vision_transformer.py

# Copyright (c) Facebook, Inc. and its affiliates.
# 
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# 
#     http://www.apache.org/licenses/LICENSE-2.0
# 
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Mostly copy-paste from timm library.
https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py

Modified to work with Hugging Face Transformers
"""
import math
from functools import partial
import torch
import torch.nn as nn
from transformers import PretrainedConfig, PreTrainedModel
from transformers.modeling_outputs import BaseModelOutput
from typing import Optional, Tuple, Union


def trunc_normal_(tensor, mean=0., std=1., a=-2., b=2.):
    """Truncated normal initialization (from timm library)"""
    def norm_cdf(x):
        return (1. + math.erf(x / math.sqrt(2.))) / 2.
    
    with torch.no_grad():
        l = norm_cdf((a - mean) / std)
        u = norm_cdf((b - mean) / std)
        tensor.uniform_(2 * l - 1, 2 * u - 1)
        tensor.erfinv_()
        tensor.mul_(std * math.sqrt(2.))
        tensor.add_(mean)
        tensor.clamp_(min=a, max=b)
        return tensor


def drop_path(x, drop_prob: float = 0., training: bool = False):
    if drop_prob == 0. or not training:
        return x
    keep_prob = 1 - drop_prob
    shape = (x.shape[0],) + (1,) * (x.ndim - 1)
    random_tensor = keep_prob + torch.rand(shape, dtype=x.dtype, device=x.device)
    random_tensor.floor_()
    output = x.div(keep_prob) * random_tensor
    return output


class DropPath(nn.Module):
    """Drop paths (Stochastic Depth) per sample"""
    def __init__(self, drop_prob=None):
        super(DropPath, self).__init__()
        self.drop_prob = drop_prob

    def forward(self, x):
        return drop_path(x, self.drop_prob, self.training)


class Mlp(nn.Module):
    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.fc2 = nn.Linear(hidden_features, out_features)
        self.drop = nn.Dropout(drop)

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


class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0.):
        super().__init__()
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or 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)

    def forward(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[0], qkv[1], qkv[2]
        attn = (q @ k.transpose(-2, -1)) * self.scale
        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, attn


class Block(nn.Module):
    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm):
        super().__init__()
        self.norm1 = norm_layer(dim)
        self.attn = Attention(
            dim, 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. 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)

    def forward(self, x, return_attention=False):
        y, attn = self.attn(self.norm1(x))
        if return_attention:
            return attn
        x = x + self.drop_path(y)
        x = x + self.drop_path(self.mlp(self.norm2(x)))
        return x


class PatchEmbed(nn.Module):
    """ Image to Patch Embedding """
    def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768):
        super().__init__()
        num_patches = (img_size // patch_size) * (img_size // patch_size)
        self.img_size = img_size
        self.patch_size = patch_size
        self.num_patches = num_patches
        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)

    def forward(self, x):
        B, C, H, W = x.shape
        x = self.proj(x).flatten(2).transpose(1, 2)
        return x


# ============================================================================
# HUGGING FACE CONFIGURATION CLASS (REQUIRED)
# ============================================================================

class VisionTransformerConfig(PretrainedConfig):
    """Configuration for Vision Transformer model"""
    
    model_type = "vit"
    
    def __init__(
        self,
        img_size=224,
        patch_size=16,
        in_chans=3,
        num_classes=0,
        embed_dim=768,
        depth=12,
        num_heads=12,
        mlp_ratio=4.0,
        qkv_bias=True,
        qk_scale=None,
        drop_rate=0.0,
        attn_drop_rate=0.0,
        drop_path_rate=0.0,
        **kwargs
    ):
        super().__init__(**kwargs)
        self.img_size = img_size
        self.patch_size = patch_size
        self.in_chans = in_chans
        self.num_classes = num_classes
        self.embed_dim = embed_dim
        self.depth = depth
        self.num_heads = num_heads
        self.mlp_ratio = mlp_ratio
        self.qkv_bias = qkv_bias
        self.qk_scale = qk_scale
        self.drop_rate = drop_rate
        self.attn_drop_rate = attn_drop_rate
        self.drop_path_rate = drop_path_rate


# ============================================================================
# HUGGING FACE COMPATIBLE WRAPPER (REQUIRED)
# ============================================================================

class VisionTransformer(PreTrainedModel):
    """
    Vision Transformer - Hugging Face compatible wrapper
    
    This wraps the original VisionTransformer to make it compatible with
    Hugging Face's AutoModel.from_pretrained()
    """
    
    config_class = VisionTransformerConfig
    base_model_prefix = "vit"
    main_input_name = "pixel_values"
    
    def __init__(self, config):
        super().__init__(config)
        self.config = config
        
