Update magic_pdf/model/v3/__init__.py, magic_pdf/model/v3/helpers.py,...

Update magic_pdf/model/v3/__init__.py, magic_pdf/model/v3/helpers.py, magic_pdf/model/pek_sub_modules/layoutlmv3/layoutlmft/models/layoutlmv3/modeling_layoutlmv3.py files

magic_pdf/model/pek_sub_modules/layoutlmv3/layoutlmft/models/layoutlmv3/modeling_layoutlmv3.py

+# coding=utf-8
+# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
+# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
+#
+# 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.
+"""PyTorch LayoutLMv3 model. """
+import math
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from transformers import apply_chunking_to_forward
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPastAndCrossAttentions,
+    BaseModelOutputWithPoolingAndCrossAttentions,
+    MaskedLMOutput,
+    TokenClassifierOutput,
+    QuestionAnsweringModelOutput,
+    SequenceClassifierOutput,
+)
+from transformers.modeling_utils import PreTrainedModel, find_pruneable_heads_and_indices, prune_linear_layer
+from transformers.models.roberta.modeling_roberta import (
+    RobertaIntermediate,
+    RobertaLMHead,
+    RobertaOutput,
+    RobertaSelfOutput,
+)
+from transformers.utils import logging
+
+from .configuration_layoutlmv3 import LayoutLMv3Config
+from timm.models.layers import to_2tuple
+
+
+logger = logging.get_logger(__name__)
+
+
+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__()
+        img_size = to_2tuple(img_size)
+        patch_size = to_2tuple(patch_size)
+        self.patch_shape = (img_size[0] // patch_size[0], img_size[1] // patch_size[1])
+        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
+        # The following variables are used in detection mycheckpointer.py
+        self.num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])
+        self.num_patches_w = self.patch_shape[0]
+        self.num_patches_h = self.patch_shape[1]
+
+    def forward(self, x, position_embedding=None):
+        x = self.proj(x)
+
+        if position_embedding is not None:
+            # interpolate the position embedding to the corresponding size
+            position_embedding = position_embedding.view(1, self.patch_shape[0], self.patch_shape[1], -1).permute(0, 3, 1, 2)
+            Hp, Wp = x.shape[2], x.shape[3]
+            position_embedding = F.interpolate(position_embedding, size=(Hp, Wp), mode='bicubic')
+            x = x + position_embedding
+
+        x = x.flatten(2).transpose(1, 2)
+        return x
+
+class LayoutLMv3Embeddings(nn.Module):
+    """
+    Same as BertEmbeddings with a tiny tweak for positional embeddings indexing.
+    """
+
+    # Copied from transformers.models.bert.modeling_bert.BertEmbeddings.__init__
+    def __init__(self, config):
+        super().__init__()
+        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
+        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
+
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+        # position_ids (1, len position emb) is contiguous in memory and exported when serialized
+        self.register_buffer("position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)))
+
+        # End copy
+        self.padding_idx = config.pad_token_id
+        self.position_embeddings = nn.Embedding(
+            config.max_position_embeddings, config.hidden_size, padding_idx=self.padding_idx
+        )
+
+        self.x_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.coordinate_size)
+        self.y_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.coordinate_size)
+        self.h_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.shape_size)
+        self.w_position_embeddings = nn.Embedding(config.max_2d_position_embeddings, config.shape_size)
+
+    def _calc_spatial_position_embeddings(self, bbox):
+        try:
+            assert torch.all(0 <= bbox) and torch.all(bbox <= 1023)
+            left_position_embeddings = self.x_position_embeddings(bbox[:, :, 0])
+            upper_position_embeddings = self.y_position_embeddings(bbox[:, :, 1])
+            right_position_embeddings = self.x_position_embeddings(bbox[:, :, 2])
+            lower_position_embeddings = self.y_position_embeddings(bbox[:, :, 3])
+        except IndexError as e:
+            raise IndexError("The :obj:`bbox` coordinate values should be within 0-1000 range.") from e
+
+        h_position_embeddings = self.h_position_embeddings(torch.clip(bbox[:, :, 3] - bbox[:, :, 1], 0, 1023))
+        w_position_embeddings = self.w_position_embeddings(torch.clip(bbox[:, :, 2] - bbox[:, :, 0], 0, 1023))
+
+        # below is the difference between LayoutLMEmbeddingsV2 (torch.cat) and LayoutLMEmbeddingsV1 (add)
+        spatial_position_embeddings = torch.cat(
+            [
+                left_position_embeddings,
+                upper_position_embeddings,
+                right_position_embeddings,
+                lower_position_embeddings,
+                h_position_embeddings,
+                w_position_embeddings,
+            ],
+            dim=-1,
+        )
+        return spatial_position_embeddings
+
+    def create_position_ids_from_input_ids(self, input_ids, padding_idx, past_key_values_length=0):
+        """
+        Replace non-padding symbols with their position numbers. Position numbers begin at padding_idx+1. Padding symbols
+        are ignored. This is modified from fairseq's `utils.make_positions`.
+
+        Args:
+            x: torch.Tensor x:
+
+        Returns: torch.Tensor
+        """
+        # The series of casts and type-conversions here are carefully balanced to both work with ONNX export and XLA.
+        mask = input_ids.ne(padding_idx).int()
+        incremental_indices = (torch.cumsum(mask, dim=1).type_as(mask) + past_key_values_length) * mask
+        return incremental_indices.long() + padding_idx
+
+    def forward(
+        self,
+        input_ids=None,
+        bbox=None,
+        token_type_ids=None,
+        position_ids=None,
+        inputs_embeds=None,
+        past_key_values_length=0,
+    ):
+        if position_ids is None:
+            if input_ids is not None:
+                # Create the position ids from the input token ids. Any padded tokens remain padded.
