Merge pull request #3721 from Topdu/dygraph

add rec_nrtr

configs/rec/rec_mtb_nrtr.yml

+Global:
+  use_gpu: True
+  epoch_num: 21
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/rec/nrtr/
+  save_epoch_step: 1
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 2000]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints: 
+  save_inference_dir:
+  use_visualdl: False
+  infer_img: doc/imgs_words_en/word_10.png
+  # for data or label process
+  character_dict_path: 
+  character_type: EN_symbol
+  max_text_length: 25
+  infer_mode: False
+  use_space_char: True
+  save_res_path: ./output/rec/predicts_nrtr.txt
+
+Optimizer:
+  name: Adam
+  beta1: 0.9
+  beta2: 0.99
+  clip_norm: 5.0
+  lr:
+    name: Cosine
+    learning_rate: 0.0005
+    warmup_epoch: 2
+  regularizer:
+    name: 'L2'
+    factor: 0.
+
+Architecture:
+  model_type: rec
+  algorithm: NRTR
+  in_channels: 1
+  Transform:
+  Backbone:
+    name: MTB
+    cnn_num: 2
+  Head:
+    name: Transformer
+    d_model: 512
+    num_encoder_layers: 6
+    beam_size: 10 # When Beam size is greater than 0, it means to use beam search when evaluation.
+    
+
+Loss:
+  name: NRTRLoss
+  smoothing: True
+
+PostProcess:
+  name: NRTRLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ./train_data/data_lmdb_release/training/
+    transforms:
+      - NRTRDecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - NRTRLabelEncode: # Class handling label
+      - NRTRRecResizeImg:
+          image_shape: [100, 32]
+          resize_type: PIL # PIL or OpenCV
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 512
+    drop_last: True
+    num_workers: 8
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ./train_data/data_lmdb_release/evaluation/
+    transforms:
+      - NRTRDecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - NRTRLabelEncode: # Class handling label
+      - NRTRRecResizeImg:
+          image_shape: [100, 32]
+          resize_type: PIL # PIL or OpenCV
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 1
+    use_shared_memory: False

doc/doc_ch/algorithm_overview.md

@@ -44,6 +44,7 @@ PaddleOCR基于动态图开源的文本识别算法列表：
 - [x]  STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html))[11]
 - [x]  RARE([paper](https://arxiv.org/abs/1603.03915v1))[12]
 - [x]  SRN([paper](https://arxiv.org/abs/2003.12294))[5]
+- [x]  NRTR([paper](https://arxiv.org/abs/1806.00926v2))
 
 参考[DTRB][3](https://arxiv.org/abs/1904.01906)文字识别训练和评估流程，使用MJSynth和SynthText两个文字识别数据集训练，在IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE数据集上进行评估，算法效果如下：
 
@@ -58,6 +59,7 @@ PaddleOCR基于动态图开源的文本识别算法列表：
 |RARE|MobileNetV3|82.5%|rec_mv3_tps_bilstm_att |[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_tps_bilstm_att_v2.0_train.tar)|
 |RARE|Resnet34_vd|83.6%|rec_r34_vd_tps_bilstm_att |[下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_tps_bilstm_att_v2.0_train.tar)|
 |SRN|Resnet50_vd_fpn| 88.52% | rec_r50fpn_vd_none_srn | [下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r50_vd_srn_train.tar) |
+|NRTR|NRTR_MTB| 84.3% | rec_mtb_nrtr | [下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mtb_nrtr_train.tar) |
 
 
 PaddleOCR文本识别算法的训练和使用请参考文档教程中[模型训练/评估中的文本识别部分](./recognition.md)。

doc/doc_ch/recognition.md

@@ -215,6 +215,7 @@ PaddleOCR支持训练和评估交替进行, 可以在 `configs/rec/rec_icdar15_t
 | rec_mv3_tps_bilstm_att.yml |  CRNN |   Mobilenet_v3 |  TPS   |  BiLSTM |  att  |
 | rec_r34_vd_tps_bilstm_att.yml |  CRNN |   Resnet34_vd |  TPS   |  BiLSTM |  att  |
 | rec_r50fpn_vd_none_srn.yml    | SRN | Resnet50_fpn_vd    | None    | rnn | srn |
+| rec_mtb_nrtr.yml    | NRTR | nrtr_mtb    | None    | transformer encoder | transformer decoder |
 
 训练中文数据，推荐使用[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)，如您希望尝试其他算法在中文数据集上的效果，请参考下列说明修改配置文件：
 

doc/doc_en/algorithm_overview_en.md

@@ -46,6 +46,7 @@ PaddleOCR open-source text recognition algorithms list:
 - [x]  STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html))[11]
 - [x]  RARE([paper](https://arxiv.org/abs/1603.03915v1))[12]
 - [x]  SRN([paper](https://arxiv.org/abs/2003.12294))[5]
+- [x]  NRTR([paper](https://arxiv.org/abs/1806.00926v2))
 
 Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation result of these above text recognition (using MJSynth and SynthText for training, evaluate on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE) is as follow:
 
@@ -60,5 +61,6 @@ Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation r
 |RARE|MobileNetV3|82.5%|rec_mv3_tps_bilstm_att |[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_tps_bilstm_att_v2.0_train.tar)|
 |RARE|Resnet34_vd|83.6%|rec_r34_vd_tps_bilstm_att |[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_tps_bilstm_att_v2.0_train.tar)|
 |SRN|Resnet50_vd_fpn| 88.52% | rec_r50fpn_vd_none_srn |[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r50_vd_srn_train.tar)|
+|NRTR|NRTR_MTB| 84.3% | rec_mtb_nrtr | [Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mtb_nrtr_train.tar) |
 
