super-gradients/src/super_gradients/training/models/classification_models/beit.py

""" BEIT: BERT Pre-Training of Image Transformers (https://arxiv.org/abs/2106.08254)

Model from official source: https://github.com/microsoft/unilm/tree/master/beit

At this point only the 1k fine-tuned classification weights and model configs have been added,
see original source above for pre-training models and procedure.

Modifications by / Copyright 2021 Ross Wightman, original copyrights below
"""
# --------------------------------------------------------
# BEIT: BERT Pre-Training of Image Transformers (https://arxiv.org/abs/2106.08254)
# Github source: https://github.com/microsoft/unilm/tree/master/beit
# Copyright (c) 2021 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# By Hangbo Bao
# Based on timm and DeiT code bases
# https://github.com/rwightman/pytorch-image-models/tree/master/timm
# https://github.com/facebookresearch/deit/
# https://github.com/facebookresearch/dino
# --------------------------------------------------------'
import math
from functools import partial
from typing import Optional, Tuple

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.utils.checkpoint import checkpoint
from torch import Tensor
from super_gradients.training.models.classification_models.vit import PatchEmbed
from super_gradients.training.utils.regularization_utils import DropPath
from super_gradients.common.abstractions.abstract_logger import get_logger
from super_gradients.training.utils import HpmStruct
from super_gradients.training.models import SgModule
logger = get_logger(__name__)


def resize_pos_embed(posemb, posemb_new, num_tokens=1, gs_new=()):
    # Rescale the grid of position embeddings when loading from state_dict. Adapted from
    # https://github.com/google-research/vision_transformer/blob/00883dd691c63a6830751563748663526e811cee/vit_jax/checkpoint.py#L224
    ntok_new = posemb_new.shape[1]
    if num_tokens:
        posemb_tok, posemb_grid = posemb[:, :num_tokens], posemb[0, num_tokens:]
        ntok_new -= num_tokens
    else:
        posemb_tok, posemb_grid = posemb[:, :0], posemb[0]
    gs_old = int(math.sqrt(len(posemb_grid)))
    if not len(gs_new):  # backwards compatibility
        gs_new = [int(math.sqrt(ntok_new))] * 2
    assert len(gs_new) >= 2
    posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2)
    posemb_grid = F.interpolate(posemb_grid, size=gs_new, mode='bicubic', align_corners=False)
    posemb_grid = posemb_grid.permute(0, 2, 3, 1).reshape(1, gs_new[0] * gs_new[1], -1)
    posemb = torch.cat([posemb_tok, posemb_grid], dim=1)
    return posemb


def _no_grad_trunc_normal_(tensor, mean, std, a, b):
    # Cut & paste from PyTorch official master until it's in a few official releases - RW
    # Method based on https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf
    # TODO: remove this function on next torch version
    def norm_cdf(x):
        # Computes standard normal cumulative distribution function
        return (1. + math.erf(x / math.sqrt(2.))) / 2.

    if (mean < a - 2 * std) or (mean > b + 2 * std):
        logger.warn("mean is more than 2 std from [a, b] in nn.init.trunc_normal_. "
                    "The distribution of values may be incorrect.",
                    stacklevel=2)

    with torch.no_grad():
        # Values are generated by using a truncated uniform distribution and
        # then using the inverse CDF for the normal distribution.
        # Get upper and lower cdf values
        lower = norm_cdf((a - mean) / std)
        upper = norm_cdf((b - mean) / std)

        # Uniformly fill tensor with values from [l, u], then translate to
        # [2l-1, 2u-1].
        tensor.uniform_(2 * lower - 1, 2 * upper - 1)

