third_parts/recognize_anything/ram/models/ram_plus.py

'''
 * The Recognize Anything Plus Model (RAM++)
 * Written by Xinyu Huang
'''
import json
import warnings

import numpy as np
import torch
from torch import nn

import torch.nn.functional as F
from .bert import BertConfig, BertLMHeadModel, BertModel
from .swin_transformer import SwinTransformer
from .utils import *

warnings.filterwarnings("ignore")



class RAM_plus(nn.Module):
    def __init__(self,
                 med_config=f'{CONFIG_PATH}/configs/med_config.json',
                 image_size=384,
                 text_encoder_type='bert-base-uncased',
                 vit='base',
                 vit_grad_ckpt=False,
                 vit_ckpt_layer=0,
                 threshold=0.68,
                 delete_tag_index=[],
                 tag_list=f'{CONFIG_PATH}/data/ram_tag_list.txt',
                 tag_list_chinese=f'{CONFIG_PATH}/data/ram_tag_list_chinese.txt',
                 stage='eval'):
        r""" The Recognize Anything Plus Model (RAM++) inference module.
        RAM++ is a strong image tagging model, which can recognize any category with high accuracy using tag categories.
        Described in the paper "Open-Set Image Tagging with Multi-Grained Text Supervision" https://arxiv.org/abs/2310.15200

        Args:
            med_config (str): path for the mixture of encoder-decoder model's configuration file
            image_size (int): input image size
            vit (str): model size of vision transformer
            threshold (int): tagging threshold
            delete_tag_index (list): delete some tags that may disturb captioning
        """
        super().__init__()

        # create image encoder
        if vit == 'swin_b':
            if image_size == 224:
                vision_config_path = f'{CONFIG_PATH}/configs/swin/config_swinB_224.json'
            elif image_size == 384:
                vision_config_path = f'{CONFIG_PATH}/configs/swin/config_swinB_384.json'
            vision_config = read_json(vision_config_path)
            assert image_size == vision_config['image_res']
            # assert config['patch_size'] == 32
            vision_width = vision_config['vision_width']

            self.visual_encoder = SwinTransformer(
                img_size=vision_config['image_res'],
                patch_size=4,
                in_chans=3,
                embed_dim=vision_config['embed_dim'],
                depths=vision_config['depths'],
                num_heads=vision_config['num_heads'],
                window_size=vision_config['window_size'],
                mlp_ratio=4.,
                qkv_bias=True,
                drop_rate=0.0,
                drop_path_rate=0.1,
                ape=False,
                patch_norm=True,
                use_checkpoint=False)

            if stage == 'train_from_scratch':
                # download from https://github.com/microsoft/Swin-Transformer
                state_dict = torch.load(vision_config['ckpt'], map_location="cpu")['model']

                for k in list(state_dict.keys()):
                    if 'relative_position_bias_table' in k:
                        dst_num_pos = (2 * vision_config['window_size'] - 1) ** 2
                        state_dict[k] = interpolate_relative_pos_embed(state_dict[k], dst_num_pos, param_name=k)
                    elif ('relative_position_index' in k) or ('attn_mask' in k):
                        del state_dict[k]

                print("### Load Vision Backbone", vit)
                msg = self.visual_encoder.load_state_dict(state_dict, strict = False)
                print("missing_keys: ", msg.missing_keys)
                print("unexpected_keys: ", msg.unexpected_keys)

        elif vit == 'swin_l':
            if image_size == 224:
                vision_config_path = f'{CONFIG_PATH}/configs/swin/config_swinL_224.json'
            elif image_size == 384:
                vision_config_path = f'{CONFIG_PATH}/configs/swin/config_swinL_384.json'
            vision_config = read_json(vision_config_path)
            assert image_size == vision_config['image_res']
            # assert config['patch_size'] == 32
            vision_width = vision_config['vision_width']

            self.visual_encoder = SwinTransformer(
                img_size=vision_config['image_res'],
                patch_size=4,
                in_chans=3,
                embed_dim=vision_config['embed_dim'],
                depths=vision_config['depths'],
                num_heads=vision_config['num_heads'],
                window_size=vision_config['window_size'],
                mlp_ratio=4.,
                qkv_bias=True,
                drop_rate=0.0,
                drop_path_rate=0.1,
                ape=False,
                patch_norm=True,
                use_checkpoint=False)