+                position_ids = self.create_position_ids_from_input_ids(
+                    input_ids, self.padding_idx, past_key_values_length).to(input_ids.device)
+            else:
+                position_ids = self.create_position_ids_from_inputs_embeds(inputs_embeds)
+
+        if input_ids is not None:
+            input_shape = input_ids.size()
+        else:
+            input_shape = inputs_embeds.size()[:-1]
+
+        if token_type_ids is None:
+            token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+        token_type_embeddings = self.token_type_embeddings(token_type_ids)
+
+        embeddings = inputs_embeds + token_type_embeddings
+        position_embeddings = self.position_embeddings(position_ids)
+        embeddings += position_embeddings
+
+        spatial_position_embeddings = self._calc_spatial_position_embeddings(bbox)
+
+        embeddings = embeddings + spatial_position_embeddings
+
+        embeddings = self.LayerNorm(embeddings)
+        embeddings = self.dropout(embeddings)
+        return embeddings
+
+    def create_position_ids_from_inputs_embeds(self, inputs_embeds):
+        """
+        We are provided embeddings directly. We cannot infer which are padded so just generate sequential position ids.
+
+        Args:
+            inputs_embeds: torch.Tensor≈
+
+        Returns: torch.Tensor
+        """
+        input_shape = inputs_embeds.size()[:-1]
+        sequence_length = input_shape[1]
+
+        position_ids = torch.arange(
+            self.padding_idx + 1, sequence_length + self.padding_idx + 1, dtype=torch.long, device=inputs_embeds.device
+        )
+        return position_ids.unsqueeze(0).expand(input_shape)
+
+
+class LayoutLMv3PreTrainedModel(PreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = LayoutLMv3Config
+    base_model_prefix = "layoutlmv3"
+
+    # Copied from transformers.models.bert.modeling_bert.BertPreTrainedModel._init_weights
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        if isinstance(module, nn.Linear):
+            # Slightly different from the TF version which uses truncated_normal for initialization
+            # cf https://github.com/pytorch/pytorch/pull/5617
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+
+class LayoutLMv3SelfAttention(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
+            raise ValueError(
+                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
+                f"heads ({config.num_attention_heads})"
+            )
+
+        self.num_attention_heads = config.num_attention_heads
+        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
+        self.all_head_size = self.num_attention_heads * self.attention_head_size
+
+        self.query = nn.Linear(config.hidden_size, self.all_head_size)
+        self.key = nn.Linear(config.hidden_size, self.all_head_size)
+        self.value = nn.Linear(config.hidden_size, self.all_head_size)
+
+        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
+        self.has_relative_attention_bias = config.has_relative_attention_bias
+        self.has_spatial_attention_bias = config.has_spatial_attention_bias
+
+    def transpose_for_scores(self, x):
+        new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
+        x = x.view(*new_x_shape)
+        return x.permute(0, 2, 1, 3)
+
+    def cogview_attn(self, attention_scores, alpha=32):
+        '''
+        https://arxiv.org/pdf/2105.13290.pdf
+        Section 2.4 Stabilization of training: Precision Bottleneck Relaxation (PB-Relax).
+        A replacement of the original nn.Softmax(dim=-1)(attention_scores)
+        Seems the new attention_probs will result in a slower speed and a little bias
+        Can use torch.allclose(standard_attention_probs, cogview_attention_probs, atol=1e-08) for comparison
+        The smaller atol (e.g., 1e-08), the better.
+        '''
+        scaled_attention_scores = attention_scores / alpha
+        max_value = scaled_attention_scores.amax(dim=(-1)).unsqueeze(-1)
+        # max_value = scaled_attention_scores.amax(dim=(-2, -1)).unsqueeze(-1).unsqueeze(-1)
+        new_attention_scores = (scaled_attention_scores - max_value) * alpha
+        return nn.Softmax(dim=-1)(new_attention_scores)
+
+    def forward(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_value=None,
+        output_attentions=False,
+        rel_pos=None,
+        rel_2d_pos=None,
+    ):
+        mixed_query_layer = self.query(hidden_states)
+
+        # If this is instantiated as a cross-attention module, the keys
+        # and values come from an encoder; the attention mask needs to be
+        # such that the encoder's padding tokens are not attended to.
+        is_cross_attention = encoder_hidden_states is not None
+
+        if is_cross_attention and past_key_value is not None:
+            # reuse k,v, cross_attentions
+            key_layer = past_key_value[0]
+            value_layer = past_key_value[1]
+            attention_mask = encoder_attention_mask
+        elif is_cross_attention:
+            key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
+            value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
+            attention_mask = encoder_attention_mask
+        elif past_key_value is not None:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+            key_layer = torch.cat([past_key_value[0], key_layer], dim=2)
+            value_layer = torch.cat([past_key_value[1], value_layer], dim=2)
+        else:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+
+        query_layer = self.transpose_for_scores(mixed_query_layer)
+
+        # Take the dot product between "query" and "key" to get the raw attention scores.
+        # The attention scores QT K/√d could be significantly larger than input elements, and result in overflow.
+        # Changing the computational order into QT(K/√d) alleviates the problem. (https://arxiv.org/pdf/2105.13290.pdf)
+        attention_scores = torch.matmul(query_layer / math.sqrt(self.attention_head_size), key_layer.transpose(-1, -2))
+
+        if self.has_relative_attention_bias and self.has_spatial_attention_bias:
+            attention_scores += (rel_pos + rel_2d_pos) / math.sqrt(self.attention_head_size)
+        elif self.has_relative_attention_bias:
+            attention_scores += rel_pos / math.sqrt(self.attention_head_size)
+
+        # if self.has_relative_attention_bias:
+        #     attention_scores += rel_pos
+        # if self.has_spatial_attention_bias:
+        #     attention_scores += rel_2d_pos
+
+        # attention_scores = attention_scores / math.sqrt(self.attention_head_size)
+        if attention_mask is not None:
+            # Apply the attention mask is (precomputed for all layers in RobertaModel forward() function)
+            attention_scores = attention_scores + attention_mask
+
+        # Normalize the attention scores to probabilities.