 Please refer to the document for training guide and use of PaddleOCR text recognition algorithms [Text recognition model training/evaluation/prediction](./recognition_en.md)

doc/doc_en/recognition_en.md

@@ -207,7 +207,7 @@ If the evaluation set is large, the test will be time-consuming. It is recommend
 | rec_mv3_tps_bilstm_att.yml |  CRNN |   Mobilenet_v3 |  TPS   |  BiLSTM |  att  |
 | rec_r34_vd_tps_bilstm_att.yml |  CRNN |   Resnet34_vd |  TPS   |  BiLSTM |  att  |
 | rec_r50fpn_vd_none_srn.yml    | SRN | Resnet50_fpn_vd    | None    | rnn | srn |
-
+| rec_mtb_nrtr.yml    | NRTR | nrtr_mtb    | None    | transformer encoder | transformer decoder |
 
 For training Chinese data, it is recommended to use
 [rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml). If you want to try the result of other algorithms on the Chinese data set, please refer to the following instructions to modify the configuration file:

ppocr/data/imaug/__init__.py

@@ -21,7 +21,7 @@ from .make_border_map import MakeBorderMap
 from .make_shrink_map import MakeShrinkMap
 from .random_crop_data import EastRandomCropData, PSERandomCrop
 
-from .rec_img_aug import RecAug, RecResizeImg, ClsResizeImg, SRNRecResizeImg
+from .rec_img_aug import RecAug, RecResizeImg, ClsResizeImg, SRNRecResizeImg, NRTRRecResizeImg
 from .randaugment import RandAugment
 from .copy_paste import CopyPaste
 from .operators import *

ppocr/data/imaug/label_ops.py

@@ -161,6 +161,34 @@ class BaseRecLabelEncode(object):
         return text_list
 
 
+class NRTRLabelEncode(BaseRecLabelEncode):
+    """ Convert between text-label and text-index """
+
+    def __init__(self,
+                 max_text_length,
+                 character_dict_path=None,
+                 character_type='EN_symbol',
+                 use_space_char=False,
+                 **kwargs):
+
+        super(NRTRLabelEncode,
+              self).__init__(max_text_length, character_dict_path,
+                             character_type, use_space_char)
+    def __call__(self, data):
+        text = data['label']
+        text = self.encode(text)
+        if text is None:
+            return None
+        data['length'] = np.array(len(text))
+        text.insert(0, 2)
+        text.append(3)
+        text = text + [0] * (self.max_text_len - len(text))
+        data['label'] = np.array(text)
+        return data
+    def add_special_char(self, dict_character):
+        dict_character = ['blank','<unk>','<s>','</s>'] + dict_character
+        return dict_character
+
 class CTCLabelEncode(BaseRecLabelEncode):
     """ Convert between text-label and text-index """
 

ppocr/data/imaug/operators.py

@@ -57,6 +57,38 @@ class DecodeImage(object):
         return data
 
 
+class NRTRDecodeImage(object):
+    """ decode image """
+
+    def __init__(self, img_mode='RGB', channel_first=False, **kwargs):
+        self.img_mode = img_mode
+        self.channel_first = channel_first
+
+    def __call__(self, data):
+        img = data['image']
+        if six.PY2:
+            assert type(img) is str and len(
+                img) > 0, "invalid input 'img' in DecodeImage"
+        else:
+            assert type(img) is bytes and len(
+                img) > 0, "invalid input 'img' in DecodeImage"
+        img = np.frombuffer(img, dtype='uint8')
+
+        img = cv2.imdecode(img, 1)
+
+        if img is None:
+            return None
+        if self.img_mode == 'GRAY':
+            img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
+        elif self.img_mode == 'RGB':
+            assert img.shape[2] == 3, 'invalid shape of image[%s]' % (img.shape)
+            img = img[:, :, ::-1]
+        img = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
+        if self.channel_first:
+            img = img.transpose((2, 0, 1))
+        data['image'] = img
+        return data
+
 class NormalizeImage(object):
     """ normalize image such as substract mean, divide std
     """

ppocr/data/imaug/rec_img_aug.py

@@ -16,7 +16,7 @@ import math
 import cv2
 import numpy as np
 import random
-
+from PIL import Image
 from .text_image_aug import tia_perspective, tia_stretch, tia_distort
 
 
@@ -43,6 +43,25 @@ class ClsResizeImg(object):
         return data
 
 
+class NRTRRecResizeImg(object):
+    def __init__(self, image_shape, resize_type, **kwargs):
+        self.image_shape = image_shape
+        self.resize_type = resize_type
+
+    def __call__(self, data):
+        img = data['image']
+        if self.resize_type == 'PIL':
+            image_pil = Image.fromarray(np.uint8(img))
+            img = image_pil.resize(self.image_shape, Image.ANTIALIAS)
+            img = np.array(img)
+        if self.resize_type == 'OpenCV':
+            img = cv2.resize(img, self.image_shape)
+        norm_img = np.expand_dims(img, -1)
+        norm_img = norm_img.transpose((2, 0, 1))
+        data['image'] = norm_img.astype(np.float32) / 128. - 1.
+        return data
+
+
 class RecResizeImg(object):
     def __init__(self,
                  image_shape,

ppocr/losses/__init__.py

@@ -25,7 +25,7 @@ from .det_sast_loss import SASTLoss
 from .rec_ctc_loss import CTCLoss
 from .rec_att_loss import AttentionLoss
 from .rec_srn_loss import SRNLoss
-
+from .rec_nrtr_loss import NRTRLoss
 # cls loss
 from .cls_loss import ClsLoss
 