            if stage == 'train_from_scratch':
                # download from https://github.com/microsoft/Swin-Transformer
                state_dict = torch.load(vision_config['ckpt'], map_location="cpu")['model']

                for k in list(state_dict.keys()):
                    if 'relative_position_bias_table' in k:
                        dst_num_pos = (2 * vision_config['window_size'] - 1) ** 2
                        state_dict[k] = interpolate_relative_pos_embed(state_dict[k], dst_num_pos, param_name=k)
                    elif ('relative_position_index' in k) or ('attn_mask' in k):
                        del state_dict[k]

                print("### Load Vision Backbone", vit)
                msg = self.visual_encoder.load_state_dict(state_dict, strict = False)
                print("missing_keys: ", msg.missing_keys)
                print("unexpected_keys: ", msg.unexpected_keys)

        else:
            self.visual_encoder, vision_width = create_vit(
                vit, image_size, vit_grad_ckpt, vit_ckpt_layer)

        # create tokenzier
        self.tokenizer = init_tokenizer(text_encoder_type)

        self.delete_tag_index = delete_tag_index

        # load tag list
        self.tag_list = self.load_tag_list(tag_list)
        self.tag_list_chinese = self.load_tag_list(tag_list_chinese)

        # create image-tag recognition decoder
        self.threshold = threshold
        self.num_class = len(self.tag_list)
        q2l_config = BertConfig.from_json_file(f'{CONFIG_PATH}/configs/q2l_config.json')
        q2l_config.encoder_width = 512
        self.tagging_head = BertModel(config=q2l_config,
                                      add_pooling_layer=False)
        self.tagging_head.resize_token_embeddings(len(self.tokenizer))

        if stage == 'train_from_scratch':
            self.label_embed = nn.Parameter(torch.load(f'{CONFIG_PATH}/data/frozen_tag_embedding/ram_plus_tag_embedding_class_4585_des_51.pth',map_location='cpu').float())
        else:
            # when eval with pretrained RAM++ model, directly load from ram_plus_swin_large_14m.pth
            self.label_embed = nn.Parameter(torch.zeros(self.num_class * 51, q2l_config.encoder_width))

        if q2l_config.hidden_size != 512:
            self.wordvec_proj = nn.Linear(512, q2l_config.hidden_size)
        else:
            self.wordvec_proj = nn.Identity()

        self.fc = nn.Linear(q2l_config.hidden_size, 1)

        self.del_selfattention()

        self.image_proj = nn.Linear(vision_width, 512)

        # adjust thresholds for some tags
        self.class_threshold = torch.ones(self.num_class) * self.threshold
        ram_class_threshold_path = f'{CONFIG_PATH}/data/ram_tag_list_threshold.txt'
        with open(ram_class_threshold_path, 'r', encoding='utf-8') as f:
            ram_class_threshold = [float(s.strip()) for s in f]
        for key,value in enumerate(ram_class_threshold):
            self.class_threshold[key] = value

        self.reweight_scale = nn.Parameter(torch.ones([]) * np.log(1 / 0.07))

        self.tagging_loss_function = AsymmetricLoss(gamma_neg=7,
                                                    gamma_pos=0,
                                                    clip=0.05)

        self.text_alignment_loss_function = AsymmetricLoss(gamma_neg=4,
                                                    gamma_pos=0,
                                                    clip=0.05)

    def load_tag_list(self, tag_list_file):
        with open(tag_list_file, 'r', encoding="utf-8") as f:
            tag_list = f.read().splitlines()
        tag_list = np.array(tag_list)
        return tag_list

    # delete self-attention layer of image-tag recognition decoder to reduce computation, follower Query2Label
    def del_selfattention(self):
        del self.tagging_head.embeddings
        for layer in self.tagging_head.encoder.layer:
            del layer.attention

    def forward(self, image, caption, image_tag, clip_feature, batch_text_embed):
        """
        call function as forward