        # Initialize the core Vision Transformer components
        self.num_features = self.embed_dim = config.embed_dim
        
        self.patch_embed = PatchEmbed(
            img_size=config.img_size,
            patch_size=config.patch_size,
            in_chans=config.in_chans,
            embed_dim=config.embed_dim
        )
        num_patches = self.patch_embed.num_patches
        
        self.cls_token = nn.Parameter(torch.zeros(1, 1, config.embed_dim))
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, config.embed_dim))
        self.pos_drop = nn.Dropout(p=config.drop_rate)
        
        # Stochastic depth decay rule
        dpr = [x.item() for x in torch.linspace(0, config.drop_path_rate, config.depth)]
        self.blocks = nn.ModuleList([
            Block(
                dim=config.embed_dim,
                num_heads=config.num_heads,
                mlp_ratio=config.mlp_ratio,
                qkv_bias=config.qkv_bias,
                qk_scale=config.qk_scale,
                drop=config.drop_rate,
                attn_drop=config.attn_drop_rate,
                drop_path=dpr[i],
                norm_layer=nn.LayerNorm
            )
            for i in range(config.depth)
        ])
        self.norm = nn.LayerNorm(config.embed_dim)
        
        # Classifier head
        self.head = nn.Linear(config.embed_dim, config.num_classes) if config.num_classes > 0 else nn.Identity()
        
        # Initialize weights
        trunc_normal_(self.pos_embed, std=.02)
        trunc_normal_(self.cls_token, std=.02)
        self.apply(self._init_weights)
    
    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)
    
    def interpolate_pos_encoding(self, x, w, h):
        npatch = x.shape[1] - 1
        N = self.pos_embed.shape[1] - 1
        if npatch == N and w == h:
            return self.pos_embed
        class_pos_embed = self.pos_embed[:, 0]
        patch_pos_embed = self.pos_embed[:, 1:]
        dim = x.shape[-1]
        w0 = w // self.patch_embed.patch_size
        h0 = h // self.patch_embed.patch_size
        w0, h0 = w0 + 0.1, h0 + 0.1
        patch_pos_embed = nn.functional.interpolate(
            patch_pos_embed.reshape(1, int(math.sqrt(N)), int(math.sqrt(N)), dim).permute(0, 3, 1, 2),
            scale_factor=(w0 / math.sqrt(N), h0 / math.sqrt(N)),
            mode='bicubic',
        )
        assert int(w0) == patch_pos_embed.shape[-2] and int(h0) == patch_pos_embed.shape[-1]
        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1)
    
    def prepare_tokens(self, x):
        B, nc, w, h = x.shape
        x = self.patch_embed(x)
        cls_tokens = self.cls_token.expand(B, -1, -1)
        x = torch.cat((cls_tokens, x), dim=1)
        x = x + self.interpolate_pos_encoding(x, w, h)
        return self.pos_drop(x)
    
    def forward(
        self,
        pixel_values: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, BaseModelOutput]:
        """
        Forward pass compatible with Hugging Face
        
        Args:
            pixel_values: Input images (batch_size, channels, height, width)
            output_attentions: Whether to return attention weights
            output_hidden_states: Whether to return all hidden states
            return_dict: Whether to return BaseModelOutput
        
        Returns:
            BaseModelOutput or tuple
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        
        x = self.prepare_tokens(pixel_values)
        
        for blk in self.blocks:
            x = blk(x)
        
        x = self.norm(x)
        
        # Return CLS token output
        pooled_output = x[:, 0]
        
        if not return_dict:
            return (x, pooled_output)
        
        return BaseModelOutput(
            last_hidden_state=x,
            hidden_states=None,
            attentions=None,
        )
    
    def forward_features(self, x):
        """
        Feature extraction method - EXACT match to local vision_transformer.py
        This ensures HuggingFace and local models give identical results
        """
        x = self.prepare_tokens(x)  # Tokenize input
        
        for blk in self.blocks:
            x = blk(x)
        
        x_norm = self.norm(x)  # Normalize tokens
        
        return {
            "x_norm_clstoken": x_norm[:, 0],  # CLS token
            "x_norm_patchtokens": x_norm[:, 1:],  # Patch tokens
            "x_prenorm": x,  # Before norm
        }
    
    def get_last_selfattention(self, x):
        """Get attention from last block"""
        x = self.prepare_tokens(x)
        for i, blk in enumerate(self.blocks):
            if i < len(self.blocks) - 1:
                x = blk(x)
            else:
                return blk(x, return_attention=True)
    
    def get_intermediate_layers(self, x, n=1):
        """Get outputs from last n blocks"""
        x = self.prepare_tokens(x)
        output = []
        for i, blk in enumerate(self.blocks):
            x = blk(x)
            if len(self.blocks) - i <= n:
                output.append(self.norm(x))
        return output


# Register for auto classes
VisionTransformerConfig.register_for_auto_class()
VisionTransformer.register_for_auto_class("AutoModel")