+        # attention_probs = nn.Softmax(dim=-1)(attention_scores)  # comment the line below and use this line for speedup
+        attention_probs = self.cogview_attn(attention_scores)  # to stablize training
+        # assert torch.allclose(attention_probs, nn.Softmax(dim=-1)(attention_scores), atol=1e-8)
+
+        # This is actually dropping out entire tokens to attend to, which might
+        # seem a bit unusual, but is taken from the original Transformer paper.
+        attention_probs = self.dropout(attention_probs)
+
+        # Mask heads if we want to
+        if head_mask is not None:
+            attention_probs = attention_probs * head_mask
+
+        context_layer = torch.matmul(attention_probs, value_layer)
+
+        context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
+        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
+        context_layer = context_layer.view(*new_context_layer_shape)
+
+        outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)
+
+        return outputs
+
+
+class LayoutLMv3Attention(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.self = LayoutLMv3SelfAttention(config)
+        self.output = RobertaSelfOutput(config)
+        self.pruned_heads = set()
+
+    def prune_heads(self, heads):
+        if len(heads) == 0:
+            return
+        heads, index = find_pruneable_heads_and_indices(
+            heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
+        )
+
+        # Prune linear layers
+        self.self.query = prune_linear_layer(self.self.query, index)
+        self.self.key = prune_linear_layer(self.self.key, index)
+        self.self.value = prune_linear_layer(self.self.value, index)
+        self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)
+
+        # Update hyper params and store pruned heads
+        self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
+        self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
+        self.pruned_heads = self.pruned_heads.union(heads)
+
+    def forward(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_value=None,
+        output_attentions=False,
+        rel_pos=None,
+        rel_2d_pos=None,
+    ):
+        self_outputs = self.self(
+            hidden_states,
+            attention_mask,
+            head_mask,
+            encoder_hidden_states,
+            encoder_attention_mask,
+            past_key_value,
+            output_attentions,
+            rel_pos=rel_pos,
+            rel_2d_pos=rel_2d_pos,
+        )
+        attention_output = self.output(self_outputs[0], hidden_states)
+        outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
+        return outputs
+
+
+class LayoutLMv3Layer(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.chunk_size_feed_forward = config.chunk_size_feed_forward
+        self.seq_len_dim = 1
+        self.attention = LayoutLMv3Attention(config)
+        assert not config.is_decoder and not config.add_cross_attention, \
+            "This version do not support decoder. Please refer to RoBERTa for implementation of is_decoder."
+        self.intermediate = RobertaIntermediate(config)
+        self.output = RobertaOutput(config)
+
+    def forward(
+        self,
+        hidden_states,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_value=None,
+        output_attentions=False,
+        rel_pos=None,
+        rel_2d_pos=None,
+    ):
+        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
+        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
+        self_attention_outputs = self.attention(
+            hidden_states,
+            attention_mask,
+            head_mask,
+            output_attentions=output_attentions,
+            past_key_value=self_attn_past_key_value,
+            rel_pos=rel_pos,
+            rel_2d_pos=rel_2d_pos,
+        )
+        attention_output = self_attention_outputs[0]
+
+        outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights
+
+        layer_output = apply_chunking_to_forward(
+            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
+        )
+        outputs = (layer_output,) + outputs
+
+        return outputs
+
+    def feed_forward_chunk(self, attention_output):
+        intermediate_output = self.intermediate(attention_output)
+        layer_output = self.output(intermediate_output, attention_output)
+        return layer_output
+
+
+class LayoutLMv3Encoder(nn.Module):
+    def __init__(self, config, detection=False, out_features=None):
+        super().__init__()
+        self.config = config
+        self.detection = detection
+        self.layer = nn.ModuleList([LayoutLMv3Layer(config) for _ in range(config.num_hidden_layers)])
+        self.gradient_checkpointing = False
+
+        self.has_relative_attention_bias = config.has_relative_attention_bias
+        self.has_spatial_attention_bias = config.has_spatial_attention_bias
+
+        if self.has_relative_attention_bias:
+            self.rel_pos_bins = config.rel_pos_bins
+            self.max_rel_pos = config.max_rel_pos
+            self.rel_pos_onehot_size = config.rel_pos_bins
+            self.rel_pos_bias = nn.Linear(self.rel_pos_onehot_size, config.num_attention_heads, bias=False)
+
+        if self.has_spatial_attention_bias:
+            self.max_rel_2d_pos = config.max_rel_2d_pos
+            self.rel_2d_pos_bins = config.rel_2d_pos_bins
+            self.rel_2d_pos_onehot_size = config.rel_2d_pos_bins
+            self.rel_pos_x_bias = nn.Linear(self.rel_2d_pos_onehot_size, config.num_attention_heads, bias=False)
+            self.rel_pos_y_bias = nn.Linear(self.rel_2d_pos_onehot_size, config.num_attention_heads, bias=False)
+
+        if self.detection:
+            self.gradient_checkpointing = True
+            embed_dim = self.config.hidden_size
+            self.out_features = out_features
+            self.out_indices = [int(name[5:]) for name in out_features]
+            self.fpn1 = nn.Sequential(
