intern_vit_6b.py

# --------------------------------------------------------
# InternVL
# Copyright (c) 2023 OpenGVLab
# Licensed under The MIT License [see LICENSE for details]
# --------------------------------------------------------
from functools import partial

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as checkpoint
from einops import rearrange
from timm.models.layers import DropPath, to_2tuple

try:
    from .flash_attention import FlashAttention
    has_flash_attn = True
except:
    print('FlashAttention is not installed.')
    has_flash_attn = False


def _freeze_params(module):
    for param in module.parameters():
        param.requires_grad = False


class CrossAttention(nn.Module):
    def __init__(
            self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0.,
            proj_drop=0., attn_head_dim=None, out_dim=None):
        super().__init__()
        if out_dim is None:
            out_dim = dim
        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 = qk_scale or head_dim ** -0.5
        assert all_head_dim == dim

        self.q = nn.Linear(dim, all_head_dim, bias=False)
        self.k = nn.Linear(dim, all_head_dim, bias=False)
        self.v = nn.Linear(dim, all_head_dim, bias=False)

        if qkv_bias:
            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
            self.k_bias = nn.Parameter(torch.zeros(all_head_dim))
            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
        else:
            self.q_bias = None
            self.k_bias = None
            self.v_bias = None

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

    def forward(self, x, k=None, v=None):
        B, N, C = x.shape
        N_k = k.shape[1]
        N_v = v.shape[1]

        q_bias, k_bias, v_bias = None, None, None
        if self.q_bias is not None:
            q_bias = self.q_bias
            k_bias = self.k_bias
            v_bias = self.v_bias

        q = F.linear(input=x, weight=self.q.weight, bias=q_bias)
        q = q.reshape(B, N, 1, self.num_heads, -1).permute(2, 0, 3, 1, 4).squeeze(0)  # (B, N_head, N_q, dim)

        k = F.linear(input=k, weight=self.k.weight, bias=k_bias)
        k = k.reshape(B, N_k, 1, self.num_heads, -1).permute(2, 0, 3, 1, 4).squeeze(0)

        v = F.linear(input=v, weight=self.v.weight, bias=v_bias)
        v = v.reshape(B, N_v, 1, self.num_heads, -1).permute(2, 0, 3, 1, 4).squeeze(0)

        q = q * self.scale
        attn = (q @ k.transpose(-2, -1))  # (B, N_head, N_q, N_k)

        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 AttentiveBlock(nn.Module):

    def __init__(self, dim, num_heads, qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., norm_layer=nn.LayerNorm, attn_head_dim=None, out_dim=None):
        super().__init__()

        self.norm1_q = norm_layer(dim)
        self.norm1_k = norm_layer(dim)
        self.norm1_v = norm_layer(dim)
        self.cross_attn = CrossAttention(
            dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop,
            proj_drop=drop, attn_head_dim=attn_head_dim, out_dim=out_dim)

        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()

    def forward(self, x_q, x_kv, pos_q, pos_k, bool_masked_pos, rel_pos_bias=None):
        x_q = self.norm1_q(x_q + pos_q)
        x_k = self.norm1_k(x_kv + pos_k)
        x_v = self.norm1_v(x_kv)
        x = self.cross_attn(x_q, k=x_k, v=x_v)

        return x


class AttentionPoolingBlock(AttentiveBlock):

    def forward(self, x):
        x_q = x.mean(1, keepdim=True)
        x_kv, pos_q, pos_k = x, 0, 0
        x = super().forward(x_q, x_kv, pos_q, pos_k, bool_masked_pos=None, rel_pos_bias=None)
        x = x.squeeze(1)
        return x


class RMSNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.variance_epsilon = eps

    def forward(self, hidden_states):
        input_dtype = hidden_states.dtype
        hidden_states = hidden_states.to(torch.float32)
        variance = hidden_states.pow(2).mean(-1, keepdim=True)
        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
        return self.weight * hidden_states.to(input_dtype)


