AllinonSAM/prompt_adapted_segment_anything/modeling/image_encoder.py

# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.

# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import math
from functools import partial, reduce
from operator import mul
import torch
import torch.nn as nn
import torch.nn.functional as F

from typing import Optional, Tuple, Type

from .common import LayerNorm2d, MLPBlock
from .svd_layers import SVDLinear, SVDConv2d
# from .SALT_layers import SALTLinear , SALTConv2d # SALT-LoRA
# from .SALT_layers_please_work import SALTLinear , SALTConv2d #SALT-2
from .SALT_layers_3 import SALTLinear , SALTConv2d # SALT-1
from .lora_layers import LoRAConv2D, LoRALinear


# This class and its supporting functions below lightly adapted from the ViTDet backbone available at: https://github.com/facebookresearch/detectron2/blob/main/detectron2/modeling/backbone/vit.py # noqa
class ImageEncoderViT(nn.Module):
    def __init__(
        self,
        img_size: int = 1024,
        patch_size: int = 16,
        in_chans: int = 3,
        embed_dim: int = 768,
        depth: int = 12,
        num_heads: int = 12,
        mlp_ratio: float = 4.0,
        out_chans: int = 256,
        qkv_bias: bool = True,
        norm_layer: Type[nn.Module] = nn.LayerNorm,
        act_layer: Type[nn.Module] = nn.GELU,
        use_abs_pos: bool = True,
        use_rel_pos: bool = False,
        rel_pos_zero_init: bool = True,
        window_size: int = 0,
        global_attn_indexes: Tuple[int, ...] = (),
        prompt_config = {
            'USE_PROMPT': False,
            'LOCATION': 'prepend',
            'DROPOUT': 0.1,
            'NUM_TOKENS': 5
        },
        mlp_transform = False,
        use_lora=False
    ) -> None:
        """
        Args:
            img_size (int): Input image size.
            patch_size (int): Patch size.
            in_chans (int): Number of input image channels.
            embed_dim (int): Patch embedding dimension.
            depth (int): Depth of ViT.
            num_heads (int): Number of attention heads in each ViT block.
            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
            qkv_bias (bool): If True, add a learnable bias to query, key, value.
            norm_layer (nn.Module): Normalization layer.
            act_layer (nn.Module): Activation layer.
            use_abs_pos (bool): If True, use absolute positional embeddings.
            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
            window_size (int): Window size for window attention blocks.
            global_attn_indexes (list): Indexes for blocks using global attention.
        """
        super().__init__()
        self.img_size = img_size
        self.embed_dim = embed_dim
        self.patch_size = (patch_size,patch_size)
        self.prompt_config = prompt_config

        self.patch_embed = PatchEmbed(
            kernel_size=(patch_size, patch_size),
            stride=(patch_size, patch_size),
            in_chans=in_chans,
            embed_dim=embed_dim,
        )

        self.pos_embed: Optional[nn.Parameter] = None
        if use_abs_pos:
            # Initialize absolute positional embedding with pretrain image size.
            #if image prompts are used, flatten the embeds along the image size dimensions
            if self.prompt_config['USE_IMAGE_PROMPT']:

                self.pos_embed = nn.Parameter(
                    torch.zeros(1, (img_size // patch_size)* (img_size // patch_size), embed_dim)
                )
            else:
                self.pos_embed = nn.Parameter(
                    torch.zeros(1, (img_size // patch_size), (img_size // patch_size), embed_dim)
                )


        self.blocks = nn.ModuleList()
        for i in range(depth):
            block = Block(
                dim=embed_dim,
                num_heads=num_heads,
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                norm_layer=norm_layer,
                act_layer=act_layer,
                use_rel_pos=use_rel_pos,
                rel_pos_zero_init=rel_pos_zero_init,
                window_size=window_size if i not in global_attn_indexes else 0,
                input_size=(img_size // patch_size, img_size // patch_size),
                mlp_transform=mlp_transform,
                use_lora = use_lora
            )
            self.blocks.append(block)