+                nn.ConvTranspose2d(embed_dim, embed_dim, kernel_size=2, stride=2),
+                # nn.SyncBatchNorm(embed_dim),
+                nn.BatchNorm2d(embed_dim),
+                nn.GELU(),
+                nn.ConvTranspose2d(embed_dim, embed_dim, kernel_size=2, stride=2),
+            )
+
+            self.fpn2 = nn.Sequential(
+                nn.ConvTranspose2d(embed_dim, embed_dim, kernel_size=2, stride=2),
+            )
+
+            self.fpn3 = nn.Identity()
+
+            self.fpn4 = nn.MaxPool2d(kernel_size=2, stride=2)
+            self.ops = [self.fpn1, self.fpn2, self.fpn3, self.fpn4]
+
+    def relative_position_bucket(self, relative_position, bidirectional=True, num_buckets=32, max_distance=128):
+        ret = 0
+        if bidirectional:
+            num_buckets //= 2
+            ret += (relative_position > 0).long() * num_buckets
+            n = torch.abs(relative_position)
+        else:
+            n = torch.max(-relative_position, torch.zeros_like(relative_position))
+        # now n is in the range [0, inf)
+
+        # half of the buckets are for exact increments in positions
+        max_exact = num_buckets // 2
+        is_small = n < max_exact
+
+        # The other half of the buckets are for logarithmically bigger bins in positions up to max_distance
+        val_if_large = max_exact + (
+                torch.log(n.float() / max_exact) / math.log(max_distance / max_exact) * (num_buckets - max_exact)
+        ).to(torch.long)
+        val_if_large = torch.min(val_if_large, torch.full_like(val_if_large, num_buckets - 1))
+
+        ret += torch.where(is_small, n, val_if_large)
+        return ret
+
+    def _cal_1d_pos_emb(self, hidden_states, position_ids, valid_span):
+        VISUAL_NUM = 196 + 1
+
+        rel_pos_mat = position_ids.unsqueeze(-2) - position_ids.unsqueeze(-1)
+
+        if valid_span is not None:
+            # for the text part, if two words are not in the same line,
+            # set their distance to the max value (position_ids.shape[-1])
+            rel_pos_mat[(rel_pos_mat > 0) & (valid_span == False)] = position_ids.shape[1]
+            rel_pos_mat[(rel_pos_mat < 0) & (valid_span == False)] = -position_ids.shape[1]
+
+            # image-text, minimum distance
+            rel_pos_mat[:, -VISUAL_NUM:, :-VISUAL_NUM] = 0
+            rel_pos_mat[:, :-VISUAL_NUM, -VISUAL_NUM:] = 0
+
+        rel_pos = self.relative_position_bucket(
+            rel_pos_mat,
+            num_buckets=self.rel_pos_bins,
+            max_distance=self.max_rel_pos,
+        )
+        rel_pos = F.one_hot(rel_pos, num_classes=self.rel_pos_onehot_size).type_as(hidden_states)
+        rel_pos = self.rel_pos_bias(rel_pos).permute(0, 3, 1, 2)
+        rel_pos = rel_pos.contiguous()
+        return rel_pos
+
+    def _cal_2d_pos_emb(self, hidden_states, bbox):
+        position_coord_x = bbox[:, :, 0]
+        position_coord_y = bbox[:, :, 3]
+        rel_pos_x_2d_mat = position_coord_x.unsqueeze(-2) - position_coord_x.unsqueeze(-1)
+        rel_pos_y_2d_mat = position_coord_y.unsqueeze(-2) - position_coord_y.unsqueeze(-1)
+        rel_pos_x = self.relative_position_bucket(
+            rel_pos_x_2d_mat,
+            num_buckets=self.rel_2d_pos_bins,
+            max_distance=self.max_rel_2d_pos,
+        )
+        rel_pos_y = self.relative_position_bucket(
+            rel_pos_y_2d_mat,
+            num_buckets=self.rel_2d_pos_bins,
+            max_distance=self.max_rel_2d_pos,
+        )
+        rel_pos_x = F.one_hot(rel_pos_x, num_classes=self.rel_2d_pos_onehot_size).type_as(hidden_states)
+        rel_pos_y = F.one_hot(rel_pos_y, num_classes=self.rel_2d_pos_onehot_size).type_as(hidden_states)
+        rel_pos_x = self.rel_pos_x_bias(rel_pos_x).permute(0, 3, 1, 2)
+        rel_pos_y = self.rel_pos_y_bias(rel_pos_y).permute(0, 3, 1, 2)
+        rel_pos_x = rel_pos_x.contiguous()
+        rel_pos_y = rel_pos_y.contiguous()
+        rel_2d_pos = rel_pos_x + rel_pos_y
+        return rel_2d_pos
+
+    def forward(
+        self,
+        hidden_states,
+        bbox=None,
+        attention_mask=None,
+        head_mask=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_values=None,
+        use_cache=None,
+        output_attentions=False,
+        output_hidden_states=False,
+        return_dict=True,
+        position_ids=None,
+        Hp=None,
+        Wp=None,
+        valid_span=None,
+    ):
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attentions = () if output_attentions else None
+        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
+
+        next_decoder_cache = () if use_cache else None
+
+        rel_pos = self._cal_1d_pos_emb(hidden_states, position_ids, valid_span) if self.has_relative_attention_bias else None
+        rel_2d_pos = self._cal_2d_pos_emb(hidden_states, bbox) if self.has_spatial_attention_bias else None
+
+        if self.detection:
+            feat_out = {}
+            j = 0
+
+        for i, layer_module in enumerate(self.layer):
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            layer_head_mask = head_mask[i] if head_mask is not None else None
+            past_key_value = past_key_values[i] if past_key_values is not None else None
+
+            if self.gradient_checkpointing and self.training:
+
+                if use_cache:
+                    logger.warning(
+                        "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                    )
+                    use_cache = False
+
+                def create_custom_forward(module):
+                    def custom_forward(*inputs):
+                        return module(*inputs)
+                        # return module(*inputs, past_key_value, output_attentions, rel_pos, rel_2d_pos)
+                        # The above line will cause error:
+                        # RuntimeError: Trying to backward through the graph a second time
+                        # (or directly access saved tensors after they have already been freed).
