revswin.py

# --------------------------------------------------------
# Swin Transformer v1
# Copyright (c) 2021 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Ze Liu, Yutong Lin, Yixuan Wei
# --------------------------------------------------------

# Modified by Tyler Zhu from https://github.com/microsoft/Swin-Transformer/blob/main/models/swin_transformer.py

import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as checkpoint
from timm.models.layers import DropPath, Mlp, to_2tuple, trunc_normal_
from torch.autograd import Function as Function


class TwoStreamFusion(nn.Module):
    def __init__(self, dim=None, kernel=3, padding=1):
        """
        Module for fusing both streams of the reversible model by concatenation,
        then applying an MLP with a hidden dim of dim*2 and output dim of dim
        to downsample.
        """
        super().__init__()
        self.dim = dim
        self.fuse_fn = Mlp(
            in_features=dim * 2,
            hidden_features=int(dim * 2),
            act_layer=nn.GELU,
            out_features=dim,
        )

    def forward(self, x):
        return self.fuse_fn(x)


def window_partition(x, window_size):
    """
    Args:
        x: (B, H, W, C)
        window_size (int): window size
    Returns:
        windows: (num_windows*B, window_size, window_size, C)
    """
    B, H, W, C = x.shape
    # print("partition", x.shape, window_size)
    x = x.view(
        B, H // window_size, window_size, W // window_size, window_size, C
    )
    windows = (
        x.permute(0, 1, 3, 2, 4, 5)
        .contiguous()
        .view(-1, window_size, window_size, C)
    )
    return windows


def window_reverse(windows, window_size, H, W):
    """
    Args:
        windows: (num_windows*B, window_size, window_size, C)
        window_size (int): Window size
        H (int): Height of image
        W (int): Width of image
    Returns:
        x: (B, H, W, C)
    """
    # print("reverse", windows.shape, window_size, H, W)
    B = int(windows.shape[0] / (H * W / window_size / window_size))
    x = windows.view(
        B, H // window_size, W // window_size, window_size, window_size, -1
    )
    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)
    return x


class WindowAttention(nn.Module):
    """Window based multi-head self attention (W-MSA) module with relative position bias.
    It supports both of shifted and non-shifted window.
    Args:
        dim (int): Number of input channels.
        window_size (tuple[int]): The height and width of the window.
        num_heads (int): Number of attention heads.
        qkv_bias (bool, optional):  If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set
        attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0
        proj_drop (float, optional): Dropout ratio of output. Default: 0.0
    """

    def __init__(
        self,
        dim,
        window_size,
        num_heads,
        qkv_bias=True,
        qk_scale=None,
        attn_drop=0.0,
        proj_drop=0.0,
    ):
        super().__init__()
        self.dim = dim
        self.window_size = window_size  # Wh, Ww
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim**-0.5

        # define a parameter table of relative position bias
        self.relative_position_bias_table = nn.Parameter(
            torch.zeros(
                (2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads
            )
        )  # 2*Wh-1 * 2*Ww-1, nH

        # get pair-wise relative position index for each token inside the window
        coords_h = torch.arange(self.window_size[0])
        coords_w = torch.arange(self.window_size[1])
        coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
        coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
        relative_coords = (
            coords_flatten[:, :, None] - coords_flatten[:, None, :]
        )  # 2, Wh*Ww, Wh*Ww
        relative_coords = relative_coords.permute(
            1, 2, 0
        ).contiguous()  # Wh*Ww, Wh*Ww, 2
        relative_coords[:, :, 0] += (
            self.window_size[0] - 1
        )  # shift to start from 0
        relative_coords[:, :, 1] += self.window_size[1] - 1
        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
        relative_position_index = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
        self.register_buffer("relative_position_index", relative_position_index)