        self.neck = Neck(embed_dim, out_chans, mlp_transform = mlp_transform, use_lora=use_lora)
        # self.neck = nn.Sequential(
        #     SVDConv2d(
        #         embed_dim,
        #         out_chans,
        #         kernel_size=1,
        #         bias=False,
        #     ),
        #     LayerNorm2d(out_chans),
        #     SVDConv2d(
        #         out_chans,
        #         out_chans,
        #         kernel_size=3,
        #         padding=1,
        #         bias=False,
        #     ),
        #     LayerNorm2d(out_chans),
        # )
        if self.prompt_config['USE_IMAGE_PROMPT']:
            val = math.sqrt(6. / float(3 * reduce(mul, self.patch_size, 1) + self.embed_dim))  # noqa
            self.prompt_dropout = nn.Dropout(self.prompt_config['DROPOUT'])
            self.prompt_embeddings = nn.Parameter(torch.zeros(1, self.prompt_config['NUM_TOKENS'], self.embed_dim))
            nn.init.uniform_(self.prompt_embeddings.data, -val,val)

            self.deep_prompt_embeddings = nn.Parameter(torch.zeros(
                len(self.blocks) - 1,
                self.prompt_config['NUM_TOKENS'],
                self.embed_dim
            ))
            nn.init.uniform_(self.deep_prompt_embeddings.data, -val, val)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = self.patch_embed(x)
        reg_loss = 0
        if self.pos_embed is not None:
            if self.prompt_config['USE_IMAGE_PROMPT']:
                x = x + self.pos_embed
            else:
                p1,p2,p3,p4 = self.pos_embed.shape
                x = x + self.pos_embed.view(p1,p2*p3,p4)

        B = x.shape[0]
        if self.prompt_config['USE_IMAGE_PROMPT']:
            x = self.incorporate_prompt(x)
            B = x.shape[0]
            # print("x shape: ",x.shape)
            num_layers = len(self.blocks)
            for i in range(num_layers):
                if i==0:
                    x, loss = self.blocks[i](x)
                    reg_loss += loss
                else:
                    x = torch.cat((
                        x[:,:1,:],
                        self.prompt_dropout(self.deep_prompt_embeddings[i-1].expand(B,-1,-1)),
                        x[:,(1+self.prompt_config['NUM_TOKENS']):,:]
                    ), dim=1)
                    x, loss = self.blocks[i](x)
                    reg_loss += loss
            
            x = torch.cat((
            x[:,:1,:],
            x[:,(1+self.prompt_config['NUM_TOKENS']):,:]
        ), dim=1)
        else:
            for blk in self.blocks:
                x, loss = blk(x)
                reg_loss += loss

        
        resize_dim = self.img_size // self.patch_size[0]
        x = x.view(B, resize_dim, resize_dim, -1)
        x = self.neck(x.permute(0, 3, 1, 2))

        return x, reg_loss

    def incorporate_prompt(self, x):
        B = x.shape[0]
        if self.prompt_config['LOCATION'] == 'prepend':
            x = torch.cat((
                x[:,:1,:],
                self.prompt_dropout(self.prompt_embeddings.expand(B,-1,-1)),
                x[:,1:,:]
            ), dim=1)
        else:
            raise ValueError("Other prompt location not supported")
        return x

class Neck(nn.Module):
    """Neck which is a MLP at the end"""
    def __init__(self, embed_dim, out_chans, mlp_transform=False, use_lora=False):
        super().__init__()
        if use_lora:
            self.conv1 = LoRAConv2D(embed_dim, out_chans, kernel_size=1, bias=False)
            self.conv2 = LoRAConv2D(out_chans, out_chans, kernel_size=3, padding=1, bias=False)
        else:
            rank_value = 150
            self.conv1 = SALTConv2d(embed_dim, out_chans, kernel_size=1, bias=False , rank=rank_value , r_lora=256 , rsLora=False , alpha=1)
            self.conv2 = SALTConv2d(out_chans, out_chans, kernel_size=3, padding=1, bias=False , rank=rank_value , r_lora=256 , rsLora=False, alpha=1)
            # self.conv1 = SVDConv2d(embed_dim, out_chans, kernel_size=1, bias=False, mlp_transform=mlp_transform)
            # self.conv2 = SVDConv2d(out_chans, out_chans, kernel_size=3, padding=1, bias=False, mlp_transform=mlp_transform)
        