        # Use inverse cdf transform for normal distribution to get truncated
        # standard normal
        tensor.erfinv_()

        # Transform to proper mean, std
        tensor.mul_(std * math.sqrt(2.))
        tensor.add_(mean)

        # Clamp to ensure it's in the proper range
        tensor.clamp_(min=a, max=b)
        return tensor


def trunc_normal_(tensor, mean=0., std=1., a=-2., b=2.):
    # type: (Tensor, float, float, float, float) -> Tensor
    r"""Fills the input Tensor with values drawn from a truncated
    normal distribution. The values are effectively drawn from the
    normal distribution :math:`\mathcal{N}(\text{mean}, \text{std}^2)`
    with values outside :math:`[a, b]` redrawn until they are within
    the bounds. The method used for generating the random values works
    best when :math:`a \leq \text{mean} \leq b`.
    Args:
        tensor: an n-dimensional `torch.Tensor`
        mean: the mean of the normal distribution
        std: the standard deviation of the normal distribution
        a: the minimum cutoff value
        b: the maximum cutoff value
    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.trunc_normal_(w)
    """
    return _no_grad_trunc_normal_(tensor, mean, std, a, b)


class Mlp(nn.Module):
    """ MLP as used in Vision Transformer, MLP-Mixer and related networks
    """

    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
        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.act = act_layer()
        self.drop1 = nn.Dropout(drop)
        self.fc2 = nn.Linear(hidden_features, out_features)
        self.drop2 = nn.Dropout(drop)

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.drop1(x)
        x = self.fc2(x)
        x = self.drop2(x)
        return x


def gen_relative_position_index(window_size: Tuple[int, int]) -> torch.Tensor:
    num_relative_distance = (2 * window_size[0] - 1) * (2 * window_size[1] - 1) + 3
    # cls to token & token 2 cls & cls to cls
    # get pair-wise relative position index for each token inside the window
    window_area = window_size[0] * window_size[1]
    coords = torch.stack(torch.meshgrid(
        [torch.arange(window_size[0]),
         torch.arange(window_size[1])]))  # 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] += window_size[0] - 1  # shift to start from 0
    relative_coords[:, :, 1] += window_size[1] - 1
    relative_coords[:, :, 0] *= 2 * window_size[1] - 1
    relative_position_index = torch.zeros(size=(window_area + 1,) * 2, dtype=relative_coords.dtype)
    relative_position_index[1:, 1:] = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
    relative_position_index[0, 0:] = num_relative_distance - 3
    relative_position_index[0:, 0] = num_relative_distance - 2
    relative_position_index[0, 0] = num_relative_distance - 1
    return relative_position_index


class Attention(nn.Module):
    def __init__(
            self, dim, num_heads=8, qkv_bias=False, attn_drop=0.,
            proj_drop=0., window_size=None, attn_head_dim=None):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        if attn_head_dim is not None:
            head_dim = attn_head_dim
        all_head_dim = head_dim * self.num_heads
        self.scale = head_dim ** -0.5

        self.qkv = nn.Linear(dim, all_head_dim * 3, bias=False)
        if qkv_bias:
            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
            self.register_buffer('k_bias', torch.zeros(all_head_dim), persistent=False)
            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
        else:
            self.q_bias = None
            self.k_bias = None
            self.v_bias = None

        if window_size:
            self.window_size = window_size
            self.num_relative_distance = (2 * window_size[0] - 1) * (2 * window_size[1] - 1) + 3
            self.relative_position_bias_table = nn.Parameter(
                torch.zeros(self.num_relative_distance, num_heads))  # 2*Wh-1 * 2*Ww-1, nH
            self.register_buffer("relative_position_index", gen_relative_position_index(window_size))
        else:
            self.window_size = None
            self.relative_position_bias_table = None
            self.relative_position_index = None

        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(all_head_dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

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

    def forward(self, x, shared_rel_pos_bias: Optional[torch.Tensor] = None):
        B, N, C = x.shape

        qkv_bias = torch.cat((self.q_bias, self.k_bias, self.v_bias)) if self.q_bias is not None else None
        qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
        qkv = qkv.reshape(B, N, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
        q, k, v = qkv.unbind(0)  # make torchscript happy (cannot use tensor as tuple)

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

        if self.relative_position_bias_table is not None:
            attn = attn + self._get_rel_pos_bias()
        if shared_rel_pos_bias is not None:
            attn = attn + shared_rel_pos_bias

        attn = attn.softmax(dim=-1)
        attn = self.attn_drop(attn)