        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        trunc_normal_(self.relative_position_bias_table, std=0.02)
        self.softmax = nn.Softmax(dim=-1)

    def forward(self, x, mask=None):
        """Forward function.
        Args:
            x: input features with shape of (num_windows*B, N, C)
            mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None
        """
        B_, N, C = x.shape
        qkv = (
            self.qkv(x)
            .reshape(B_, N, 3, self.num_heads, C // self.num_heads)
            .permute(2, 0, 3, 1, 4)
        )
        # make torchscript happy (cannot use tensor as tuple)
        q, k, v = qkv[0], qkv[1], qkv[2]

        q = q * self.scale
        attn = q @ k.transpose(-2, -1)

        relative_position_bias = self.relative_position_bias_table[
            self.relative_position_index.view(-1)
        ].view(
            self.window_size[0] * self.window_size[1],
            self.window_size[0] * self.window_size[1],
            -1,
        )  # Wh*Ww,Wh*Ww,nH
        relative_position_bias = relative_position_bias.permute(
            2, 0, 1
        ).contiguous()  # nH, Wh*Ww, Wh*Ww
        attn = attn + relative_position_bias.unsqueeze(0)

        if mask is not None:
            nW = mask.shape[0]
            attn = attn.view(
                B_ // nW, nW, self.num_heads, N, N
            ) + mask.unsqueeze(1).unsqueeze(0)
            attn = attn.view(-1, self.num_heads, N, N)
            attn = self.softmax(attn)
        else:
            attn = self.softmax(attn)

        attn = self.attn_drop(attn)

        x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x


class ReversibleSwinTransformerBlock(nn.Module):
    """Reversible Swin Transformer Block.
    Args:
        dim (int): Number of input channels.
        num_heads (int): Number of attention heads.
        window_size (int): Window size.
        shift_size (int): Shift size for SW-MSA.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float, optional): Stochastic depth rate. Default: 0.0
        act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
        enable_amp (bool, optional): Enable mixed precision.  Default: False.
    """

    def __init__(
        self,
        dim,
        num_heads,
        window_size=7,
        shift_size=0,
        mlp_ratio=4.0,
        qkv_bias=True,
        qk_scale=None,
        drop=0.0,
        attn_drop=0.0,
        drop_path=0.0,
        act_layer=nn.GELU,
        norm_layer=nn.LayerNorm,
        enable_amp=False,
    ):
        super().__init__()
        self.dim = dim
        self.num_heads = num_heads
        self.window_size = window_size
        self.shift_size = shift_size
        self.mlp_ratio = mlp_ratio
        assert (
            0 <= self.shift_size < self.window_size
        ), "shift_size must in 0-window_size"
        self.norm1 = norm_layer(dim)
        self.attn = WindowAttention(
            dim,
            window_size=to_2tuple(self.window_size),
            num_heads=num_heads,
            qkv_bias=qkv_bias,
            qk_scale=qk_scale,
            attn_drop=attn_drop,
            proj_drop=drop,
        )

        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,
        )

        self.enable_amp = enable_amp
        self.H = None
        self.W = None

        self.seeds = {}

    def seed_cuda(self, key):
        """
        Fix seeds to allow for stochastic elements such as
        dropout to be reproduced exactly in activation
        recomputation in the backward pass.

        From RevViT.
        """

        # randomize seeds
        # use cuda generator if available
        if (
            hasattr(torch.cuda, "default_generators")
            and len(torch.cuda.default_generators) > 0
        ):
            # GPU
            device_idx = torch.cuda.current_device()
            seed = torch.cuda.default_generators[device_idx].seed()
        else:
            # CPU
            seed = int(torch.seed() % sys.maxsize)

        self.seeds[key] = seed
        torch.manual_seed(self.seeds[key])

    def F(self, x, mask_matrix):
        """Forward function for windowed attention. Implements an extra mask_matrix.
        No drop_path.
         Args:
            x: Input feature, tensor size (B, H*W, C).
            H, W: Spatial resolution of the input feature.
            mask_matrix: Attention mask for cyclic shift.
        """
        B, L, C = x.shape
        H, W = self.H, self.W
        assert L == H * W, "input feature has wrong size"

        with torch.cuda.amp.autocast(enabled=self.enable_amp):
            x = self.norm1(x)
            x = x.view(B, H, W, C)

            # pad feature maps to multiples of window size
            pad_l = pad_t = 0
            pad_r = (self.window_size - W % self.window_size) % self.window_size
            pad_b = (self.window_size - H % self.window_size) % self.window_size
            x = F.pad(x, (0, 0, pad_l, pad_r, pad_t, pad_b))
            _, Hp, Wp, _ = x.shape