@@ -44,8 +44,9 @@ from .table_att_loss import TableAttentionLoss
 def build_loss(config):
     support_dict = [
         'DBLoss', 'EASTLoss', 'SASTLoss', 'CTCLoss', 'ClsLoss', 'AttentionLoss',
-        'SRNLoss', 'PGLoss', 'CombinedLoss', 'TableAttentionLoss'
+        'SRNLoss', 'PGLoss', 'CombinedLoss', 'NRTRLoss', 'TableAttentionLoss'
     ]
+
     config = copy.deepcopy(config)
     module_name = config.pop('name')
     assert module_name in support_dict, Exception('loss only support {}'.format(

ppocr/losses/rec_nrtr_loss.py

+import paddle
+from paddle import nn
+import paddle.nn.functional as F
+
+
+class NRTRLoss(nn.Layer):
+    def __init__(self, smoothing=True, **kwargs):
+        super(NRTRLoss, self).__init__()
+        self.loss_func = nn.CrossEntropyLoss(reduction='mean', ignore_index=0)
+        self.smoothing = smoothing
+
+    def forward(self, pred, batch):
+        pred = pred.reshape([-1, pred.shape[2]])
+        max_len = batch[2].max()
+        tgt = batch[1][:, 1:2 + max_len]
+        tgt = tgt.reshape([-1])
+        if self.smoothing:
+            eps = 0.1
+            n_class = pred.shape[1]
+            one_hot = F.one_hot(tgt, pred.shape[1])
+            one_hot = one_hot * (1 - eps) + (1 - one_hot) * eps / (n_class - 1)
+            log_prb = F.log_softmax(pred, axis=1)
+            non_pad_mask = paddle.not_equal(
+                tgt, paddle.zeros(
+                    tgt.shape, dtype='int64'))
+            loss = -(one_hot * log_prb).sum(axis=1)
+            loss = loss.masked_select(non_pad_mask).mean()
+        else:
+            loss = self.loss_func(pred, tgt)
+        return {'loss': loss}

ppocr/metrics/rec_metric.py

@@ -57,3 +57,4 @@ class RecMetric(object):
         self.correct_num = 0
         self.all_num = 0
         self.norm_edit_dis = 0
+        

ppocr/modeling/architectures/base_model.py

@@ -14,7 +14,6 @@
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
-
 from paddle import nn
 from ppocr.modeling.transforms import build_transform
 from ppocr.modeling.backbones import build_backbone

ppocr/modeling/backbones/__init__.py

@@ -26,8 +26,9 @@ def build_backbone(config, model_type):
         from .rec_resnet_vd import ResNet
         from .rec_resnet_fpn import ResNetFPN
         from .rec_mv1_enhance import MobileNetV1Enhance
+        from .rec_nrtr_mtb import MTB
         support_dict = [
-            "MobileNetV1Enhance", "MobileNetV3", "ResNet", "ResNetFPN"
+            'MobileNetV1Enhance', 'MobileNetV3', 'ResNet', 'ResNetFPN', 'MTB'
         ]
     elif model_type == "e2e":
         from .e2e_resnet_vd_pg import ResNet

ppocr/modeling/backbones/rec_nrtr_mtb.py

+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+from paddle import nn
+
+
+class MTB(nn.Layer):
+    def __init__(self, cnn_num, in_channels):
+        super(MTB, self).__init__()
+        self.block = nn.Sequential()
+        self.out_channels = in_channels
+        self.cnn_num = cnn_num
+        if self.cnn_num == 2:
+            for i in range(self.cnn_num):
+                self.block.add_sublayer(
+                    'conv_{}'.format(i),
+                    nn.Conv2D(
+                        in_channels=in_channels
+                        if i == 0 else 32 * (2**(i - 1)),
+                        out_channels=32 * (2**i),
+                        kernel_size=3,
+                        stride=2,
+                        padding=1))
+                self.block.add_sublayer('relu_{}'.format(i), nn.ReLU())
+                self.block.add_sublayer('bn_{}'.format(i),
+                                        nn.BatchNorm2D(32 * (2**i)))
+
+    def forward(self, images):
+        x = self.block(images)
+        if self.cnn_num == 2:
+            # (b, w, h, c)
+            x = x.transpose([0, 3, 2, 1])
+            x_shape = x.shape
+            x = x.reshape([x_shape[0], x_shape[1], x_shape[2] * x_shape[3]])
+        return x

ppocr/modeling/heads/__init__.py

@@ -26,12 +26,14 @@ def build_head(config):
     from .rec_ctc_head import CTCHead
     from .rec_att_head import AttentionHead
     from .rec_srn_head import SRNHead
+    from .rec_nrtr_head import Transformer
 
     # cls head
     from .cls_head import ClsHead
     support_dict = [
         'DBHead', 'EASTHead', 'SASTHead', 'CTCHead', 'ClsHead', 'AttentionHead',
-        'SRNHead', 'PGHead', 'TableAttentionHead']
+        'SRNHead', 'PGHead', 'Transformer', 'TableAttentionHead'
+    ]
 