        
        self.ln1 = LayerNorm2d(out_chans)
        self.ln2 = LayerNorm2d(out_chans)        

    def forward(self, x):
        out, reg_loss1 = self.conv1(x)
        out = self.ln1(out)
        out, reg_loss2 = self.conv2(out)
        out = self.ln2(out)
        return out


class Block(nn.Module):
    """Transformer blocks with support of window attention and residual propagation blocks"""

    def __init__(
        self,
        dim: int,
        num_heads: int,
        mlp_ratio: float = 4.0,
        qkv_bias: bool = True,
        norm_layer: Type[nn.Module] = nn.LayerNorm,
        act_layer: Type[nn.Module] = nn.GELU,
        use_rel_pos: bool = False,
        rel_pos_zero_init: bool = True,
        window_size: int = 0,
        input_size: Optional[Tuple[int, int]] = None,
        mlp_transform = False,
        use_lora = False
    ) -> None:
        """
        Args:
            dim (int): Number of input channels.
            num_heads (int): Number of attention heads in each ViT block.
            mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
            qkv_bias (bool): If True, add a learnable bias to query, key, value.
            norm_layer (nn.Module): Normalization layer.
            act_layer (nn.Module): Activation layer.
            use_rel_pos (bool): If True, add relative positional embeddings to the attention map.
            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
            window_size (int): Window size for window attention blocks. If it equals 0, then
                use global attention.
            input_size (int or None): Input resolution for calculating the relative positional
                parameter size.
        """
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim,
            num_heads=num_heads,
            qkv_bias=qkv_bias,
            use_rel_pos=use_rel_pos,
            rel_pos_zero_init=rel_pos_zero_init,
            input_size=input_size if window_size == 0 else (window_size, window_size),
            mlp_transform = mlp_transform,
            use_lora = use_lora
        )

        self.norm2 = norm_layer(dim)
        self.mlp = MLPBlock(embedding_dim=dim, mlp_dim=int(dim * mlp_ratio), act=act_layer, mlp_transform=mlp_transform, use_lora=use_lora)

        self.window_size = window_size

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        shortcut = x
        x = self.norm1(x)
        # Window partition
        if self.window_size > 0:
            H, W = x.shape[1], x.shape[2]
            x, pad_hw = window_partition(x, self.window_size)

        x, reg_loss1 = self.attn(x)
        # Reverse window partition
        if self.window_size > 0:
            x = window_unpartition(x, self.window_size, pad_hw, (H, W))

        x = shortcut + x
        mlp_out, reg_loss2 = self.mlp(self.norm2(x))
        x = x + mlp_out

        return x, (reg_loss1 + reg_loss2)


class Attention(nn.Module):
    """Multi-head Attention block with relative position embeddings."""