            # cyclic shift
            if self.shift_size > 0:
                attn_mask = mask_matrix
                shifted_x = torch.roll(
                    x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)
                )
                # partition windows
                x_windows = window_partition(
                    shifted_x, self.window_size
                )  # nW*B, window_size, window_size, C
            else:
                attn_mask = None
                shifted_x = x
                # partition windows
                x_windows = window_partition(
                    shifted_x, self.window_size
                )  # nW*B, window_size, window_size, C

            x_windows = x_windows.view(
                -1, self.window_size * self.window_size, C
            )  # nW*B, window_size*window_size, C

            # W-MSA/SW-MSA
            # nW*B, window_size*window_size, C
            attn_windows = self.attn(x_windows, mask=attn_mask)

            # merge windows
            attn_windows = attn_windows.view(
                -1, self.window_size, self.window_size, C
            )
            shifted_x = window_reverse(
                attn_windows, self.window_size, Hp, Wp
            )  # B H' W' C

            # reverse cyclic shift
            if self.shift_size > 0:
                shifted_x = window_reverse(
                    attn_windows, self.window_size, H, W
                )  # B H' W' C
                x = torch.roll(
                    shifted_x,
                    shifts=(self.shift_size, self.shift_size),
                    dims=(1, 2),
                )
            else:
                shifted_x = window_reverse(
                    attn_windows, self.window_size, H, W
                )  # B H' W' C
            x = shifted_x

        if pad_r > 0 or pad_b > 0:
            x = x[:, :H, :W, :].contiguous()

        x = x.view(B, H * W, C)

        return x

    def G(self, x):
        """Forward function for mlp."""
        with torch.cuda.amp.autocast(enabled=self.enable_amp):
            x_out = self.mlp(self.norm2(x))
        return x_out

    def forward(self, X1, X2, mask_matrix):
        """Reversible forward function with rewiring.

        Y_1 = X_1 + Attn(X_2)
        Y_2 = X_2 + MLP(Y_1)

        Args:
            x1, x2: Input feature, tensor size (B, H*W, C).
            H, W: Spatial resolution of the input feature.
            mask_matrix: Attention mask for cyclic shift.
        """
        assert X1.shape == X2.shape, "Input shapes are different."

        self.seed_cuda("attn")
        attn_out = self.F(X2, mask_matrix)

        self.seed_cuda("droppath")
        Y1 = X1 + self.drop_path(attn_out)

        # Free memory
        del X1

        self.seed_cuda("mlp")
        mlp_out = self.G(Y1)

        torch.manual_seed(self.seeds["droppath"])
        Y2 = X2 + self.drop_path(mlp_out)

        del X2

        return Y1, Y2

    def backward_pass(self, Y_1, Y_2, dY_1, dY_2, mask_matrix):
        """
        equations for recovering activations:
        X2 = Y2 - MLP(Y1)
        X1 = Y1 - Attn(X2)
        """

        # temporarily record intermediate activation for G
        # and use them for gradient calculcation of G
        with torch.enable_grad():
            Y_1.requires_grad = True

            torch.manual_seed(self.seeds["mlp"])
            g_Y_1 = self.G(Y_1)

            torch.manual_seed(self.seeds["droppath"])
            g_Y_1 = self.drop_path(g_Y_1)

            g_Y_1.backward(dY_2, retain_graph=True)

        # activation recomputation is by design and not part of
        # the computation graph in forward pass.
        with torch.no_grad():
            X_2 = Y_2 - g_Y_1
            del g_Y_1

            dY_1 = dY_1 + Y_1.grad
            Y_1.grad = None

        # record F activations and calc gradients on F
        with torch.enable_grad():
            X_2.requires_grad = True

            torch.manual_seed(self.seeds["attn"])
            f_X_2 = self.F(X_2, mask_matrix)

            torch.manual_seed(self.seeds["droppath"])
            f_X_2 = self.drop_path(f_X_2)

            f_X_2.backward(dY_1, retain_graph=True)