     #table head
     from .table_att_head import TableAttentionHead

ppocr/modeling/heads/multiheadAttention.py

+# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+import paddle
+from paddle import nn
+import paddle.nn.functional as F
+from paddle.nn import Linear
+from paddle.nn.initializer import XavierUniform as xavier_uniform_
+from paddle.nn.initializer import Constant as constant_
+from paddle.nn.initializer import XavierNormal as xavier_normal_
+
+zeros_ = constant_(value=0.)
+ones_ = constant_(value=1.)
+
+
+class MultiheadAttention(nn.Layer):
+    """Allows the model to jointly attend to information
+    from different representation subspaces.
+    See reference: Attention Is All You Need
+
+    .. math::
+        \text{MultiHead}(Q, K, V) = \text{Concat}(head_1,\dots,head_h)W^O
+        \text{where} head_i = \text{Attention}(QW_i^Q, KW_i^K, VW_i^V)
+
+    Args:
+        embed_dim: total dimension of the model
+        num_heads: parallel attention layers, or heads
+
+    """
+
+    def __init__(self,
+                 embed_dim,
+                 num_heads,
+                 dropout=0.,
+                 bias=True,
+                 add_bias_kv=False,
+                 add_zero_attn=False):
+        super(MultiheadAttention, self).__init__()
+        self.embed_dim = embed_dim
+        self.num_heads = num_heads
+        self.dropout = dropout
+        self.head_dim = embed_dim // num_heads
+        assert self.head_dim * num_heads == self.embed_dim, "embed_dim must be divisible by num_heads"
+        self.scaling = self.head_dim**-0.5
+        self.out_proj = Linear(embed_dim, embed_dim, bias_attr=bias)
+        self._reset_parameters()
+        self.conv1 = paddle.nn.Conv2D(
+            in_channels=embed_dim, out_channels=embed_dim, kernel_size=(1, 1))
+        self.conv2 = paddle.nn.Conv2D(
+            in_channels=embed_dim, out_channels=embed_dim, kernel_size=(1, 1))
+        self.conv3 = paddle.nn.Conv2D(
+            in_channels=embed_dim, out_channels=embed_dim, kernel_size=(1, 1))
+
+    def _reset_parameters(self):
+        xavier_uniform_(self.out_proj.weight)
+
+    def forward(self,
+                query,
+                key,
+                value,
+                key_padding_mask=None,
+                incremental_state=None,
+                need_weights=True,
+                static_kv=False,
+                attn_mask=None):
+        """
+        Inputs of forward function
+            query: [target length, batch size, embed dim]
+            key: [sequence length, batch size, embed dim]
+            value: [sequence length, batch size, embed dim]
+            key_padding_mask: if True, mask padding based on batch size
+            incremental_state: if provided, previous time steps are cashed
+            need_weights: output attn_output_weights
+            static_kv: key and value are static
+
+        Outputs of forward function
+            attn_output: [target length, batch size, embed dim]
+            attn_output_weights: [batch size, target length, sequence length]
+        """
+        tgt_len, bsz, embed_dim = query.shape
+        assert embed_dim == self.embed_dim
+        assert list(query.shape) == [tgt_len, bsz, embed_dim]
+        assert key.shape == value.shape
+
+        q = self._in_proj_q(query)
+        k = self._in_proj_k(key)
+        v = self._in_proj_v(value)
+        q *= self.scaling
+
+        q = q.reshape([tgt_len, bsz * self.num_heads, self.head_dim]).transpose(
+            [1, 0, 2])
+        k = k.reshape([-1, bsz * self.num_heads, self.head_dim]).transpose(
+            [1, 0, 2])
+        v = v.reshape([-1, bsz * self.num_heads, self.head_dim]).transpose(
+            [1, 0, 2])
+
+        src_len = k.shape[1]
+
+        if key_padding_mask is not None:
+            assert key_padding_mask.shape[0] == bsz
+            assert key_padding_mask.shape[1] == src_len
+
+        attn_output_weights = paddle.bmm(q, k.transpose([0, 2, 1]))
+        assert list(attn_output_weights.
+                    shape) == [bsz * self.num_heads, tgt_len, src_len]
+
+        if attn_mask is not None:
+            attn_mask = attn_mask.unsqueeze(0)
+            attn_output_weights += attn_mask
+        if key_padding_mask is not None:
+            attn_output_weights = attn_output_weights.reshape(
+                [bsz, self.num_heads, tgt_len, src_len])
+            key = key_padding_mask.unsqueeze(1).unsqueeze(2).astype('float32')
+            y = paddle.full(shape=key.shape, dtype='float32', fill_value='-inf')
+            y = paddle.where(key == 0., key, y)
+            attn_output_weights += y
+            attn_output_weights = attn_output_weights.reshape(
+                [bsz * self.num_heads, tgt_len, src_len])
+
+        attn_output_weights = F.softmax(
+            attn_output_weights.astype('float32'),
+            axis=-1,
+            dtype=paddle.float32 if attn_output_weights.dtype == paddle.float16
+            else attn_output_weights.dtype)
+        attn_output_weights = F.dropout(
+            attn_output_weights, p=self.dropout, training=self.training)
+
+        attn_output = paddle.bmm(attn_output_weights, v)
+        assert list(attn_output.
+                    shape) == [bsz * self.num_heads, tgt_len, self.head_dim]
+        attn_output = attn_output.transpose([1, 0, 2]).reshape(
+            [tgt_len, bsz, embed_dim])
+        attn_output = self.out_proj(attn_output)
+
+        if need_weights:
+            # average attention weights over heads
+            attn_output_weights = attn_output_weights.reshape(
+                [bsz, self.num_heads, tgt_len, src_len])
+            attn_output_weights = attn_output_weights.sum(
+                axis=1) / self.num_heads
+        else:
+            attn_output_weights = None
+        return attn_output, attn_output_weights
+
+    def _in_proj_q(self, query):
+        query = query.transpose([1, 2, 0])
+        query = paddle.unsqueeze(query, axis=2)
+        res = self.conv1(query)
+        res = paddle.squeeze(res, axis=2)
+        res = res.transpose([2, 0, 1])
+        return res
+
+    def _in_proj_k(self, key):
+        key = key.transpose([1, 2, 0])
+        key = paddle.unsqueeze(key, axis=2)
+        res = self.conv2(key)
+        res = paddle.squeeze(res, axis=2)
+        res = res.transpose([2, 0, 1])
+        return res
+
+    def _in_proj_v(self, value):
+        value = value.transpose([1, 2, 0])  #(1, 2, 0)
+        value = paddle.unsqueeze(value, axis=2)
+        res = self.conv3(value)
+        res = paddle.squeeze(res, axis=2)
+        res = res.transpose([2, 0, 1])
+        return res