try:
    from apex.normalization import FusedRMSNorm

    RMSNorm = FusedRMSNorm  # noqa

    print('Discovered apex.normalization.FusedRMSNorm - will use it instead of RMSNorm')
except ImportError:
    # using the normal RMSNorm
    pass
except Exception:
    print('discovered apex but it failed to load, falling back to RMSNorm')
    pass


class LayerScale(nn.Module):
    def __init__(self, dim, init_values=1e-5, inplace=False, force_fp32=False):
        super().__init__()
        self.inplace = inplace
        self.gamma = nn.Parameter(init_values * torch.ones(dim))
        self.force_fp32 = force_fp32

    @torch.cuda.amp.autocast(enabled=False)
    def forward(self, x):
        if self.force_fp32:
            output_type = x.dtype
            out = x.float().mul_(self.gamma.float()) if self.inplace else x.float() * self.gamma.float()
            return out.to(dtype=output_type)
        else:
            out = x.mul_(self.gamma) if self.inplace else x * self.gamma
            return out


class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, attn_drop=0., proj_drop=0., use_flash_attn=False,
                 causal=False, norm_layer=nn.LayerNorm, qk_normalization=False):
        super().__init__()
        assert dim % num_heads == 0, 'dim should be divisible by num_heads'
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = head_dim ** -0.5

        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)

        self.use_flash_attn = use_flash_attn
        if use_flash_attn:
            self.causal = causal
            self.inner_attn = FlashAttention(attention_dropout=attn_drop)

        self.qk_normalization = qk_normalization
        self.q_norm = norm_layer(dim) if qk_normalization else nn.Identity()
        self.k_norm = norm_layer(dim) if qk_normalization else nn.Identity()

    def _naive_attn(self, x):
        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)
        q, k, v = qkv.unbind(0)  # make torchscript happy (cannot use tensor as tuple)

        if self.qk_normalization:
            B_, H_, N_, D_ = q.shape
            q = self.q_norm(q.transpose(1, 2).flatten(-2, -1)).view(B_, N_, H_, D_).transpose(1, 2)
            k = self.k_norm(k.transpose(1, 2).flatten(-2, -1)).view(B_, N_, H_, D_).transpose(1, 2)

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

    def _flash_attn(self, x, key_padding_mask=None, need_weights=False):
        qkv = self.qkv(x)
        qkv = rearrange(qkv, 'b s (three h d) -> b s three h d', three=3, h=self.num_heads)

        if self.qk_normalization:
            q, k, v = qkv.unbind(2)
            q = self.q_norm(q.flatten(-2, -1)).view(q.shape)
            k = self.k_norm(k.flatten(-2, -1)).view(k.shape)
            qkv = torch.stack([q, k, v], dim=2)

        context, _ = self.inner_attn(
            qkv, key_padding_mask=key_padding_mask, need_weights=need_weights, causal=self.causal
        )
        outs = self.proj(rearrange(context, 'b s h d -> b s (h d)'))
        outs = self.proj_drop(outs)
        return outs

    def forward(self, x):
        x = self._naive_attn(x) if not self.use_flash_attn else self._flash_attn(x)
        return x


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,
                 bias=True, drop=0.):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        bias = to_2tuple(bias)
        drop_probs = to_2tuple(drop)

        self.fc1 = nn.Linear(in_features, hidden_features, bias=bias[0])
        self.act = act_layer()
        self.drop1 = nn.Dropout(drop_probs[0])
        self.fc2 = nn.Linear(hidden_features, out_features, bias=bias[1])
        self.drop2 = nn.Dropout(drop_probs[1])