        # propagate reverse computed acitvations at the start of
        # the previou block for backprop.s
        with torch.no_grad():
            X_1 = Y_1 - f_X_2

            del f_X_2, Y_1
            dY_2 = dY_2 + X_2.grad

            X_2.grad = None
            X_2 = X_2.detach()

        return X_1, X_2, dY_1, dY_2

    def backward_pass_recover(self, Y_1, Y_2, mask_matrix):
        """
        Use equations to recover activations and return them.
        Used for streaming the backward pass.
        """
        with torch.enable_grad():
            Y_1.requires_grad = True

            torch.manual_seed(self.seeds["mlp"])
            g_Y_1 = self.G(Y_1)

            torch.manual_seed(self.seeds["droppath"])
            g_Y_1 = self.drop_path(g_Y_1)

        with torch.no_grad():
            X_2 = Y_2 - g_Y_1

        with torch.enable_grad():
            X_2.requires_grad = True

            torch.manual_seed(self.seeds["attn"])
            f_X_2 = self.F(X_2, mask_matrix)

            torch.manual_seed(self.seeds["droppath"])
            f_X_2 = self.drop_path(f_X_2)

        with torch.no_grad():
            X_1 = Y_1 - f_X_2

        # Keep tensors around to do backprop on the graph.
        ctx = [X_1, X_2, Y_1, g_Y_1, f_X_2]
        return ctx

    def backward_pass_grads(self, X_1, X_2, Y_1, g_Y_1, f_X_2, dY_1, dY_2):
        """
        Receive intermediate activations and inputs to backprop through.
        """

        with torch.enable_grad():
            g_Y_1.backward(dY_2)

        with torch.no_grad():
            dY_1 = dY_1 + Y_1.grad
            Y_1.grad = None

        with torch.enable_grad():
            f_X_2.backward(dY_1)

        with torch.no_grad():
            dY_2 = dY_2 + X_2.grad
            X_2.grad = None
            X_2.detach()

        return dY_1, dY_2


class PatchMerging(nn.Module):
    """Patch Merging Layer
    Args:
        dim (int): Number of input channels.
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
    """

    def __init__(self, dim, norm_layer=nn.LayerNorm):
        super().__init__()
        self.dim = dim
        self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)
        self.norm = norm_layer(4 * dim)

    def forward(self, x, H, W):
        """Forward function.
        Args:
            x: Input feature, tensor size (B, H*W, C).
            H, W: Spatial resolution of the input feature.
        """
        B, L, C = x.shape
        assert L == H * W, "input feature has wrong size"

        x = x.view(B, H, W, C)

        # padding
        pad_input = (H % 2 == 1) or (W % 2 == 1)
        if pad_input:
            x = F.pad(x, (0, 0, 0, W % 2, 0, H % 2))

        x0 = x[:, 0::2, 0::2, :]  # B H/2 W/2 C
        x1 = x[:, 1::2, 0::2, :]  # B H/2 W/2 C
        x2 = x[:, 0::2, 1::2, :]  # B H/2 W/2 C
        x3 = x[:, 1::2, 1::2, :]  # B H/2 W/2 C
        x = torch.cat([x0, x1, x2, x3], -1)  # B H/2 W/2 4*C
        x = x.view(B, -1, 4 * C)  # B H/2*W/2 4*C

        x = self.norm(x)
        x = self.reduction(x)

        return x


class RevBackProp(Function):
    """
    Custom Backpropagation function to allow (A) flushing memory in foward
    and (B) activation recomputation reversibly in backward for gradient calculation.

    Inspired by https://github.com/RobinBruegger/RevTorch/blob/master/revtorch/revtorch.py
    """

    @staticmethod
    def forward(ctx, x, blocks, buffer):
        """
        Reversible Forward pass. Any intermediate activations from `buffer_layers` are
        cached in ctx for forward pass. This is not necessary for standard usecases.
        Each reversible layer implements its own forward pass logic.
        """

