featup/featurizers/MAE.py

from functools import partial

import numpy as np
import torch
import torch.nn as nn
import os
from timm.models.vision_transformer import Block
import torch.nn.functional as F


class PatchEmbed(nn.Module):
    """ 2D Image to Patch Embedding
    """

    def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768, norm_layer=None, flatten=True):
        super().__init__()
        img_size = (img_size, img_size)
        patch_size = (patch_size, patch_size)
        self.img_size = img_size
        self.patch_size = patch_size
        self.grid_size = (img_size[0] // patch_size[0], img_size[1] // patch_size[1])
        self.num_patches = self.grid_size[0] * self.grid_size[1]
        self.flatten = flatten

        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity()

    def forward(self, x):
        B, C, H, W = x.shape
        # assert H == self.img_size[0] and W == self.img_size[1], \
        #     f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})."
        x = self.proj(x)
        if self.flatten:
            x = x.flatten(2).transpose(1, 2)  # BCHW -> BNC
        x = self.norm(x)
        return x


def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):
    """
    grid_size: int of the grid height and width
    return:
    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)
    """
    grid_h = np.arange(grid_size, dtype=np.float32)
    grid_w = np.arange(grid_size, dtype=np.float32)
    grid = np.meshgrid(grid_w, grid_h)  # here w goes first
    grid = np.stack(grid, axis=0)

    grid = grid.reshape([2, 1, grid_size, grid_size])
    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
    if cls_token:
        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)
    return pos_embed


def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
    assert embed_dim % 2 == 0

    # use half of dimensions to encode grid_h
    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)
    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)

    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)
    return emb


def get_1d_sincos_pos_embed_from_grid(embed_dim, pos):
    """
    embed_dim: output dimension for each position
    pos: a list of positions to be encoded: size (M,)
    out: (M, D)
    """
    assert embed_dim % 2 == 0
    omega = np.arange(embed_dim // 2, dtype=np.float)
    omega /= embed_dim / 2.
    omega = 1. / 10000 ** omega  # (D/2,)

    pos = pos.reshape(-1)  # (M,)
    out = np.einsum('m,d->md', pos, omega)  # (M, D/2), outer product

    emb_sin = np.sin(out)  # (M, D/2)
    emb_cos = np.cos(out)  # (M, D/2)

    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)
    return emb


# --------------------------------------------------------
# Interpolate position embeddings for high-resolution
# References:
# DeiT: https://github.com/facebookresearch/deit
# --------------------------------------------------------
def interpolate_pos_embed(model, checkpoint_model):
    if 'pos_embed' in checkpoint_model:
        pos_embed_checkpoint = checkpoint_model['pos_embed']
        embedding_size = pos_embed_checkpoint.shape[-1]
        num_patches = model.patch_embed.num_patches
        num_extra_tokens = model.pos_embed.shape[-2] - num_patches
        # height (== width) for the checkpoint position embedding
        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)
        # height (== width) for the new position embedding
        new_size = int(num_patches ** 0.5)
        # class_token and dist_token are kept unchanged
        if orig_size != new_size:
            print("Position interpolate from %dx%d to %dx%d" % (orig_size, orig_size, new_size, new_size))
            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]
            # only the position tokens are interpolated
            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]
            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)
            pos_tokens = torch.nn.functional.interpolate(
                pos_tokens, size=(new_size, new_size), mode='bicubic', align_corners=False)
            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)
            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)
            checkpoint_model['pos_embed'] = new_pos_embed


def sample(t: torch.Tensor, coords: torch.Tensor):
    return F.grid_sample(t, coords.permute(0, 2, 1, 3), padding_mode='border', align_corners=True)


class MaskedAutoencoderViT(nn.Module):
    """ Masked Autoencoder with VisionTransformer backbone
    """

    def __init__(self, img_size=224, patch_size=16, in_chans=3,
                 embed_dim=1024, depth=24, num_heads=16,
                 decoder_embed_dim=512, decoder_depth=8, decoder_num_heads=16,
                 mlp_ratio=4., norm_layer=nn.LayerNorm, norm_pix_loss=False):
        super().__init__()

        # --------------------------------------------------------------------------
        # MAE encoder specifics
        self.embed_dim = embed_dim
        self.patch_embed = PatchEmbed(img_size, patch_size, in_chans, embed_dim)
        num_patches = self.patch_embed.num_patches