ppocr/modeling/heads/rec_nrtr_head.py

+# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+import math
+import paddle
+import copy
+from paddle import nn
+import paddle.nn.functional as F
+from paddle.nn import LayerList
+from paddle.nn.initializer import XavierNormal as xavier_uniform_
+from paddle.nn import Dropout, Linear, LayerNorm, Conv2D
+import numpy as np
+from ppocr.modeling.heads.multiheadAttention import MultiheadAttention
+from paddle.nn.initializer import Constant as constant_
+from paddle.nn.initializer import XavierNormal as xavier_normal_
+
+zeros_ = constant_(value=0.)
+ones_ = constant_(value=1.)
+
+
+class Transformer(nn.Layer):
+    """A transformer model. User is able to modify the attributes as needed. The architechture
+    is based on the paper "Attention Is All You Need". Ashish Vaswani, Noam Shazeer,
+    Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and
+    Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information
+    Processing Systems, pages 6000-6010.
+
+    Args:
+        d_model: the number of expected features in the encoder/decoder inputs (default=512).
+        nhead: the number of heads in the multiheadattention models (default=8).
+        num_encoder_layers: the number of sub-encoder-layers in the encoder (default=6).
+        num_decoder_layers: the number of sub-decoder-layers in the decoder (default=6).
+        dim_feedforward: the dimension of the feedforward network model (default=2048).
+        dropout: the dropout value (default=0.1).
+        custom_encoder: custom encoder (default=None).
+        custom_decoder: custom decoder (default=None).
+
+    """
+
+    def __init__(self,
+                 d_model=512,
+                 nhead=8,
+                 num_encoder_layers=6,
+                 beam_size=0,
+                 num_decoder_layers=6,
+                 dim_feedforward=1024,
+                 attention_dropout_rate=0.0,
+                 residual_dropout_rate=0.1,
+                 custom_encoder=None,
+                 custom_decoder=None,
+                 in_channels=0,
+                 out_channels=0,
+                 dst_vocab_size=99,
+                 scale_embedding=True):
+        super(Transformer, self).__init__()
+        self.embedding = Embeddings(
+            d_model=d_model,
+            vocab=dst_vocab_size,
+            padding_idx=0,
+            scale_embedding=scale_embedding)
+        self.positional_encoding = PositionalEncoding(
+            dropout=residual_dropout_rate,
+            dim=d_model, )
+        if custom_encoder is not None:
+            self.encoder = custom_encoder
+        else:
+            if num_encoder_layers > 0:
+                encoder_layer = TransformerEncoderLayer(
+                    d_model, nhead, dim_feedforward, attention_dropout_rate,
+                    residual_dropout_rate)
+                self.encoder = TransformerEncoder(encoder_layer,
+                                                  num_encoder_layers)
+            else:
+                self.encoder = None
+
+        if custom_decoder is not None:
+            self.decoder = custom_decoder
+        else:
+            decoder_layer = TransformerDecoderLayer(
+                d_model, nhead, dim_feedforward, attention_dropout_rate,
+                residual_dropout_rate)
+            self.decoder = TransformerDecoder(decoder_layer, num_decoder_layers)
+
+        self._reset_parameters()
+        self.beam_size = beam_size
+        self.d_model = d_model
+        self.nhead = nhead
+        self.tgt_word_prj = nn.Linear(d_model, dst_vocab_size, bias_attr=False)
+        w0 = np.random.normal(0.0, d_model**-0.5,
+                              (d_model, dst_vocab_size)).astype(np.float32)
+        self.tgt_word_prj.weight.set_value(w0)
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m):
+
+        if isinstance(m, nn.Conv2D):
+            xavier_normal_(m.weight)
+            if m.bias is not None:
+                zeros_(m.bias)
+
+    def forward_train(self, src, tgt):
+        tgt = tgt[:, :-1]
+
+        tgt_key_padding_mask = self.generate_padding_mask(tgt)
+        tgt = self.embedding(tgt).transpose([1, 0, 2])
+        tgt = self.positional_encoding(tgt)
+        tgt_mask = self.generate_square_subsequent_mask(tgt.shape[0])
+
+        if self.encoder is not None:
+            src = self.positional_encoding(src.transpose([1, 0, 2]))
+            memory = self.encoder(src)
+        else:
+            memory = src.squeeze(2).transpose([2, 0, 1])
+        output = self.decoder(
+            tgt,
+            memory,
+            tgt_mask=tgt_mask,
+            memory_mask=None,
+            tgt_key_padding_mask=tgt_key_padding_mask,
+            memory_key_padding_mask=None)
+        output = output.transpose([1, 0, 2])
+        logit = self.tgt_word_prj(output)
+        return logit
+
+    def forward(self, src, targets=None):
+        """Take in and process masked source/target sequences.
+        Args:
+            src: the sequence to the encoder (required).
+            tgt: the sequence to the decoder (required).
+        Shape:
+            - src: :math:`(S, N, E)`.
+            - tgt: :math:`(T, N, E)`.
+        Examples:
+            >>> output = transformer_model(src, tgt)
+        """
+
+        if self.training:
+            max_len = targets[1].max()
+            tgt = targets[0][:, :2 + max_len]
+            return self.forward_train(src, tgt)
+        else:
+            if self.beam_size > 0:
+                return self.forward_beam(src)
+            else:
+                return self.forward_test(src)