        H, W, attn_mask = buffer

        X_1, X_2 = torch.chunk(x, 2, dim=-1)
        for _, blk in enumerate(blocks):
            blk.H, blk.W = H, W
            X_1, X_2 = blk(X_1, X_2, attn_mask)

        all_tensors = [X_1.detach(), X_2.detach()]
        ctx.save_for_backward(*all_tensors)
        ctx.blocks = blocks
        ctx.attn_mask = attn_mask

        return torch.cat([X_1, X_2], dim=-1)

    @staticmethod
    def backward(ctx, dx):
        """
        Reversible Backward pass. Any intermediate activations from `buffer_layers` are
        recovered from ctx. Each layer implements its own logic for backward pass (both
        activation recomputation and grad calculation).
        """
        dX_1, dX_2 = torch.chunk(dx, 2, dim=-1)

        # retrieve params from ctx for backward
        X_1, X_2 = ctx.saved_tensors
        blocks = ctx.blocks
        attn_mask = ctx.attn_mask

        for _, blk in enumerate(blocks[::-1]):
            X_1, X_2, dX_1, dX_2 = blk.backward_pass(
                Y_1=X_1,
                Y_2=X_2,
                dY_1=dX_1,
                dY_2=dX_2,
                mask_matrix=attn_mask,
            )

        dx = torch.cat([dX_1, dX_2], dim=-1)

        del dX_1, dX_2, X_1, X_2

        return dx, None, None


class RevBackPropFast(RevBackProp):
    @staticmethod
    def backward(ctx, dx):
        """Overwrite backward by using PyTorch Streams to parallelize."""

        dX_1, dX_2 = torch.chunk(dx, 2, dim=-1)

        # retrieve params from ctx for backward
        X_1, X_2, *int_tensors = ctx.saved_tensors

        layers = ctx.blocks
        attn_mask = ctx.attn_mask

        # Keep a dictionary of events to synchronize on
        # Each is for the completion of a recalculation (f) or gradient calculation (b)
        events = {}
        for i in range(len(layers)):
            events[f"f{i}"] = torch.cuda.Event()
            events[f"b{i}"] = torch.cuda.Event()

        # Run backward staggered on two streams, which were defined globally in __init__
        # Initial pass
        with torch.cuda.stream(s1):
            layer = layers[-1]
            prev = layer.backward_pass_recover(
                Y_1=X_1, Y_2=X_2, mask_matrix=attn_mask
            )

            events["f0"].record(s1)

        # Stagger streams based on iteration
        for i, (this_layer, next_layer) in enumerate(
            zip(layers[1:][::-1], layers[:-1][::-1])
        ):
            if i % 2 == 0:
                stream1 = s1
                stream2 = s2
            else:
                stream1 = s2
                stream2 = s1

            with torch.cuda.stream(stream1):
                # b{i} waits on b{i-1}
                if i > 0:
                    events[f"b{i-1}"].synchronize()

                if i % 2 == 0:
                    dY_1, dY_2 = this_layer.backward_pass_grads(
                        *prev, dX_1, dX_2
                    )
                else:
                    dX_1, dX_2 = this_layer.backward_pass_grads(
                        *prev, dY_1, dY_2
                    )

                events[f"b{i}"].record(stream1)

            with torch.cuda.stream(stream2):
                # f{i} waits on f{i-1}
                events[f"f{i}"].synchronize()

                prev = next_layer.backward_pass_recover(
                    Y_1=prev[0], Y_2=prev[1], mask_matrix=attn_mask
                )

                events[f"f{i+1}"].record(stream2)

        # Last iteration
        if len(layers) - 1 % 2 == 0:
            stream2 = s1
        else:
            stream2 = s2
        next_layer = layers[0]

        with torch.cuda.stream(stream2):
            # stream2.wait_event(events[f"b{len(layers)-2}_end"])
            events[f"b{len(layers)-2}"].synchronize()
            if len(layers) - 1 % 2 == 0:
                dY_1, dY_2 = next_layer.backward_pass_grads(*prev, dX_1, dX_2)
                dx = torch.cat([dY_1, dY_2], dim=-1)
            else:
                dX_1, dX_2 = next_layer.backward_pass_grads(*prev, dY_1, dY_2)
                dx = torch.cat([dX_1, dX_2], dim=-1)
            events[f"b{len(layers)-1}"].record(stream2)