        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim),
                                      requires_grad=False)  # fixed sin-cos embedding

        self.blocks = nn.ModuleList([
            Block(embed_dim, num_heads, mlp_ratio, qkv_bias=True, norm_layer=norm_layer)
            for i in range(depth)])
        self.norm = norm_layer(embed_dim)
        # --------------------------------------------------------------------------

        # --------------------------------------------------------------------------
        # MAE decoder specifics
        self.decoder_embed = nn.Linear(embed_dim, decoder_embed_dim, bias=True)

        self.mask_token = nn.Parameter(torch.zeros(1, 1, decoder_embed_dim))

        self.decoder_pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, decoder_embed_dim),
                                              requires_grad=False)  # fixed sin-cos embedding

        self.decoder_blocks = nn.ModuleList([
            Block(decoder_embed_dim, decoder_num_heads, mlp_ratio, qkv_bias=True, norm_layer=norm_layer)
            for i in range(decoder_depth)])

        self.decoder_norm = norm_layer(decoder_embed_dim)
        self.decoder_pred = nn.Linear(decoder_embed_dim, patch_size ** 2 * in_chans, bias=True)  # decoder to patch
        # --------------------------------------------------------------------------

        self.norm_pix_loss = norm_pix_loss

        self.initialize_weights()

    def initialize_weights(self):
        # initialization
        # initialize (and freeze) pos_embed by sin-cos embedding
        pos_embed = get_2d_sincos_pos_embed(self.pos_embed.shape[-1], int(self.patch_embed.num_patches ** .5),
                                            cls_token=True)
        self.pos_embed.data.copy_(torch.from_numpy(pos_embed).float().unsqueeze(0))

        decoder_pos_embed = get_2d_sincos_pos_embed(self.decoder_pos_embed.shape[-1],
                                                    int(self.patch_embed.num_patches ** .5), cls_token=True)
        self.decoder_pos_embed.data.copy_(torch.from_numpy(decoder_pos_embed).float().unsqueeze(0))

        # initialize patch_embed like nn.Linear (instead of nn.Conv2d)
        w = self.patch_embed.proj.weight.data
        torch.nn.init.xavier_uniform_(w.view([w.shape[0], -1]))

        # timm's trunc_normal_(std=.02) is effectively normal_(std=0.02) as cutoff is too big (2.)
        torch.nn.init.normal_(self.cls_token, std=.02)
        torch.nn.init.normal_(self.mask_token, std=.02)

        # initialize nn.Linear and nn.LayerNorm
        self.apply(self._init_weights)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            # we use xavier_uniform following official JAX ViT:
            torch.nn.init.xavier_uniform_(m.weight)
            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)

    def patchify(self, imgs):
        """
        imgs: (N, 3, H, W)
        x: (N, L, patch_size**2 *3)
        """
        p = self.patch_embed.patch_size[0]
        assert imgs.shape[2] == imgs.shape[3] and imgs.shape[2] % p == 0

        h = w = imgs.shape[2] // p
        x = imgs.reshape(shape=(imgs.shape[0], 3, h, p, w, p))
        x = torch.einsum('nchpwq->nhwpqc', x)
        x = x.reshape(shape=(imgs.shape[0], h * w, p ** 2 * 3))
        return x

    def unpatchify(self, x):
        """
        x: (N, L, patch_size**2 *3)
        imgs: (N, 3, H, W)
        """
        p = self.patch_embed.patch_size[0]
        h = w = int(x.shape[1] ** .5)
        assert h * w == x.shape[1]

        x = x.reshape(shape=(x.shape[0], h, w, p, p, 3))
        x = torch.einsum('nhwpqc->nchpwq', x)
        imgs = x.reshape(shape=(x.shape[0], 3, h * p, h * p))
        return imgs

    def random_masking(self, x, mask_ratio):
        """
        Perform per-sample random masking by per-sample shuffling.
        Per-sample shuffling is done by argsort random noise.
        x: [N, L, D], sequence
        """
        N, L, D = x.shape  # batch, length, dim
        len_keep = int(L * (1 - mask_ratio))

        noise = torch.rand(N, L, device=x.device)  # noise in [0, 1]

