 8, 4, 4],
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        depths=[2, 2, 2, 2],
        sr_ratios=[8, 4, 2, 1],
        **kwargs,
    )
    model.default_cfg = _cfg()
    if pretrained:
        state_dict = load_state_dict_from_url(
            "https://github.com/whai362/PVT/releases/download/v2/pvt_v2_b1.pth", progress=True
        )
        state_dict.pop("head.weight")
        state_dict.pop("head.bias")
        model.load_state_dict(state_dict)
    return model


def pvt_v2_eff_b2(pretrained=False, **kwargs):
    model = PyramidVisionTransformerV2(
        patch_size=4,
        embed_dims=[64, 128, 320, 512],
        num_heads=[1, 2, 5, 8],
        mlp_ratios=[8, 8, 4, 4],
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        depths=[3, 4, 6, 3],
        sr_ratios=[8, 4, 2, 1],
        **kwargs,
    )
    model.default_cfg = _cfg()
    if pretrained:
        state_dict = load_state_dict_from_url(
            "https://github.com/whai362/PVT/releases/download/v2/pvt_v2_b2.pth", progress=True
        )
        state_dict.pop("head.weight")
        state_dict.pop("head.bias")
        model.load_state_dict(state_dict)
    return model


def pvt_v2_eff_b3(pretrained=False, **kwargs):
    model = PyramidVisionTransformerV2(
        patch_size=4,
        embed_dims=[64, 128, 320, 512],
        num_heads=[1, 2, 5, 8],
        mlp_ratios=[8, 8, 4, 4],
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        depths=[3, 4, 18, 3],
        sr_ratios=[8, 4, 2, 1],
        **kwargs,
    )
    model.default_cfg = _cfg()
    if pretrained:
        state_dict = load_state_dict_from_url(
            "https://github.com/whai362/PVT/releases/download/v2/pvt_v2_b3.pth", progress=True
        )
        state_dict.pop("head.weight")
        state_dict.pop("head.bias")
        model.load_state_dict(state_dict)

    return model


def pvt_v2_eff_b4(pretrained=False, **kwargs):
    model = PyramidVisionTransformerV2(
        patch_size=4,
        embed_dims=[64, 128, 320, 512],
        num_heads=[1, 2, 5, 8],
        mlp_ratios=[8, 8, 4, 4],
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        depths=[3, 8, 27, 3],
        sr_ratios=[8, 4, 2, 1],
        **kwargs,
    )
    model.default_cfg = _cfg()
    if pretrained:
        state_dict = load_state_dict_from_url(
            "https://github.com/whai362/PVT/releases/download/v2/pvt_v2_b4.pth", progress=True
        )
        state_dict.pop("head.weight")
        state_dict.pop("head.bias")
        model.load_state_dict(state_dict)
    return model


def pvt_v2_eff_b5(pretrained=False, **kwargs):
    model = PyramidVisionTransformerV2(
        patch_size=4,
        embed_dims=[64, 128, 320, 512],
        num_heads=[1, 2, 5, 8],
        mlp_ratios=[4, 4, 4, 4],
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        depths=[3, 6, 40, 3],
        sr_ratios=[8, 4, 2, 1],
        **kwargs,
    )
    model.default_cfg = _cfg()
    if pretrained:
        state_dict = load_state_dict_from_url(
            "https://github.com/whai362/PVT/releases/download/v2/pvt_v2_b5.pth", progress=True
        )
        state_dict.pop("head.weight")
        state_dict.pop("head.bias")
        model.load_state_dict(state_dict)
    return model


def pvt_v2_eff_b2_li(pretrained=False, **kwargs):
    model = PyramidVisionTransformerV2(
        patch_size=4,
        embed_dims=[64, 128, 320, 512],
        num_heads=[1, 2, 5, 8],
        mlp_ratios=[8, 8, 4, 4],
        qkv_bias=True,
        norm_layer=partial(nn.LayerNorm, eps=1e-6),
        depths=[3, 4, 6, 3],
        sr_ratios=[8, 4, 2, 1],
        linear=True,
        **kwargs,
    )
    model.default_cfg = _cfg()

    return model