    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


class Block(nn.Module):

    def __init__(
            self, dim, num_heads, mlp_ratio=4., qkv_bias=False, drop=0., attn_drop=0., init_values=None,
            drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm, use_flash_attn=False, with_cp=False,
            qk_normalization=False, layerscale_force_fp32=False):
        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,
                              use_flash_attn=use_flash_attn, causal=False, norm_layer=norm_layer,
                              qk_normalization=qk_normalization)
        self.ls1 = LayerScale(dim, init_values=init_values,
                              force_fp32=layerscale_force_fp32) if init_values else nn.Identity()
        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
        self.drop_path1 = 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)
        self.ls2 = LayerScale(dim, init_values=init_values,
                              force_fp32=layerscale_force_fp32) if init_values else nn.Identity()
        self.drop_path2 = DropPath(drop_path) if drop_path > 0. else nn.Identity()

        self.with_cp = with_cp

    def forward(self, x):

        def _inner_forward(x):
            x = x + self.drop_path1(self.ls1(self.attn(self.norm1(x))))
            x = x + self.drop_path2(self.ls2(self.mlp(self.norm2(x))))
            return x

        if self.with_cp:
            return checkpoint.checkpoint(_inner_forward, x)
        else:
            return _inner_forward(x)


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 = to_2tuple(img_size)
        patch_size = to_2tuple(patch_size)
        num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])
        self.patch_shape = (img_size[0] // patch_size[0], img_size[1] // patch_size[1])
        self.img_size = img_size
        self.patch_size = patch_size
        self.num_patches = num_patches
        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, **kwargs):
        x = self.proj(x)
        _, _, H, W = x.shape
        if self.flatten:
            x = x.flatten(2).transpose(1, 2)  # BCHW -> BNC
        x = self.norm(x)
        return x, H, W


class InternViT6B(nn.Module):

    def __init__(self, in_chans=3, patch_size=14, img_size=224, pretrain_size=224, qkv_bias=False, drop_path_rate=0.0,
                 embed_dim=3200, num_heads=25, mlp_ratio=4, init_values=0.1, qk_normalization=True, depth=48,
                 use_flash_attn=True, with_cp=True, layerscale_force_fp32=False, freeze_vit=True,
                 cls_target='cls_patch_concat', num_classes=1000, attn_pool_num_heads=16, clip_embed_dim=768,
                 head_norm_type='bn', pretrained=None):
        super().__init__()
        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models

        self.pretrain_size = pretrain_size
        self.drop_path_rate = drop_path_rate
        self.img_size = img_size
        self.patch_size = patch_size
        self.cls_target = cls_target
        self.depth = depth

        use_flash_attn = use_flash_attn and has_flash_attn
        if use_flash_attn and not has_flash_attn:
            print('Warning: Flash Attention is not available, use_flash_attn is set to False.')
        use_flash_attn = [use_flash_attn] * depth if not isinstance(use_flash_attn, list) else use_flash_attn

        norm_layer_for_blocks = partial(RMSNorm, eps=1e-6)
        self.norm_layer_for_blocks = norm_layer_for_blocks
        self.patch_embed = PatchEmbed(img_size, patch_size, in_chans, embed_dim)
        num_patches = self.patch_embed.num_patches
        self.num_patches = num_patches
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim))
        self.pos_drop = nn.Identity()
        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))

        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]

        self.blocks = nn.ModuleList([
            Block(embed_dim, num_heads, mlp_ratio, qkv_bias=qkv_bias,
                  norm_layer=norm_layer_for_blocks,
                  drop_path=dpr[i], init_values=init_values, attn_drop=0.,
                  use_flash_attn=use_flash_attn[i],
                  with_cp=with_cp,
                  qk_normalization=qk_normalization,
                  layerscale_force_fp32=layerscale_force_fp32)
            for i in range(depth)])

        if cls_target == 'clip_projector':
            self.clip_projector = AttentionPoolingBlock(
                dim=embed_dim, num_heads=attn_pool_num_heads, qkv_bias=True, qk_scale=None,
                drop=0., attn_drop=0., norm_layer=partial(nn.LayerNorm, eps=1e-5), out_dim=clip_embed_dim)

        self.init_weights(pretrained)

        if freeze_vit:
            _freeze_params(self)