+
+    def forward_test(self, src):
+        bs = src.shape[0]
+        if self.encoder is not None:
+            src = self.positional_encoding(src.transpose([1, 0, 2]))
+            memory = self.encoder(src)
+        else:
+            memory = src.squeeze(2).transpose([2, 0, 1])
+        dec_seq = paddle.full((bs, 1), 2, dtype=paddle.int64)
+        for len_dec_seq in range(1, 25):
+            src_enc = memory.clone()
+            tgt_key_padding_mask = self.generate_padding_mask(dec_seq)
+            dec_seq_embed = self.embedding(dec_seq).transpose([1, 0, 2])
+            dec_seq_embed = self.positional_encoding(dec_seq_embed)
+            tgt_mask = self.generate_square_subsequent_mask(dec_seq_embed.shape[
+                0])
+            output = self.decoder(
+                dec_seq_embed,
+                src_enc,
+                tgt_mask=tgt_mask,
+                memory_mask=None,
+                tgt_key_padding_mask=tgt_key_padding_mask,
+                memory_key_padding_mask=None)
+            dec_output = output.transpose([1, 0, 2])
+
+            dec_output = dec_output[:,
+                                    -1, :]  # Pick the last step: (bh * bm) * d_h
+            word_prob = F.log_softmax(self.tgt_word_prj(dec_output), axis=1)
+            word_prob = word_prob.reshape([1, bs, -1])
+            preds_idx = word_prob.argmax(axis=2)
+
+            if paddle.equal_all(
+                    preds_idx[-1],
+                    paddle.full(
+                        preds_idx[-1].shape, 3, dtype='int64')):
+                break
+
+            preds_prob = word_prob.max(axis=2)
+            dec_seq = paddle.concat(
+                [dec_seq, preds_idx.reshape([-1, 1])], axis=1)
+
+        return dec_seq
+
+    def forward_beam(self, images):
+        ''' Translation work in one batch '''
+
+        def get_inst_idx_to_tensor_position_map(inst_idx_list):
+            ''' Indicate the position of an instance in a tensor. '''
+            return {
+                inst_idx: tensor_position
+                for tensor_position, inst_idx in enumerate(inst_idx_list)
+            }
+
+        def collect_active_part(beamed_tensor, curr_active_inst_idx,
+                                n_prev_active_inst, n_bm):
+            ''' Collect tensor parts associated to active instances. '''
+
+            _, *d_hs = beamed_tensor.shape
+            n_curr_active_inst = len(curr_active_inst_idx)
+            new_shape = (n_curr_active_inst * n_bm, *d_hs)
+
+            beamed_tensor = beamed_tensor.reshape([n_prev_active_inst, -1])
+            beamed_tensor = beamed_tensor.index_select(
+                paddle.to_tensor(curr_active_inst_idx), axis=0)
+            beamed_tensor = beamed_tensor.reshape([*new_shape])
+
+            return beamed_tensor
+
+        def collate_active_info(src_enc, inst_idx_to_position_map,
+                                active_inst_idx_list):
+            # Sentences which are still active are collected,
+            # so the decoder will not run on completed sentences.
+
+            n_prev_active_inst = len(inst_idx_to_position_map)
+            active_inst_idx = [
+                inst_idx_to_position_map[k] for k in active_inst_idx_list
+            ]
+            active_inst_idx = paddle.to_tensor(active_inst_idx, dtype='int64')
+            active_src_enc = collect_active_part(
+                src_enc.transpose([1, 0, 2]), active_inst_idx,
+                n_prev_active_inst, n_bm).transpose([1, 0, 2])
+            active_inst_idx_to_position_map = get_inst_idx_to_tensor_position_map(
+                active_inst_idx_list)
+            return active_src_enc, active_inst_idx_to_position_map
+
+        def beam_decode_step(inst_dec_beams, len_dec_seq, enc_output,
+                             inst_idx_to_position_map, n_bm,
+                             memory_key_padding_mask):
+            ''' Decode and update beam status, and then return active beam idx '''
+
+            def prepare_beam_dec_seq(inst_dec_beams, len_dec_seq):
+                dec_partial_seq = [
+                    b.get_current_state() for b in inst_dec_beams if not b.done
+                ]
+                dec_partial_seq = paddle.stack(dec_partial_seq)
+
+                dec_partial_seq = dec_partial_seq.reshape([-1, len_dec_seq])
+                return dec_partial_seq
+
+            def prepare_beam_memory_key_padding_mask(
+                    inst_dec_beams, memory_key_padding_mask, n_bm):
+                keep = []
+                for idx in (memory_key_padding_mask):
+                    if not inst_dec_beams[idx].done:
+                        keep.append(idx)
+                memory_key_padding_mask = memory_key_padding_mask[
+                    paddle.to_tensor(keep)]
+                len_s = memory_key_padding_mask.shape[-1]
+                n_inst = memory_key_padding_mask.shape[0]
+                memory_key_padding_mask = paddle.concat(
+                    [memory_key_padding_mask for i in range(n_bm)], axis=1)
+                memory_key_padding_mask = memory_key_padding_mask.reshape(
+                    [n_inst * n_bm, len_s])  #repeat(1, n_bm)
+                return memory_key_padding_mask
+
+            def predict_word(dec_seq, enc_output, n_active_inst, n_bm,
+                             memory_key_padding_mask):
+                tgt_key_padding_mask = self.generate_padding_mask(dec_seq)
+                dec_seq = self.embedding(dec_seq).transpose([1, 0, 2])
+                dec_seq = self.positional_encoding(dec_seq)