+                    return custom_forward
+
+                layer_outputs = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(layer_module),
+                    hidden_states,
+                    attention_mask,
+                    layer_head_mask,
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                    past_key_value,
+                    output_attentions,
+                    rel_pos,
+                    rel_2d_pos
+                )
+            else:
+                layer_outputs = layer_module(
+                    hidden_states,
+                    attention_mask,
+                    layer_head_mask,
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                    past_key_value,
+                    output_attentions,
+                    rel_pos=rel_pos,
+                    rel_2d_pos=rel_2d_pos,
+                )
+
+            hidden_states = layer_outputs[0]
+            if use_cache:
+                next_decoder_cache += (layer_outputs[-1],)
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (layer_outputs[1],)
+                if self.config.add_cross_attention:
+                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)
+
+            if self.detection and i in self.out_indices:
+                xp = hidden_states[:, -Hp*Wp:, :].permute(0, 2, 1).reshape(len(hidden_states), -1, Hp, Wp)
+                feat_out[self.out_features[j]] = self.ops[j](xp.contiguous())
+                j += 1
+
+        if self.detection:
+            return feat_out
+
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [
+                    hidden_states,
+                    next_decoder_cache,
+                    all_hidden_states,
+                    all_self_attentions,
+                    all_cross_attentions,
+                ]
+                if v is not None
+            )
+        return BaseModelOutputWithPastAndCrossAttentions(
+            last_hidden_state=hidden_states,
+            past_key_values=next_decoder_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+            cross_attentions=all_cross_attentions,
+        )
+
+
+class LayoutLMv3Model(LayoutLMv3PreTrainedModel):
+    """
+    """
+
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    # Copied from transformers.models.bert.modeling_bert.BertModel.__init__ with Bert->Roberta
+    def __init__(self, config, detection=False, out_features=None, image_only=False):
+        super().__init__(config)
+        self.config = config
+        assert not config.is_decoder and not config.add_cross_attention, \
+            "This version do not support decoder. Please refer to RoBERTa for implementation of is_decoder."
+        self.detection = detection
+        if not self.detection:
+            self.image_only = False
+        else:
+            assert config.visual_embed
+            self.image_only = image_only
+
+        if not self.image_only:
+            self.embeddings = LayoutLMv3Embeddings(config)
+        self.encoder = LayoutLMv3Encoder(config, detection=detection, out_features=out_features)
+
+        if config.visual_embed:
+            embed_dim = self.config.hidden_size
+            # use the default pre-training parameters for fine-tuning (e.g., input_size)
+            # when the input_size is larger in fine-tuning, we will interpolate the position embedding in forward
+            self.patch_embed = PatchEmbed(embed_dim=embed_dim)
+
+            patch_size = 16
+            size = int(self.config.input_size / patch_size)
+            self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
+            self.pos_embed = nn.Parameter(torch.zeros(1, size * size + 1, embed_dim))
+            self.pos_drop = nn.Dropout(p=0.)
+
+            self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+            self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+            if self.config.has_relative_attention_bias or self.config.has_spatial_attention_bias:
+                self._init_visual_bbox(img_size=(size, size))
+
+            from functools import partial
+            norm_layer = partial(nn.LayerNorm, eps=1e-6)
+            self.norm = norm_layer(embed_dim)
+
+        self.init_weights()
+
+    def get_input_embeddings(self):
+        return self.embeddings.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.embeddings.word_embeddings = value
+
+    def _prune_heads(self, heads_to_prune):
+        """
+        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
+        class PreTrainedModel
+        """
+        for layer, heads in heads_to_prune.items():
+            self.encoder.layer[layer].attention.prune_heads(heads)
+
+    def _init_visual_bbox(self, img_size=(14, 14), max_len=1000):
+        visual_bbox_x = torch.div(torch.arange(0, max_len * (img_size[1] + 1), max_len),
+                                  img_size[1], rounding_mode='trunc')
+        visual_bbox_y = torch.div(torch.arange(0, max_len * (img_size[0] + 1), max_len),
+                                  img_size[0], rounding_mode='trunc')
+        visual_bbox = torch.stack(
+            [
+                visual_bbox_x[:-1].repeat(img_size[0], 1),
+                visual_bbox_y[:-1].repeat(img_size[1], 1).transpose(0, 1),
+                visual_bbox_x[1:].repeat(img_size[0], 1),
+                visual_bbox_y[1:].repeat(img_size[1], 1).transpose(0, 1),
+            ],
+            dim=-1,
+        ).view(-1, 4)
+
+        cls_token_box = torch.tensor([[0 + 1, 0 + 1, max_len - 1, max_len - 1]])
+        self.visual_bbox = torch.cat([cls_token_box, visual_bbox], dim=0)
+
+    def _calc_visual_bbox(self, device, dtype, bsz):  # , img_size=(14, 14), max_len=1000):
+        visual_bbox = self.visual_bbox.repeat(bsz, 1, 1)
+        visual_bbox = visual_bbox.to(device).type(dtype)
+        return visual_bbox
+
+    def forward_image(self, x):
+        if self.detection:
+            x = self.patch_embed(x, self.pos_embed[:, 1:, :] if self.pos_embed is not None else None)
+        else:
+            x = self.patch_embed(x)
+        batch_size, seq_len, _ = x.size()
+
+        cls_tokens = self.cls_token.expand(batch_size, -1, -1)  # stole cls_tokens impl from Phil Wang, thanks
+        if self.pos_embed is not None and self.detection:
+            cls_tokens = cls_tokens + self.pos_embed[:, :1, :]
+
+        x = torch.cat((cls_tokens, x), dim=1)
+        if self.pos_embed is not None and not self.detection:
+            x = x + self.pos_embed
+        x = self.pos_drop(x)
+
+        x = self.norm(x)
+        return x
+
+    # Copied from transformers.models.bert.modeling_bert.BertModel.forward
+    def forward(
+        self,
+        input_ids=None,
+        bbox=None,
+        attention_mask=None,
+        token_type_ids=None,
+        valid_span=None,
+        position_ids=None,
+        head_mask=None,
+        inputs_embeds=None,
+        encoder_hidden_states=None,
+        encoder_attention_mask=None,
+        past_key_values=None,
+        use_cache=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+        images=None,
+    ):
+        r"""
+        encoder_hidden_states  (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`):
+            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
+            the model is configured as a decoder.