        if cls_target == 'cls_patch_concat':
            if head_norm_type == 'bn':
                self.norm = nn.SyncBatchNorm(embed_dim * 2, eps=1e-6)
            else:
                self.norm = nn.LayerNorm(embed_dim * 2, eps=1e-6)
            self.head = nn.Linear(embed_dim * 2, num_classes) if num_classes > 0 else nn.Identity()
        elif cls_target == 'clip_projector':
            if head_norm_type == 'bn':
                self.norm = nn.SyncBatchNorm(clip_embed_dim, eps=1e-6)
            else:
                self.norm = nn.LayerNorm(clip_embed_dim, eps=1e-6)
            self.head = nn.Linear(clip_embed_dim, num_classes) if num_classes > 0 else nn.Identity()
        else:
            raise NotImplementedError

        if type(self.head) != nn.Identity:
            self.head.weight.data.normal_(mean=0.0, std=0.01)
            self.head.bias.data.zero_()

    def init_weights(self, pretrained=None):
        print(f'pretrained: {pretrained}')

        def resize_pos_embed(pos_embed, H, W):
            cls = pos_embed[:, :1, :]
            pos_embed = pos_embed[:, 1:, :].reshape(
                1, self.pretrain_size // 14, self.pretrain_size // 14, -1).permute(0, 3, 1, 2)
            pos_embed = F.interpolate(pos_embed, size=(H, W), mode='bicubic', align_corners=False). \
                reshape(1, -1, H * W).permute(0, 2, 1)
            pos_embed = torch.cat([cls, pos_embed], dim=1)
            return pos_embed

        if isinstance(pretrained, str):
            checkpoint = torch.load(pretrained, map_location='cpu')
            if 'module' in checkpoint:
                checkpoint = checkpoint['module']

            # resize pos_embed
            pos_embed = checkpoint['pos_embed']
            checkpoint['pos_embed'] = resize_pos_embed(
                pos_embed, self.img_size // self.patch_size, self.img_size // self.patch_size)
            # resize patch_embed
            patch_embed = checkpoint['patch_embed.proj.weight']
            checkpoint['patch_embed.proj.weight'] = F.interpolate(
                patch_embed, size=(self.patch_size, self.patch_size),
                mode='bicubic', align_corners=False)
            message = self.load_state_dict(checkpoint, strict=False)
            print(message)

    @property
    def dtype(self):
        return self.patch_embed.proj.weight.dtype

    def forward_features(self, x):
        x, _, _ = self.patch_embed(x.type(self.dtype))
        batch_size, seq_len, _ = x.size()
        cls_tokens = self.cls_token.expand(batch_size, -1, -1)
        x = torch.cat((cls_tokens, x), dim=1)
        x = x + self.pos_embed

        for idx, blk in enumerate(self.blocks):
            x = blk(x)
        return x

    def forward(self, x):
        x = self.forward_features(x)
        if self.cls_target == 'cls_patch_concat':
            x = torch.cat((x[:, 0, :], x[:, 1:, :].mean(dim=1)), dim=-1)
        elif self.cls_target == 'clip_projector':
            x = self.clip_projector(x)
        else:
            raise NotImplementedError
        x = self.norm(x)
        x = self.head(x)
        return x

    @torch.jit.ignore
    def lr_decay_keywords(self, decay_ratio=0.95):
        lr_ratios = {}

        # blocks
        for idx in range(self.depth):
            tag = 'blocks.{}.'.format(idx)
            decay = 1.0 * (decay_ratio ** (self.depth - idx))
            lr_ratios[tag] = decay

        # patch_embed
        lr_ratios['patch_embed'] = 1.0 * (decay_ratio ** (self.depth + 1))
        lr_ratios['pos_embed'] = 1.0 * (decay_ratio ** (self.depth + 1))
        lr_ratios['cls_token'] = 1.0 * (decay_ratio ** (self.depth + 1))

        return lr_ratios