        Args:
            image: type: torch.Tensor  shape: batch_size * 3 * 384 * 384
            caption: type: list[string]  len: batch_size
            tag: type: torch.Tensor   shape: batch * class_num (e.g. 3429)   value: positive sample is 1.0, negative sample is 0.0

        Returns:
            loss: type: torch.Tensor
        """

        image_embeds = self.image_proj(self.visual_encoder(image))
        image_atts = torch.ones(image_embeds.size()[:-1],
                                dtype=torch.long).to(image.device)

        ##================= Distillation from CLIP ================##
        image_cls_embeds = image_embeds[:, 0, :]
        image_spatial_embeds = image_embeds[:, 1:, :]

        loss_dis = F.l1_loss(image_cls_embeds, clip_feature)

        ###===========multi tag des reweight==============###
        bs = image_embeds.shape[0]

        des_per_class = int(self.label_embed.shape[0] / self.num_class)

        image_cls_embeds = image_cls_embeds / image_cls_embeds.norm(dim=-1, keepdim=True)
        reweight_scale = self.reweight_scale.exp()
        logits_per_image = (reweight_scale * image_cls_embeds @ self.label_embed.t())
        logits_per_image = logits_per_image.view(bs, -1,des_per_class)

        weight_normalized = F.softmax(logits_per_image, dim=2)
        label_embed_reweight = torch.empty(bs, self.num_class, 512).to(image.device).to(image.dtype)

        for i in range(bs):
            reshaped_value = self.label_embed.view(-1, des_per_class, 512)
            product = weight_normalized[i].unsqueeze(-1) * reshaped_value
            label_embed_reweight[i] = product.sum(dim=1)

        label_embed = torch.nn.functional.relu(self.wordvec_proj(label_embed_reweight))

        ##================= Image Tagging ================##

        tagging_embed = self.tagging_head(
            encoder_embeds=label_embed,
            encoder_hidden_states=image_embeds,
            encoder_attention_mask=image_atts,
            return_dict=False,
            mode='tagging',
        )

        logits = self.fc(tagging_embed[0]).squeeze(-1)

        loss_tag = self.tagging_loss_function(logits, image_tag)

        ##================= Image-text Alignment ================##

        batch_text_embed = torch.nn.functional.relu(self.wordvec_proj(batch_text_embed.to(self.label_embed.dtype)))
        batch_text_embed = batch_text_embed.unsqueeze(0).repeat(bs, 1, 1)
        alignment_embedding = self.tagging_head(
            encoder_embeds=batch_text_embed,
            encoder_hidden_states=image_embeds,
            encoder_attention_mask=image_atts,
            return_dict=False,
            mode='tagging',
        )
        alignment_logits = self.fc(alignment_embedding[0]).squeeze(-1)

        with torch.no_grad():
            alignment_targets = torch.zeros(alignment_logits.size()).to(image.device)
            alignment_targets.fill_diagonal_(1)

        loss_alignment = self.text_alignment_loss_function(alignment_logits,alignment_targets)

        return loss_tag, loss_dis, loss_alignment

    def generate_tag(self,
                 image
                 ):

        image_embeds = self.image_proj(self.visual_encoder(image))
        image_atts = torch.ones(image_embeds.size()[:-1],
                                dtype=torch.long).to(image.device)

        image_cls_embeds = image_embeds[:, 0, :]
        image_spatial_embeds = image_embeds[:, 1:, :]

        bs = image_spatial_embeds.shape[0]

        des_per_class = int(self.label_embed.shape[0] / self.num_class)

        image_cls_embeds = image_cls_embeds / image_cls_embeds.norm(dim=-1, keepdim=True)
        reweight_scale = self.reweight_scale.exp()
        logits_per_image = (reweight_scale * image_cls_embeds @ self.label_embed.t())
        logits_per_image = logits_per_image.view(bs, -1,des_per_class)

        weight_normalized = F.softmax(logits_per_image, dim=2)
        label_embed_reweight = torch.empty(bs, self.num_class, 512).to(image.device).to(image.dtype)

        for i in range(bs):
            # 这里对 value_ori 进行 reshape，然后使用 broadcasting
            reshaped_value = self.label_embed.view(-1, des_per_class, 512)
            product = weight_normalized[i].unsqueeze(-1) * reshaped_value
            label_embed_reweight[i] = product.sum(dim=1)

        label_embed = torch.nn.functional.relu(self.wordvec_proj(label_embed_reweight))