        # Synchronize, for PyTorch 1.9
        torch.cuda.current_stream().wait_stream(s1)
        torch.cuda.current_stream().wait_stream(s2)
        events[f"b{len(layers)-1}"].synchronize()

        del int_tensors
        del dX_1, dX_2, dY_1, dY_2, X_1, X_2, prev[:]
        return dx, None, None


class ReversibleLayer(nn.Module):
    """A Reversible Swin Transformer layer for one stage.
    Args:
        dim (int): Number of feature channels
        depth (int): Depths of this stage.
        num_heads (int): Number of attention head.
        window_size (int): Local window size. Default: 7.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0
        norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
        downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None
        fast_backprop (bool, optional): Whether to use fast reversible backprop, i.e. PaReprop. Default: False
        enable_amp (bool, optional): Enable mixed precision. Default: False
    """

    def __init__(
        self,
        dim,
        depth,
        num_heads,
        window_size=7,
        mlp_ratio=4.0,
        qkv_bias=True,
        qk_scale=None,
        drop=0.0,
        attn_drop=0.0,
        drop_path=0.0,
        norm_layer=nn.LayerNorm,
        downsample=None,
        fast_backprop=False,
        enable_amp=False,
    ):
        super().__init__()
        self.window_size = window_size
        self.shift_size = window_size // 2
        self.depth = depth

        # build blocks
        self.blocks = nn.ModuleList(
            [
                ReversibleSwinTransformerBlock(
                    dim=dim,
                    num_heads=num_heads,
                    window_size=window_size,
                    shift_size=0 if (i % 2 == 0) else window_size // 2,
                    mlp_ratio=mlp_ratio,
                    qkv_bias=qkv_bias,
                    qk_scale=qk_scale,
                    drop=drop,
                    attn_drop=attn_drop,
                    drop_path=drop_path[i]
                    if isinstance(drop_path, list)
                    else drop_path,
                    norm_layer=norm_layer,
                    enable_amp=enable_amp,
                )
                for i in range(depth)
            ]
        )

        # patch merging layer
        if downsample is not None:
            # lateral fusion strategy: input [B, L, 2*C] -> [B, L, C]
            self.lateral_fuse = TwoStreamFusion(dim=dim)
            self.downsample = downsample(dim=dim, norm_layer=norm_layer)
        else:
            self.downsample = None

        self.use_vanilla_backward = False
        self.use_fast_backprop = fast_backprop

    @staticmethod
    def vanilla_backward(h, blocks, buffer):
        """
        Use rev layers without rev backprop, for debugging. Use w/ self.use_vanilla_backward
        """
        # split into hidden states (h) and attention_output (a)
        H, W, attn_mask = buffer

        h, a = torch.chunk(h, 2, dim=-1)
        for _, blk in enumerate(blocks):
            blk.H, blk.W = H, W
            a, h = blk(a, h, attn_mask)

        return torch.cat([a, h], dim=-1)  # fuse by averaging

    def forward(self, x, H, W):
        """Forward function.
        Args:
            x: Input feature, tensor size (B, H*W, C).
            H, W: Spatial resolution of the input feature.
        """

        # calculate attention mask for SW-MSA
        Hp = int(np.ceil(H / self.window_size)) * self.window_size
        Wp = int(np.ceil(W / self.window_size)) * self.window_size
        img_mask = torch.zeros((1, Hp, Wp, 1), device=x.device)  # 1 Hp Wp 1
        h_slices = (
            slice(0, -self.window_size),
            slice(-self.window_size, -self.shift_size),
            slice(-self.shift_size, None),
        )
        w_slices = (
            slice(0, -self.window_size),
            slice(-self.window_size, -self.shift_size),
            slice(-self.shift_size, None),
        )
        cnt = 0
        for h in h_slices:
            for w in w_slices:
                img_mask[:, h, w, :] = cnt
                cnt += 1

        mask_windows = window_partition(
            img_mask, self.window_size
        )  # nW, window_size, window_size, 1
        mask_windows = mask_windows.view(
            -1, self.window_size * self.window_size
        )
        attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
        attn_mask = attn_mask.masked_fill(
            attn_mask != 0, float(-100.0)
        ).masked_fill(attn_mask == 0, float(0.0))