        # sort noise for each sample
        ids_shuffle = torch.argsort(noise, dim=1)  # ascend: small is keep, large is remove
        ids_restore = torch.argsort(ids_shuffle, dim=1)

        # keep the first subset
        ids_keep = ids_shuffle[:, :len_keep]
        x_masked = torch.gather(x, dim=1, index=ids_keep.unsqueeze(-1).repeat(1, 1, D))

        # generate the binary mask: 0 is keep, 1 is remove
        mask = torch.ones([N, L], device=x.device)
        mask[:, :len_keep] = 0
        # unshuffle to get the binary mask
        mask = torch.gather(mask, dim=1, index=ids_restore)

        return x_masked, mask, ids_restore

    def sample_pe(self, img, pe):
        p = self.patch_embed.patch_size[0]

        H = img.shape[2] // p
        W = img.shape[3] // p

        original_num_patches = 224 // p
        embed_dim = pe.shape[-1]

        reshaped_pe = pe.squeeze(0)[1:] \
            .reshape(1, original_num_patches, original_num_patches, embed_dim) \
            .permute(0, 3, 1, 2)

        XX, YY = torch.meshgrid(torch.linspace(-1, 1, H, device=img.device, dtype=img.dtype),
                                torch.linspace(-1, 1, W, device=img.device, dtype=img.dtype))

        coords = torch.cat([XX.unsqueeze(-1), YY.unsqueeze(-1)], dim=-1).unsqueeze(0)

        return sample(reshaped_pe, coords).reshape(embed_dim, H * W).permute(1, 0).unsqueeze(0)

    def featurize(self, img, n_decoder_blocks=None):
        p = self.patch_embed.patch_size[0]
        H = img.shape[2] // p
        W = img.shape[3] // p

        # embed patches
        x = self.patch_embed(img)

        # add pos embed w/o cls token
        x = x + self.sample_pe(img, self.pos_embed)

        # append cls token
        cls_token = self.cls_token + self.pos_embed[:, :1, :]
        cls_tokens = cls_token.expand(x.shape[0], -1, -1)
        x = torch.cat((cls_tokens, x), dim=1)

        # apply Transformer blocks
        for blk in self.blocks:
            x = blk(x)
        x = self.norm(x)


        # embed tokens
        #x = self.decoder_embed(x)
        #
        # # add pos embed
        # cls_token = x[:, :1] + self.decoder_pos_embed[0, :1]
        # x = x[:, 1:] + self.sample_pe(img, self.decoder_pos_embed)
        # x = torch.cat((cls_token, x), dim=1)

        # apply Transformer blocks

        # if n_decoder_blocks == "all":
        #     for blk in self.decoder_blocks:
        #         x = blk(x)
        #     x = self.decoder_norm(x)
        # else:
        # for blk in self.decoder_blocks[:7]:
        #     x = blk(x)

        # # predictor projection
        # x = self.decoder_pred(x)

        # # remove cls token
        # x = x[:, 1:, :]
        #
        # return x

        return x[:, 1:, :].reshape(shape=(x.shape[0], H, W, -1)) \
                   .permute(0, 3, 1, 2), x[:, 0, :]

    def forward_encoder(self, img, mask_ratio):
        # embed patches
        x = self.patch_embed(img)

        # add pos embed w/o cls token
        x = x + self.sample_pe(img, self.pos_embed)


        # masking: length -> length * mask_ratio
        x, mask, ids_restore = self.random_masking(x, mask_ratio)

        # append cls token
        cls_token = self.cls_token + self.pos_embed[:, :1, :]
        cls_tokens = cls_token.expand(x.shape[0], -1, -1)
        x = torch.cat((cls_tokens, x), dim=1)

        # apply Transformer blocks
        for blk in self.blocks:
            x = blk(x)
        x = self.norm(x)

        return x, mask, ids_restore

    def forward_decoder(self, x, ids_restore, img):
        # embed tokens
        x = self.decoder_embed(x)

        # append mask tokens to sequence
        mask_tokens = self.mask_token.repeat(x.shape[0], ids_restore.shape[1] + 1 - x.shape[1], 1)
        x_ = torch.cat([x[:, 1:, :], mask_tokens], dim=1)  # no cls token
        x_ = torch.gather(x_, dim=1, index=ids_restore.unsqueeze(-1).repeat(1, 1, x.shape[2]))  # unshuffle
        x = torch.cat([x[:, :1, :], x_], dim=1)  # append cls token