+                tgt_mask = self.generate_square_subsequent_mask(dec_seq.shape[
+                    0])
+                dec_output = self.decoder(
+                    dec_seq,
+                    enc_output,
+                    tgt_mask=tgt_mask,
+                    tgt_key_padding_mask=tgt_key_padding_mask,
+                    memory_key_padding_mask=memory_key_padding_mask,
+                ).transpose([1, 0, 2])
+                dec_output = dec_output[:,
+                                        -1, :]  # Pick the last step: (bh * bm) * d_h
+                word_prob = F.log_softmax(self.tgt_word_prj(dec_output), axis=1)
+                word_prob = word_prob.reshape([n_active_inst, n_bm, -1])
+                return word_prob
+
+            def collect_active_inst_idx_list(inst_beams, word_prob,
+                                             inst_idx_to_position_map):
+                active_inst_idx_list = []
+                for inst_idx, inst_position in inst_idx_to_position_map.items():
+                    is_inst_complete = inst_beams[inst_idx].advance(word_prob[
+                        inst_position])
+                    if not is_inst_complete:
+                        active_inst_idx_list += [inst_idx]
+
+                return active_inst_idx_list
+
+            n_active_inst = len(inst_idx_to_position_map)
+            dec_seq = prepare_beam_dec_seq(inst_dec_beams, len_dec_seq)
+            memory_key_padding_mask = None
+            word_prob = predict_word(dec_seq, enc_output, n_active_inst, n_bm,
+                                     memory_key_padding_mask)
+            # Update the beam with predicted word prob information and collect incomplete instances
+            active_inst_idx_list = collect_active_inst_idx_list(
+                inst_dec_beams, word_prob, inst_idx_to_position_map)
+            return active_inst_idx_list
+
+        def collect_hypothesis_and_scores(inst_dec_beams, n_best):
+            all_hyp, all_scores = [], []
+            for inst_idx in range(len(inst_dec_beams)):
+                scores, tail_idxs = inst_dec_beams[inst_idx].sort_scores()
+                all_scores += [scores[:n_best]]
+                hyps = [
+                    inst_dec_beams[inst_idx].get_hypothesis(i)
+                    for i in tail_idxs[:n_best]
+                ]
+                all_hyp += [hyps]
+            return all_hyp, all_scores
+
+        with paddle.no_grad():
+            #-- Encode
+
+            if self.encoder is not None:
+                src = self.positional_encoding(images.transpose([1, 0, 2]))
+                src_enc = self.encoder(src).transpose([1, 0, 2])
+            else:
+                src_enc = images.squeeze(2).transpose([0, 2, 1])
+
+            #-- Repeat data for beam search
+            n_bm = self.beam_size
+            n_inst, len_s, d_h = src_enc.shape
+            src_enc = paddle.concat([src_enc for i in range(n_bm)], axis=1)
+            src_enc = src_enc.reshape([n_inst * n_bm, len_s, d_h]).transpose(
+                [1, 0, 2])
+            #-- Prepare beams
+            inst_dec_beams = [Beam(n_bm) for _ in range(n_inst)]
+
+            #-- Bookkeeping for active or not
+            active_inst_idx_list = list(range(n_inst))
+            inst_idx_to_position_map = get_inst_idx_to_tensor_position_map(
+                active_inst_idx_list)
+            #-- Decode
+            for len_dec_seq in range(1, 25):
+                src_enc_copy = src_enc.clone()
+                active_inst_idx_list = beam_decode_step(
+                    inst_dec_beams, len_dec_seq, src_enc_copy,
+                    inst_idx_to_position_map, n_bm, None)
+                if not active_inst_idx_list:
+                    break  # all instances have finished their path to <EOS>
+                src_enc, inst_idx_to_position_map = collate_active_info(
+                    src_enc_copy, inst_idx_to_position_map,
+                    active_inst_idx_list)
+        batch_hyp, batch_scores = collect_hypothesis_and_scores(inst_dec_beams,
+                                                                1)
+        result_hyp = []
+        for bs_hyp in batch_hyp:
+            bs_hyp_pad = bs_hyp[0] + [3] * (25 - len(bs_hyp[0]))
+            result_hyp.append(bs_hyp_pad)
+        return paddle.to_tensor(np.array(result_hyp), dtype=paddle.int64)
+
+    def generate_square_subsequent_mask(self, sz):
+        """Generate a square mask for the sequence. The masked positions are filled with float('-inf').
+            Unmasked positions are filled with float(0.0).
+        """
+        mask = paddle.zeros([sz, sz], dtype='float32')
+        mask_inf = paddle.triu(
+            paddle.full(
+                shape=[sz, sz], dtype='float32', fill_value='-inf'),
+            diagonal=1)
+        mask = mask + mask_inf
+        return mask
+
+    def generate_padding_mask(self, x):
+        padding_mask = x.equal(paddle.to_tensor(0, dtype=x.dtype))
+        return padding_mask
+
+    def _reset_parameters(self):
+        """Initiate parameters in the transformer model."""
+
+        for p in self.parameters():
+            if p.dim() > 1:
+                xavier_uniform_(p)
+
+
+class TransformerEncoder(nn.Layer):
+    """TransformerEncoder is a stack of N encoder layers
+    Args:
+        encoder_layer: an instance of the TransformerEncoderLayer() class (required).
+        num_layers: the number of sub-encoder-layers in the encoder (required).
+        norm: the layer normalization component (optional).
+    """
+
+    def __init__(self, encoder_layer, num_layers):
+        super(TransformerEncoder, self).__init__()