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


class Block(nn.Module):

    def __init__(
            self, dim, num_heads, mlp_ratio=4., qkv_bias=False, drop=0., attn_drop=0.,
            drop_path=0., init_values=None, act_layer=nn.GELU, norm_layer=nn.LayerNorm,
            window_size=None, attn_head_dim=None):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim, num_heads=num_heads, qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop,
            window_size=window_size, attn_head_dim=attn_head_dim)
        # 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. 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)

        if init_values:
            self.gamma_1 = nn.Parameter(init_values * torch.ones(dim), requires_grad=True)
            self.gamma_2 = nn.Parameter(init_values * torch.ones(dim), requires_grad=True)
        else:
            self.gamma_1, self.gamma_2 = None, None

    def forward(self, x, shared_rel_pos_bias: Optional[torch.Tensor] = None):
        if self.gamma_1 is None:
            x = x + self.drop_path(self.attn(self.norm1(x), shared_rel_pos_bias=shared_rel_pos_bias))
            x = x + self.drop_path(self.mlp(self.norm2(x)))
        else:
            x = x + self.drop_path(self.gamma_1 * self.attn(self.norm1(x), shared_rel_pos_bias=shared_rel_pos_bias))
            x = x + self.drop_path(self.gamma_2 * self.mlp(self.norm2(x)))
        return x


class RelativePositionBias(nn.Module):

    def __init__(self, window_size, num_heads):
        super().__init__()
        self.window_size = window_size
        self.window_area = window_size[0] * window_size[1]
        num_relative_distance = (2 * window_size[0] - 1) * (2 * window_size[1] - 1) + 3
        self.relative_position_bias_table = nn.Parameter(torch.zeros(num_relative_distance, num_heads))
        # trunc_normal_(self.relative_position_bias_table, std=.02)
        self.register_buffer("relative_position_index", gen_relative_position_index(window_size))

    def forward(self):
        relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
            self.window_area + 1, self.window_area + 1, -1)  # Wh*Ww,Wh*Ww,nH
        return relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww


class Beit(SgModule):
    """ Vision Transformer with support for patch or hybrid CNN input stage
    """

    def __init__(
            self, image_size=(224, 224), patch_size=16, in_chans=3, num_classes=1000, global_pool='avg',
            embed_dim=768, depth=12, num_heads=12, mlp_ratio=4., qkv_bias=True, drop_rate=0.,
            attn_drop_rate=0., drop_path_rate=0., norm_layer=partial(nn.LayerNorm, eps=1e-6),
            init_values=None, use_abs_pos_emb=True, use_rel_pos_bias=False, use_shared_rel_pos_bias=False,
            head_init_scale=0.001, **kwargs):
        super().__init__()
        self.num_classes = num_classes
        self.global_pool = global_pool
        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models
        self.grad_checkpointing = False

        self.patch_embed = PatchEmbed(
            img_size=image_size, patch_size=patch_size, in_channels=in_chans, hidden_dim=embed_dim)
        num_patches = self.patch_embed.num_patches

        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
        # self.mask_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim)) if use_abs_pos_emb else None
        self.pos_drop = nn.Dropout(p=drop_rate)

        if use_shared_rel_pos_bias:
            self.rel_pos_bias = RelativePositionBias(window_size=self.patch_embed.grid_size, num_heads=num_heads)
        else:
            self.rel_pos_bias = None

        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]  # stochastic depth decay rule
        self.blocks = nn.ModuleList([
            Block(
                dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias,
                drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer,
                init_values=init_values, window_size=self.patch_embed.grid_size if use_rel_pos_bias else None)
            for i in range(depth)])
        use_fc_norm = self.global_pool == 'avg'
        self.norm = nn.Identity() if use_fc_norm else norm_layer(embed_dim)
        self.fc_norm = norm_layer(embed_dim) if use_fc_norm else None
        self.head = nn.Linear(embed_dim, num_classes) if num_classes > 0 else nn.Identity()