        # # add pos embed
        # x = x + self.decoder_pos_embed

        # add pos embed
        cls_token = x[:, :1] + self.decoder_pos_embed[0, :1]
        x = x[:, 1:] + self.sample_pe(img, self.decoder_pos_embed)
        x = torch.cat((cls_token, x), dim=1)
        print("foo")

        # apply Transformer blocks
        for blk in self.decoder_blocks:
            x = blk(x)
        x = self.decoder_norm(x)

        # predictor projection
        x = self.decoder_pred(x)

        # remove cls token
        x = x[:, 1:, :]

        return x

    def forward_loss(self, imgs, pred, mask):
        """
        imgs: [N, 3, H, W]
        pred: [N, L, p*p*3]
        mask: [N, L], 0 is keep, 1 is remove,
        """
        target = self.patchify(imgs)
        if self.norm_pix_loss:
            mean = target.mean(dim=-1, keepdim=True)
            var = target.var(dim=-1, keepdim=True)
            target = (target - mean) / (var + 1.e-6) ** .5

        loss = (pred - target) ** 2
        loss = loss.mean(dim=-1)  # [N, L], mean loss per patch

        loss = (loss * mask).sum() / mask.sum()  # mean loss on removed patches
        return loss

    def forward(self, imgs, mask_ratio=0.75):
        latent, mask, ids_restore = self.forward_encoder(imgs, mask_ratio)
        pred = self.forward_decoder(latent, ids_restore, imgs)  # [N, L, p*p*3]
        loss = self.forward_loss(imgs, pred, mask)
        return loss, pred, mask


class MAEFeaturizer(nn.Module):

    def __init__(self, arch="mae_vit_large_patch16_gan"):
        super().__init__()
        # build model
        shared_args = dict(
            decoder_embed_dim=512,
            decoder_depth=8,
            decoder_num_heads=16,
            mlp_ratio=4,
            norm_layer=partial(nn.LayerNorm, eps=1e-6)
        )
        if arch == "mae_vit_base_patch16":
            self.model = MaskedAutoencoderViT(
                patch_size=16, embed_dim=768, depth=12, num_heads=12, **shared_args)
            chkpoint_dir = '../models/mae_visualize_vit_base.pth'
        elif arch == "mae_vit_large_patch16":
            self.model = MaskedAutoencoderViT(
                patch_size=16, embed_dim=1024, depth=24, num_heads=16, **shared_args)
            chkpoint_dir = '../models/mae_visualize_vit_large.pth'
        elif arch == "mae_vit_large_patch16_gan":
            self.model = MaskedAutoencoderViT(
                patch_size=16, embed_dim=1024, depth=24, num_heads=16, **shared_args)
            chkpoint_dir = '../models/mae_visualize_vit_large_ganloss.pth'
        elif arch == "mae_vit_huge_patch14":
            self.model = MaskedAutoencoderViT(
                patch_size=14, embed_dim=1280, depth=32, num_heads=16, **shared_args)
            chkpoint_dir = '../models/mae_visualize_vit_huge.pth'
        else:
            raise ValueError("Unknown model arch {}".format(arch))

        # load model
        chkpoint_dir = os.path.join(os.path.dirname(os.path.realpath(__file__)), chkpoint_dir)

        checkpoint = torch.load(chkpoint_dir)
        self.model.load_state_dict(checkpoint['model'], strict=False)

    def get_cls_token(self, img):
        feats, cls_token = self.model.featurize(img)
        return cls_token

    def forward(self, img):
        feats, cls_token = self.model.featurize(img)
        return feats


if __name__ == "__main__":
    import torchvision.transforms as T
    from PIL import Image
    from shared import norm, crop_to_divisor

    device = "cuda" if torch.cuda.is_available() else "cpu"

    image = Image.open("../samples/lex1.jpg")
    load_size = 224  # * 3
    transform = T.Compose([
        T.Resize(load_size, Image.BILINEAR),
        # T.CenterCrop(load_size),
        T.ToTensor(),
        lambda x: crop_to_divisor(x, 16),
        norm])

    model = MAEFeaturizer().cuda()

    results = model(transform(image).cuda().unsqueeze(0))

    print(results.shape)