+        encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
+            the cross-attention if the model is configured as a decoder. Mask values selected in ``[0, 1]``:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+        past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
+            Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.
+
+            If :obj:`past_key_values` are used, the user can optionally input only the last :obj:`decoder_input_ids`
+            (those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)`
+            instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`.
+        use_cache (:obj:`bool`, `optional`):
+            If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
+            decoding (see :obj:`past_key_values`).
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        use_cache = False
+
+        # if input_ids is not None and inputs_embeds is not None:
+        #     raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        if input_ids is not None:
+            input_shape = input_ids.size()
+            batch_size, seq_length = input_shape
+            device = input_ids.device
+        elif inputs_embeds is not None:
+            input_shape = inputs_embeds.size()[:-1]
+            batch_size, seq_length = input_shape
+            device = inputs_embeds.device
+        elif images is not None:
+            batch_size = len(images)
+            device = images.device
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds or images")
+
+        if not self.image_only:
+            # past_key_values_length
+            past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0
+
+            if attention_mask is None:
+                attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device)
+            if token_type_ids is None:
+                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
+
+        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
+        # ourselves in which case we just need to make it broadcastable to all heads.
+        # extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape, device)
+
+        encoder_extended_attention_mask = None
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape bsz x n_heads x N x N
+        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
+        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
+        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
+
+        if not self.image_only:
+            if bbox is None:
+                bbox = torch.zeros(tuple(list(input_shape) + [4]), dtype=torch.long, device=device)
+
+            embedding_output = self.embeddings(
+                input_ids=input_ids,
+                bbox=bbox,
+                position_ids=position_ids,
+                token_type_ids=token_type_ids,
+                inputs_embeds=inputs_embeds,
+                past_key_values_length=past_key_values_length,
+            )
+
+        final_bbox = final_position_ids = None
+        Hp = Wp = None
+        if images is not None:
+            patch_size = 16
+            Hp, Wp = int(images.shape[2] / patch_size), int(images.shape[3] / patch_size)
+            visual_emb = self.forward_image(images)
+            if self.detection:
+                visual_attention_mask = torch.ones((batch_size, visual_emb.shape[1]), dtype=torch.long, device=device)
+                if self.image_only:
+                    attention_mask = visual_attention_mask
+                else:
+                    attention_mask = torch.cat([attention_mask, visual_attention_mask], dim=1)
+            elif self.image_only:
+                attention_mask = torch.ones((batch_size, visual_emb.shape[1]), dtype=torch.long, device=device)
+
+            if self.config.has_relative_attention_bias or self.config.has_spatial_attention_bias:
+                if self.config.has_spatial_attention_bias:
+                    visual_bbox = self._calc_visual_bbox(device, dtype=torch.long, bsz=batch_size)
+                    if self.image_only:
+                        final_bbox = visual_bbox
+                    else:
+                        final_bbox = torch.cat([bbox, visual_bbox], dim=1)
+
+                visual_position_ids = torch.arange(0, visual_emb.shape[1], dtype=torch.long, device=device).repeat(
+                    batch_size, 1)
+                if self.image_only:
+                    final_position_ids = visual_position_ids
+                else:
+                    position_ids = torch.arange(0, input_shape[1], device=device).unsqueeze(0)
+                    position_ids = position_ids.expand_as(input_ids)
+                    final_position_ids = torch.cat([position_ids, visual_position_ids], dim=1)
+
+            if self.image_only:
+                embedding_output = visual_emb
+            else:
+                embedding_output = torch.cat([embedding_output, visual_emb], dim=1)
+            embedding_output = self.LayerNorm(embedding_output)
+            embedding_output = self.dropout(embedding_output)
+        elif self.config.has_relative_attention_bias or self.config.has_spatial_attention_bias:
+            if self.config.has_spatial_attention_bias:
+                final_bbox = bbox
+            if self.config.has_relative_attention_bias:
+                position_ids = self.embeddings.position_ids[:, :input_shape[1]]
+                position_ids = position_ids.expand_as(input_ids)
+                final_position_ids = position_ids
+
+        extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, None, device)
+
+        encoder_outputs = self.encoder(
+            embedding_output,
+            bbox=final_bbox,
+            position_ids=final_position_ids,
+            attention_mask=extended_attention_mask,
+            head_mask=head_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_extended_attention_mask,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            Hp=Hp,
+            Wp=Wp,
+            valid_span=valid_span,
+        )
+
+        if self.detection:
+            return encoder_outputs
+
+        sequence_output = encoder_outputs[0]
+        pooled_output = None
+
+        if not return_dict:
+            return (sequence_output, pooled_output) + encoder_outputs[1:]
+
+        return BaseModelOutputWithPoolingAndCrossAttentions(
+            last_hidden_state=sequence_output,
+            pooler_output=pooled_output,
+            past_key_values=encoder_outputs.past_key_values,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+            cross_attentions=encoder_outputs.cross_attentions,
+        )
+
+
+class LayoutLMv3ClassificationHead(nn.Module):
+    """
+    Head for sentence-level classification tasks.