        x = torch.cat([x, x], dim=-1)

        if self.use_vanilla_backward:
            executing_fn = ReversibleLayer.vanilla_backward
        elif self.use_fast_backprop:
            executing_fn = RevBackPropFast.apply
        else:
            executing_fn = RevBackProp.apply

        x = executing_fn(x, self.blocks, (H, W, attn_mask))

        if self.downsample is not None:
            # Fuses [B, L, 2C] (from 2 streams) to [B, L, C]
            x = self.lateral_fuse(x)
            x_down = self.downsample(x, H, W)
            Wh, Ww = (H + 1) // 2, (W + 1) // 2
            return x_down, Wh, Ww
            # return x, H, W, x_down, Wh, Ww
        else:
            return x, H, W
            # return x, H, W, x, H, W


class PatchEmbed(nn.Module):
    """Image to Patch Embedding
    Args:
        patch_size (int): Patch token size. Default: 4.
        in_chans (int): Number of input image channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
        norm_layer (nn.Module, optional): Normalization layer. Default: None
    """

    def __init__(self, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None):
        super().__init__()
        patch_size = to_2tuple(patch_size)
        self.patch_size = patch_size

        self.in_chans = in_chans
        self.embed_dim = embed_dim

        self.proj = nn.Conv2d(
            in_chans, embed_dim, kernel_size=patch_size, stride=patch_size
        )
        if norm_layer is not None:
            self.norm = norm_layer(embed_dim)
        else:
            self.norm = None

    def forward(self, x):
        """Forward function."""
        # padding
        _, _, H, W = x.size()
        if W % self.patch_size[1] != 0:
            x = F.pad(x, (0, self.patch_size[1] - W % self.patch_size[1]))
        if H % self.patch_size[0] != 0:
            x = F.pad(x, (0, 0, 0, self.patch_size[0] - H % self.patch_size[0]))

        x = self.proj(x)  # B C Wh Ww
        if self.norm is not None:
            Wh, Ww = x.size(2), x.size(3)
            x = x.flatten(2).transpose(1, 2)
            x = self.norm(x)
            x = x.transpose(1, 2).view(-1, self.embed_dim, Wh, Ww)

        return x


class ReversibleSwinTransformer(nn.Module):
    """Reversible Swin Transformer backbone.
        A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows`  -
          https://arxiv.org/pdf/2103.14030
    Args:
        img_size (int): Input image size for training the pretrained model,
            used in absolute postion embedding. Default 224.
        patch_size (int | tuple(int)): Patch size. Default: 4.
        in_chans (int): Number of input image channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
        depths (tuple[int]): Depths of each Swin Transformer stage.
        num_heads (tuple[int]): Number of attention head of each stage.
        window_size (int): Window size. Default: 7.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4.
        qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float): Override default qk scale of head_dim ** -0.5 if set.
        drop_rate (float): Dropout rate.
        attn_drop_rate (float): Attention dropout rate. Default: 0.
        drop_path_rate (float): Stochastic depth rate. Default: 0.2.
        norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
        ape (bool): If True, add absolute position embedding to the patch embedding. Default: False.
        patch_norm (bool): If True, add normalization after patch embedding. Default: True.
        enable_amp (bool): False by default, no amp supported yet.
    """

    def __init__(
        self,
        img_size=224,
        patch_size=4,
        in_chans=3,
        num_classes=1000,
        embed_dim=96,
        depths=[2, 2, 6, 2],
        num_heads=[3, 6, 12, 24],
        window_size=7,
        mlp_ratio=4.0,
        qkv_bias=True,
        qk_scale=None,
        drop_rate=0.0,
        attn_drop_rate=0.0,
        drop_path_rate=0.2,
        norm_layer=nn.LayerNorm,
        ape=False,
        patch_norm=True,
        fast_backprop=False,
        enable_amp=False,
    ):
        super().__init__()

        self.img_size = img_size
        self.num_layers = len(depths)
        self.embed_dim = embed_dim
        self.ape = ape
        self.patch_norm = patch_norm
        # self.out_indices = out_indices
        # self.frozen_stages = frozen_stages