        self.apply(self._init_weights)
        if self.pos_embed is not None:
            trunc_normal_(self.pos_embed, std=.02)
        trunc_normal_(self.cls_token, std=.02)
        # trunc_normal_(self.mask_token, std=.02)
        self.fix_init_weight()
        if isinstance(self.head, nn.Linear):
            trunc_normal_(self.head.weight, std=.02)
            self.head.weight.data.mul_(head_init_scale)
            self.head.bias.data.mul_(head_init_scale)

    def fix_init_weight(self):
        def rescale(param, layer_id):
            param.div_(math.sqrt(2.0 * layer_id))

        for layer_id, layer in enumerate(self.blocks):
            rescale(layer.attn.proj.weight.data, layer_id + 1)
            rescale(layer.mlp.fc2.weight.data, layer_id + 1)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=.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)

    @torch.jit.ignore
    def no_weight_decay(self):
        nwd = {'pos_embed', 'cls_token'}
        for n, _ in self.named_parameters():
            if 'relative_position_bias_table' in n:
                nwd.add(n)
        return nwd

    @torch.jit.ignore
    def set_grad_checkpointing(self, enable=True):
        self.grad_checkpointing = enable

    @torch.jit.ignore
    def group_matcher(self, coarse=False):
        matcher = dict(
            stem=r'^cls_token|pos_embed|patch_embed|rel_pos_bias',  # stem and embed
            blocks=[(r'^blocks\.(\d+)', None), (r'^norm', (99999,))],
        )
        return matcher

    @torch.jit.ignore
    def get_classifier(self):
        return self.head

    def reset_classifier(self, num_classes, global_pool=None):
        self.num_classes = num_classes
        if global_pool is not None:
            self.global_pool = global_pool
        self.head = nn.Linear(self.embed_dim, num_classes) if num_classes > 0 else nn.Identity()

    def forward_features(self, x):
        x = self.patch_embed(x)
        x = torch.cat((self.cls_token.expand(x.shape[0], -1, -1), x), dim=1)
        if self.pos_embed is not None:
            x = x + self.pos_embed
        x = self.pos_drop(x)

        rel_pos_bias = self.rel_pos_bias() if self.rel_pos_bias is not None else None
        for blk in self.blocks:
            if self.grad_checkpointing and not torch.jit.is_scripting():
                x = checkpoint(blk, x, shared_rel_pos_bias=rel_pos_bias)
            else:
                x = blk(x, shared_rel_pos_bias=rel_pos_bias)
        x = self.norm(x)
        return x

    def forward_head(self, x, pre_logits: bool = False):
        if self.fc_norm is not None:
            x = x[:, 1:].mean(dim=1)
            x = self.fc_norm(x)
        else:
            x = x[:, 0]
        return x if pre_logits else self.head(x)

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

    def replace_head(self, new_num_classes=None, new_head=None):
        if new_num_classes is None and new_head is None:
            raise ValueError("At least one of new_num_classes, new_head must be given to replace output layer.")
        if new_head is not None:
            self.head = new_head
        else:
            self.head = nn.Linear(self.head.in_features, new_num_classes)


def beit_base_patch16_224(arch_params: HpmStruct):
    model_kwargs = HpmStruct(patch_size=(16, 16),
                             embed_dim=768,
                             depth=12,
                             num_heads=12,
                             mlp_ratio=4,
                             use_abs_pos_emb=False,
                             use_rel_pos_bias=True,
                             init_values=0.1)
    model_kwargs.override(**arch_params.to_dict())
    model = Beit(**model_kwargs.to_dict())
    return model


def beit_large_patch16_224(arch_params: HpmStruct):
    model_kwargs = HpmStruct(patch_size=(16, 16),
                             embed_dim=1024,
                             depth=24,
                             num_heads=16,
                             mlp_ratio=4,
                             qkv_bias=True,
                             use_abs_pos_emb=False,
                             use_rel_pos_bias=True,
                             init_values=1e-5)

    model_kwargs.override(**arch_params.to_dict())
    model = Beit(**model_kwargs.to_dict())
    return model