+    Reference: RobertaClassificationHead
+    """
+
+    def __init__(self, config, pool_feature=False):
+        super().__init__()
+        self.pool_feature = pool_feature
+        if pool_feature:
+            self.dense = nn.Linear(config.hidden_size*3, config.hidden_size)
+        else:
+            self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        classifier_dropout = (
+            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
+        )
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.out_proj = nn.Linear(config.hidden_size, config.num_labels)
+
+    def forward(self, x):
+        # x = features[:, 0, :]  # take <s> token (equiv. to [CLS])
+        x = self.dropout(x)
+        x = self.dense(x)
+        x = torch.tanh(x)
+        x = self.dropout(x)
+        x = self.out_proj(x)
+        return x
+
+
+class LayoutLMv3ForTokenClassification(LayoutLMv3PreTrainedModel):
+    _keys_to_ignore_on_load_unexpected = [r"pooler"]
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.layoutlmv3 = LayoutLMv3Model(config)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        if config.num_labels < 10:
+            self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+        else:
+            self.classifier = LayoutLMv3ClassificationHead(config, pool_feature=False)
+
+        self.init_weights()
+
+    def forward(
+        self,
+        input_ids=None,
+        bbox=None,
+        attention_mask=None,
+        token_type_ids=None,
+        position_ids=None,
+        valid_span=None,
+        head_mask=None,
+        inputs_embeds=None,
+        labels=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+        images=None,
+    ):
+        r"""
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
+            Labels for computing the token classification loss. Indices should be in ``[0, ..., config.num_labels -
+            1]``.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.layoutlmv3(
+            input_ids,
+            bbox=bbox,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            images=images,
+            valid_span=valid_span,
+        )
+
+        sequence_output = outputs[0]
+
+        sequence_output = self.dropout(sequence_output)
+        logits = self.classifier(sequence_output)
+
+        loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+            # Only keep active parts of the loss
+            if attention_mask is not None:
+                active_loss = attention_mask.view(-1) == 1
+                active_logits = logits.view(-1, self.num_labels)
+                active_labels = torch.where(
+                    active_loss, labels.view(-1), torch.tensor(loss_fct.ignore_index).type_as(labels)
+                )
+                loss = loss_fct(active_logits, active_labels)
+            else:
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+class LayoutLMv3ForQuestionAnswering(LayoutLMv3PreTrainedModel):
+    _keys_to_ignore_on_load_unexpected = [r"pooler"]
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.layoutlmv3 = LayoutLMv3Model(config)
+        # self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
+        self.qa_outputs = LayoutLMv3ClassificationHead(config, pool_feature=False)
+
+        self.init_weights()
+
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        token_type_ids=None,
+        position_ids=None,
+        valid_span=None,
+        head_mask=None,
+        inputs_embeds=None,
+        start_positions=None,
+        end_positions=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+        bbox=None,
+        images=None,
+    ):
+        r"""
+        start_positions (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
+            Labels for position (index) of the start of the labelled span for computing the token classification loss.
+            Positions are clamped to the length of the sequence (:obj:`sequence_length`). Position outside of the
+            sequence are not taken into account for computing the loss.
+        end_positions (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
+            Labels for position (index) of the end of the labelled span for computing the token classification loss.
+            Positions are clamped to the length of the sequence (:obj:`sequence_length`). Position outside of the
+            sequence are not taken into account for computing the loss.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.layoutlmv3(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            bbox=bbox,
+            images=images,
+            valid_span=valid_span,
+        )
+
+        sequence_output = outputs[0]
+
+        logits = self.qa_outputs(sequence_output)
+        start_logits, end_logits = logits.split(1, dim=-1)
+        start_logits = start_logits.squeeze(-1).contiguous()
+        end_logits = end_logits.squeeze(-1).contiguous()
+
+        total_loss = None
+        if start_positions is not None and end_positions is not None:
+            # If we are on multi-GPU, split add a dimension
+            if len(start_positions.size()) > 1:
+                start_positions = start_positions.squeeze(-1)
+            if len(end_positions.size()) > 1:
+                end_positions = end_positions.squeeze(-1)
+            # sometimes the start/end positions are outside our model inputs, we ignore these terms
+            ignored_index = start_logits.size(1)
+            start_positions = start_positions.clamp(0, ignored_index)
+            end_positions = end_positions.clamp(0, ignored_index)
+
+            loss_fct = CrossEntropyLoss(ignore_index=ignored_index)
+            start_loss = loss_fct(start_logits, start_positions)
+            end_loss = loss_fct(end_logits, end_positions)
+            total_loss = (start_loss + end_loss) / 2
+
+        if not return_dict:
+            output = (start_logits, end_logits) + outputs[2:]
+            return ((total_loss,) + output) if total_loss is not None else output
+
+        return QuestionAnsweringModelOutput(
+            loss=total_loss,