        # split image into non-overlapping patches
        self.patch_embed = PatchEmbed(
            patch_size=patch_size,
            in_chans=in_chans,
            embed_dim=embed_dim,
            norm_layer=norm_layer if self.patch_norm else None,
        )

        # absolute position embedding
        if self.ape:
            img_size = to_2tuple(img_size)
            patch_size = to_2tuple(patch_size)
            patches_resolution = [
                img_size[0] // patch_size[0],
                img_size[1] // patch_size[1],
            ]

            self.absolute_pos_embed = nn.Parameter(
                torch.zeros(
                    1, embed_dim, patches_resolution[0], patches_resolution[1]
                )
            )
            trunc_normal_(self.absolute_pos_embed, std=0.02)

        self.pos_drop = nn.Dropout(p=drop_rate)

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

        # Initialize global stream for device
        if fast_backprop:
            global s1, s2
            s1 = torch.cuda.default_stream(device=torch.cuda.current_device())
            s2 = torch.cuda.Stream(device=torch.cuda.current_device())

        # build layers
        self.layers = nn.ModuleList()
        for i_layer in range(self.num_layers):
            layer = ReversibleLayer(
                dim=int(embed_dim * 2**i_layer),
                depth=depths[i_layer],
                num_heads=num_heads[i_layer],
                window_size=window_size,
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                qk_scale=qk_scale,
                drop=drop_rate,
                attn_drop=attn_drop_rate,
                drop_path=dpr[
                    sum(depths[:i_layer]) : sum(depths[: i_layer + 1])
                ],
                norm_layer=norm_layer,
                downsample=PatchMerging
                if (i_layer < self.num_layers - 1)
                else None,
                fast_backprop=fast_backprop,
                enable_amp=enable_amp,
            )
            self.layers.append(layer)

        # num_features = [int(embed_dim * 2 ** i)
        #                 for i in range(self.num_layers)]
        # self.num_features = num_features
        # if not rev, then self.num_features = [int(embed_dim * 2 ** (self.num_layers - 1))]
        # but rev has 2x channels, so features is * 2 as last layer isn't downsampled
        self.num_features = int(embed_dim * 2**self.num_layers)
        self.mlp_ratio = mlp_ratio

        self.norm = norm_layer(self.num_features)
        self.avgpool = nn.AdaptiveAvgPool1d(1)
        self.head = (
            nn.Linear(self.num_features, num_classes)
            if num_classes > 0
            else nn.Identity()
        )

        self.apply(self._init_weights)

    # def _freeze_stages(self):
    #     if self.frozen_stages >= 0:
    #         self.patch_embed.eval()
    #         for param in self.patch_embed.parameters():
    #             param.requires_grad = False

    #     if self.frozen_stages >= 1 and self.ape:
    #         self.absolute_pos_embed.requires_grad = False

    #     if self.frozen_stages >= 2:
    #         self.pos_drop.eval()
    #         for i in range(0, self.frozen_stages - 1):
    #             m = self.layers[i]
    #             m.eval()
    #             for param in m.parameters():
    #                 param.requires_grad = False

    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, 0.02)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0.02)
            nn.init.constant_(m.weight, 1.0)

    def forward_features(self, x):
        """Forward function."""
        x = self.patch_embed(x)

        Wh, Ww = x.size(2), x.size(3)
        if self.ape:
            # interpolate the position embedding to the corresponding size
            absolute_pos_embed = F.interpolate(
                self.absolute_pos_embed, size=(Wh, Ww), mode="bicubic"
            )
            x = (x + absolute_pos_embed).flatten(2).transpose(1, 2)  # B Wh*Ww C
        else:
            x = x.flatten(2).transpose(1, 2)
        x = self.pos_drop(x)

        for i in range(self.num_layers):
            layer = self.layers[i]
            x, Wh, Ww = layer(x, Wh, Ww)

        x = self.norm(x)
        x = self.avgpool(x.transpose(1, 2))  # B C 1
        x = torch.flatten(x, 1)
        return x

    def forward(self, x):
        x = self.forward_features(x)
        x = self.head(x)
        return x