    def __init__(
        self,
        dim: int,
        num_heads: int = 8,
        qkv_bias: bool = True,
        use_rel_pos: bool = False,
        rel_pos_zero_init: bool = True,
        input_size: Optional[Tuple[int, int]] = None,
        mlp_transform = False,
        use_lora=False
    ) -> None:
        """
        Args:
            dim (int): Number of input channels.
            num_heads (int): Number of attention heads.
            qkv_bias (bool:  If True, add a learnable bias to query, key, value.
            rel_pos (bool): If True, add relative positional embeddings to the attention map.
            rel_pos_zero_init (bool): If True, zero initialize relative positional parameters.
            input_size (int or None): Input resolution for calculating the relative positional
                parameter size.
        """
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = head_dim**-0.5
        if use_lora:
            self.qkv = LoRALinear(dim, dim * 3, bias=qkv_bias)
            self.proj = LoRALinear(dim, dim)
        else:
            rank_value = 500
            self.qkv = SALTLinear(dim, dim * 3, bias=qkv_bias , r_lora=256 , rank=rank_value , rsLora=False,alpha=1)
            self.proj = SALTLinear(dim, dim , r_lora=256 , rank=rank_value , rsLora=False,alpha=1)
            # self.qkv = SVDLinear(dim, dim * 3, bias=qkv_bias, mlp_transform=mlp_transform)
            # self.proj = SVDLinear(dim, dim, mlp_transform=mlp_transform)

        self.use_rel_pos = use_rel_pos
        if self.use_rel_pos:
            assert (
                input_size is not None
            ), "Input size must be provided if using relative positional encoding."
            # initialize relative positional embeddings
            self.rel_pos_h = nn.Parameter(torch.zeros(2 * input_size[0] - 1, head_dim))
            self.rel_pos_w = nn.Parameter(torch.zeros(2 * input_size[1] - 1, head_dim))

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        B, HW, _ = x.shape
        # qkv with shape (3, B, nHead, H * W, C)
        qkv, reg_loss1 = self.qkv(x)
        qkv = qkv.reshape(B, HW, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
        # q, k, v with shape (B * nHead, H * W, C)
        q, k, v = qkv.reshape(3, B * self.num_heads, HW, -1).unbind(0)

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

        # if self.use_rel_pos:
        #     attn = add_decomposed_rel_pos(attn, q, self.rel_pos_h, self.rel_pos_w, (H, W), (H, W))

        attn = attn.softmax(dim=-1)
        x = (attn @ v).view(B, self.num_heads, HW, -1).permute(0, 2, 1, 3).reshape(B, HW, -1)
        x, reg_loss2 = self.proj(x)

        return x, (reg_loss2 + reg_loss1)


def window_partition(x: torch.Tensor, window_size: int) -> Tuple[torch.Tensor, Tuple[int, int]]:
    """
    Partition into non-overlapping windows with padding if needed.
    Args:
        x (tensor): input tokens with [B, H, W, C].
        window_size (int): window size.

    Returns:
        windows: windows after partition with [B * num_windows, window_size, window_size, C].
        (Hp, Wp): padded height and width before partition
    """
    B, H, W, C = x.shape

    pad_h = (window_size - H % window_size) % window_size
    pad_w = (window_size - W % window_size) % window_size
    if pad_h > 0 or pad_w > 0:
        x = F.pad(x, (0, 0, 0, pad_w, 0, pad_h))
    Hp, Wp = H + pad_h, W + pad_w

    x = x.view(B, Hp // window_size, window_size, Wp // window_size, window_size, C)
    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
    return windows, (Hp, Wp)


def window_unpartition(
    windows: torch.Tensor, window_size: int, pad_hw: Tuple[int, int], hw: Tuple[int, int]
) -> torch.Tensor:
    """
    Window unpartition into original sequences and removing padding.
    Args:
        x (tensor): input tokens with [B * num_windows, window_size, window_size, C].
        window_size (int): window size.
        pad_hw (Tuple): padded height and width (Hp, Wp).
        hw (Tuple): original height and width (H, W) before padding.

    Returns:
        x: unpartitioned sequences with [B, H, W, C].
    """
    Hp, Wp = pad_hw
    H, W = hw
    B = windows.shape[0] // (Hp * Wp // window_size // window_size)
    x = windows.view(B, Hp // window_size, Wp // window_size, window_size, window_size, -1)
    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, Hp, Wp, -1)

    if Hp > H or Wp > W:
        x = x[:, :H, :W, :].contiguous()
    return x


def get_rel_pos(q_size: int, k_size: int, rel_pos: torch.Tensor) -> torch.Tensor:
    """
    Get relative positional embeddings according to the relative positions of
        query and key sizes.
    Args:
        q_size (int): size of query q.
        k_size (int): size of key k.
        rel_pos (Tensor): relative position embeddings (L, C).