+            start_logits=start_logits,
+            end_logits=end_logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+class LayoutLMv3ForSequenceClassification(LayoutLMv3PreTrainedModel):
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.config = config
+        self.layoutlmv3 = LayoutLMv3Model(config)
+        self.classifier = LayoutLMv3ClassificationHead(config, pool_feature=False)
+
+        self.init_weights()
+
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        token_type_ids=None,
+        position_ids=None,
+        valid_span=None,
+        head_mask=None,
+        inputs_embeds=None,
+        labels=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+        bbox=None,
+        images=None,
+    ):
+        r"""
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
+            Labels for computing the sequence classification/regression loss. Indices should be in :obj:`[0, ...,
+            config.num_labels - 1]`. If :obj:`config.num_labels == 1` a regression loss is computed (Mean-Square loss),
+            If :obj:`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.layoutlmv3(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            bbox=bbox,
+            images=images,
+            valid_span=valid_span,
+        )
+
+        sequence_output = outputs[0][:, 0, :]
+        logits = self.classifier(sequence_output)
+
+        loss = None
+        if labels is not None:
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(logits, labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )

magic_pdf/model/v3/helpers.py

+from collections import defaultdict
+from typing import List, Dict
+
+import torch
+from transformers import LayoutLMv3ForTokenClassification
+
+MAX_LEN = 510
+CLS_TOKEN_ID = 0
+UNK_TOKEN_ID = 3
+EOS_TOKEN_ID = 2
+
+
+class DataCollator:
+    def __call__(self, features: List[dict]) -> Dict[str, torch.Tensor]:
+        bbox = []
+        labels = []
+        input_ids = []
+        attention_mask = []
+
+        # clip bbox and labels to max length, build input_ids and attention_mask
+        for feature in features:
+            _bbox = feature["source_boxes"]
+            if len(_bbox) > MAX_LEN:
+                _bbox = _bbox[:MAX_LEN]
+            _labels = feature["target_index"]
+            if len(_labels) > MAX_LEN:
+                _labels = _labels[:MAX_LEN]
+            _input_ids = [UNK_TOKEN_ID] * len(_bbox)
+            _attention_mask = [1] * len(_bbox)
+            assert len(_bbox) == len(_labels) == len(_input_ids) == len(_attention_mask)
+            bbox.append(_bbox)
+            labels.append(_labels)
+            input_ids.append(_input_ids)
+            attention_mask.append(_attention_mask)
+
+        # add CLS and EOS tokens
+        for i in range(len(bbox)):
+            bbox[i] = [[0, 0, 0, 0]] + bbox[i] + [[0, 0, 0, 0]]
+            labels[i] = [-100] + labels[i] + [-100]
+            input_ids[i] = [CLS_TOKEN_ID] + input_ids[i] + [EOS_TOKEN_ID]
+            attention_mask[i] = [1] + attention_mask[i] + [1]
+
+        # padding to max length
+        max_len = max(len(x) for x in bbox)
+        for i in range(len(bbox)):
+            bbox[i] = bbox[i] + [[0, 0, 0, 0]] * (max_len - len(bbox[i]))
+            labels[i] = labels[i] + [-100] * (max_len - len(labels[i]))
+            input_ids[i] = input_ids[i] + [EOS_TOKEN_ID] * (max_len - len(input_ids[i]))
+            attention_mask[i] = attention_mask[i] + [0] * (
+                max_len - len(attention_mask[i])
+            )
+
+        ret = {
+            "bbox": torch.tensor(bbox),
+            "attention_mask": torch.tensor(attention_mask),
+            "labels": torch.tensor(labels),
+            "input_ids": torch.tensor(input_ids),
+        }
+        # set label > MAX_LEN to -100, because original labels may be > MAX_LEN
+        ret["labels"][ret["labels"] > MAX_LEN] = -100
+        # set label > 0 to label-1, because original labels are 1-indexed
+        ret["labels"][ret["labels"] > 0] -= 1
+        return ret
+
+
+def boxes2inputs(boxes: List[List[int]]) -> Dict[str, torch.Tensor]:
+    bbox = [[0, 0, 0, 0]] + boxes + [[0, 0, 0, 0]]
+    input_ids = [CLS_TOKEN_ID] + [UNK_TOKEN_ID] * len(boxes) + [EOS_TOKEN_ID]
+    attention_mask = [1] + [1] * len(boxes) + [1]
+    return {
+        "bbox": torch.tensor([bbox]),
+        "attention_mask": torch.tensor([attention_mask]),
+        "input_ids": torch.tensor([input_ids]),
+    }
+
+
+def prepare_inputs(
+    inputs: Dict[str, torch.Tensor], model: LayoutLMv3ForTokenClassification
+) -> Dict[str, torch.Tensor]:
+    ret = {}
+    for k, v in inputs.items():
+        v = v.to(model.device)
+        if torch.is_floating_point(v):
+            v = v.to(model.dtype)
+        ret[k] = v
+    return ret
+
+
+def parse_logits(logits: torch.Tensor, length: int) -> List[int]:
+    """
+    parse logits to orders
+
+    :param logits: logits from model
+    :param length: input length
+    :return: orders
+    """
+    logits = logits[1 : length + 1, :length]
+    orders = logits.argsort(descending=False).tolist()
+    ret = [o.pop() for o in orders]
+    while True:
+        order_to_idxes = defaultdict(list)
+        for idx, order in enumerate(ret):
+            order_to_idxes[order].append(idx)
+        # filter idxes len > 1
+        order_to_idxes = {k: v for k, v in order_to_idxes.items() if len(v) > 1}
+        if not order_to_idxes:
+            break
+        # filter
+        for order, idxes in order_to_idxes.items():
+            # find original logits of idxes
+            idxes_to_logit = {}
+            for idx in idxes:
+                idxes_to_logit[idx] = logits[idx, order]
+            idxes_to_logit = sorted(
+                idxes_to_logit.items(), key=lambda x: x[1], reverse=True
+            )
+            # keep the highest logit as order, set others to next candidate
+            for idx, _ in idxes_to_logit[1:]:
+                ret[idx] = orders[idx].pop()
+
+    return ret
+
+
+def check_duplicate(a: List[int]) -> bool:
+    return len(a) != len(set(a))