        # recognized image tags using alignment decoder
        tagging_embed = self.tagging_head(
            encoder_embeds=label_embed,
            encoder_hidden_states=image_embeds,
            encoder_attention_mask=image_atts,
            return_dict=False,
            mode='tagging',
        )

        logits = self.fc(tagging_embed[0]).squeeze(-1)

        targets = torch.where(
            torch.sigmoid(logits) > self.class_threshold.to(image.device),
            torch.tensor(1.0).to(image.device),
            torch.zeros(self.num_class).to(image.device))

        tag = targets.cpu().numpy()
        tag[:,self.delete_tag_index] = 0
        tag_output = []
        tag_output_chinese = []
        for b in range(bs):
            index = np.argwhere(tag[b] == 1)
            token = self.tag_list[index].squeeze(axis=1)
            tag_output.append(' | '.join(token))
            token_chinese = self.tag_list_chinese[index].squeeze(axis=1)
            tag_output_chinese.append(' | '.join(token_chinese))


        return tag_output, tag_output_chinese

    def generate_tag_openset(self,
                 image,
                 threshold=0.68,
                 tag_input=None,
                 ):

        image_embeds = self.image_proj(self.visual_encoder(image))
        image_atts = torch.ones(image_embeds.size()[:-1],
                                dtype=torch.long).to(image.device)

        image_cls_embeds = image_embeds[:, 0, :]
        image_spatial_embeds = image_embeds[:, 1:, :]

        bs = image_spatial_embeds.shape[0]

        des_per_class = int(self.label_embed.shape[0] / self.num_class)

        image_cls_embeds = image_cls_embeds / image_cls_embeds.norm(dim=-1, keepdim=True)
        reweight_scale = self.reweight_scale.exp()
        logits_per_image = (reweight_scale * image_cls_embeds @ self.label_embed.t())
        logits_per_image = logits_per_image.view(bs, -1,des_per_class)

        weight_normalized = F.softmax(logits_per_image, dim=2)
        label_embed_reweight = torch.empty(bs, self.num_class, 512).to(image.device).to(image.dtype)
                     
        for i in range(bs):
            # 这里对 value_ori 进行 reshape，然后使用 broadcasting
            reshaped_value = self.label_embed.view(-1, des_per_class, 512)
            product = weight_normalized[i].unsqueeze(-1) * reshaped_value
            label_embed_reweight[i] = product.sum(dim=1)

        label_embed = torch.nn.functional.relu(self.wordvec_proj(label_embed_reweight))

        # recognized image tags using alignment decoder
        tagging_embed = self.tagging_head(
            encoder_embeds=label_embed,
            encoder_hidden_states=image_embeds,
            encoder_attention_mask=image_atts,
            return_dict=False,
            mode='tagging',
        )

        logits = self.fc(tagging_embed[0]).squeeze(-1)

        targets = torch.where(
            torch.sigmoid(logits) > self.class_threshold.to(image.device),
            torch.tensor(1.0).to(image.device),
            torch.zeros(self.num_class).to(image.device))

        tag = targets.cpu().numpy()
        tag[:,self.delete_tag_index] = 0
        tag_output = []
        for b in range(bs):
            index = np.argwhere(tag[b] == 1)
            token = self.tag_list[index].squeeze(axis=1)
            tag_output.append(' | '.join(token))

        return tag_output


# load RAM++ pretrained model parameters
def ram_plus(pretrained='', **kwargs):
    model = RAM_plus(**kwargs)
    if pretrained:
        if kwargs['vit'] == 'swin_b':
            model, msg = load_checkpoint_swinbase(model, pretrained, kwargs)
        elif kwargs['vit'] == 'swin_l':
            model, msg = load_checkpoint_swinlarge(model, pretrained, kwargs)
        else:
            model, msg = load_checkpoint(model, pretrained)
        print('vit:', kwargs['vit'])
#         print('msg', msg)
    return model