    Returns:
        Extracted positional embeddings according to relative positions.
    """
    max_rel_dist = int(2 * max(q_size, k_size) - 1)
    # Interpolate rel pos if needed.
    if rel_pos.shape[0] != max_rel_dist:
        # Interpolate rel pos.
        rel_pos_resized = F.interpolate(
            rel_pos.reshape(1, rel_pos.shape[0], -1).permute(0, 2, 1),
            size=max_rel_dist,
            mode="linear",
        )
        rel_pos_resized = rel_pos_resized.reshape(-1, max_rel_dist).permute(1, 0)
    else:
        rel_pos_resized = rel_pos

    # Scale the coords with short length if shapes for q and k are different.
    q_coords = torch.arange(q_size)[:, None] * max(k_size / q_size, 1.0)
    k_coords = torch.arange(k_size)[None, :] * max(q_size / k_size, 1.0)
    relative_coords = (q_coords - k_coords) + (k_size - 1) * max(q_size / k_size, 1.0)

    return rel_pos_resized[relative_coords.long()]


def add_decomposed_rel_pos(
    attn: torch.Tensor,
    q: torch.Tensor,
    rel_pos_h: torch.Tensor,
    rel_pos_w: torch.Tensor,
    q_size: Tuple[int, int],
    k_size: Tuple[int, int],
) -> torch.Tensor:
    """
    Calculate decomposed Relative Positional Embeddings from :paper:`mvitv2`.
    https://github.com/facebookresearch/mvit/blob/19786631e330df9f3622e5402b4a419a263a2c80/mvit/models/attention.py   # noqa B950
    Args:
        attn (Tensor): attention map.
        q (Tensor): query q in the attention layer with shape (B, q_h * q_w, C).
        rel_pos_h (Tensor): relative position embeddings (Lh, C) for height axis.
        rel_pos_w (Tensor): relative position embeddings (Lw, C) for width axis.
        q_size (Tuple): spatial sequence size of query q with (q_h, q_w).
        k_size (Tuple): spatial sequence size of key k with (k_h, k_w).

    Returns:
        attn (Tensor): attention map with added relative positional embeddings.
    """
    q_h, q_w = q_size
    k_h, k_w = k_size
    Rh = get_rel_pos(q_h, k_h, rel_pos_h)
    Rw = get_rel_pos(q_w, k_w, rel_pos_w)

    B, _, dim = q.shape
    r_q = q.reshape(B, q_h, q_w, dim)
    rel_h = torch.einsum("bhwc,hkc->bhwk", r_q, Rh)
    rel_w = torch.einsum("bhwc,wkc->bhwk", r_q, Rw)

    attn = (
        attn.view(B, q_h, q_w, k_h, k_w) + rel_h[:, :, :, :, None] + rel_w[:, :, :, None, :]
    ).view(B, q_h * q_w, k_h * k_w)

    return attn


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

    def __init__(
        self,
        kernel_size: Tuple[int, int] = (16, 16),
        stride: Tuple[int, int] = (16, 16),
        padding: Tuple[int, int] = (0, 0),
        in_chans: int = 3,
        embed_dim: int = 768,
    ) -> None:
        """
        Args:
            kernel_size (Tuple): kernel size of the projection layer.
            stride (Tuple): stride of the projection layer.
            padding (Tuple): padding size of the projection layer.
            in_chans (int): Number of input image channels.
            embed_dim (int):  embed_dim (int): Patch embedding dimension.
        """
        super().__init__()

        self.proj = nn.Conv2d(
            in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = self.proj(x)
        # B C H W -> B H W C
        x = x.permute(0, 2, 3, 1)
        B,H,W,C = x.shape
        x = x.view((B,H*W,C))
        return x