+        self.layers = _get_clones(encoder_layer, num_layers)
+        self.num_layers = num_layers
+
+    def forward(self, src):
+        """Pass the input through the endocder layers in turn.
+        Args:
+            src: the sequnce to the encoder (required).
+            mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+        """
+        output = src
+
+        for i in range(self.num_layers):
+            output = self.layers[i](output,
+                                    src_mask=None,
+                                    src_key_padding_mask=None)
+
+        return output
+
+
+class TransformerDecoder(nn.Layer):
+    """TransformerDecoder is a stack of N decoder layers
+
+    Args:
+        decoder_layer: an instance of the TransformerDecoderLayer() class (required).
+        num_layers: the number of sub-decoder-layers in the decoder (required).
+        norm: the layer normalization component (optional).
+
+    """
+
+    def __init__(self, decoder_layer, num_layers):
+        super(TransformerDecoder, self).__init__()
+        self.layers = _get_clones(decoder_layer, num_layers)
+        self.num_layers = num_layers
+
+    def forward(self,
+                tgt,
+                memory,
+                tgt_mask=None,
+                memory_mask=None,
+                tgt_key_padding_mask=None,
+                memory_key_padding_mask=None):
+        """Pass the inputs (and mask) through the decoder layer in turn.
+
+        Args:
+            tgt: the sequence to the decoder (required).
+            memory: the sequnce from the last layer of the encoder (required).
+            tgt_mask: the mask for the tgt sequence (optional).
+            memory_mask: the mask for the memory sequence (optional).
+            tgt_key_padding_mask: the mask for the tgt keys per batch (optional).
+            memory_key_padding_mask: the mask for the memory keys per batch (optional).
+        """
+        output = tgt
+        for i in range(self.num_layers):
+            output = self.layers[i](
+                output,
+                memory,
+                tgt_mask=tgt_mask,
+                memory_mask=memory_mask,
+                tgt_key_padding_mask=tgt_key_padding_mask,
+                memory_key_padding_mask=memory_key_padding_mask)
+
+        return output
+
+
+class TransformerEncoderLayer(nn.Layer):
+    """TransformerEncoderLayer is made up of self-attn and feedforward network.
+    This standard encoder layer is based on the paper "Attention Is All You Need".
+    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez,
+    Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in
+    Neural Information Processing Systems, pages 6000-6010. Users may modify or implement
+    in a different way during application.
+
+    Args:
+        d_model: the number of expected features in the input (required).
+        nhead: the number of heads in the multiheadattention models (required).
+        dim_feedforward: the dimension of the feedforward network model (default=2048).
+        dropout: the dropout value (default=0.1).
+
+    """
+
+    def __init__(self,
+                 d_model,
+                 nhead,
+                 dim_feedforward=2048,
+                 attention_dropout_rate=0.0,
+                 residual_dropout_rate=0.1):
+        super(TransformerEncoderLayer, self).__init__()
+        self.self_attn = MultiheadAttention(
+            d_model, nhead, dropout=attention_dropout_rate)
+
+        self.conv1 = Conv2D(
+            in_channels=d_model,
+            out_channels=dim_feedforward,
+            kernel_size=(1, 1))
+        self.conv2 = Conv2D(
+            in_channels=dim_feedforward,
+            out_channels=d_model,
+            kernel_size=(1, 1))
+
+        self.norm1 = LayerNorm(d_model)
+        self.norm2 = LayerNorm(d_model)
+        self.dropout1 = Dropout(residual_dropout_rate)
+        self.dropout2 = Dropout(residual_dropout_rate)
+
+    def forward(self, src, src_mask=None, src_key_padding_mask=None):
+        """Pass the input through the endocder layer.
+        Args:
+            src: the sequnce to the encoder layer (required).
+            src_mask: the mask for the src sequence (optional).
+            src_key_padding_mask: the mask for the src keys per batch (optional).
+        """
+        src2 = self.self_attn(
+            src,
+            src,
+            src,
+            attn_mask=src_mask,
+            key_padding_mask=src_key_padding_mask)[0]
+        src = src + self.dropout1(src2)
+        src = self.norm1(src)
+
+        src = src.transpose([1, 2, 0])
+        src = paddle.unsqueeze(src, 2)
+        src2 = self.conv2(F.relu(self.conv1(src)))
+        src2 = paddle.squeeze(src2, 2)
+        src2 = src2.transpose([2, 0, 1])
+        src = paddle.squeeze(src, 2)
+        src = src.transpose([2, 0, 1])
+
+        src = src + self.dropout2(src2)
+        src = self.norm2(src)
+        return src
+
+
+class TransformerDecoderLayer(nn.Layer):
+    """TransformerDecoderLayer is made up of self-attn, multi-head-attn and feedforward network.
+    This standard decoder layer is based on the paper "Attention Is All You Need".
+    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez,
+    Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in
+    Neural Information Processing Systems, pages 6000-6010. Users may modify or implement
+    in a different way during application.
+
+    Args:
+        d_model: the number of expected features in the input (required).
+        nhead: the number of heads in the multiheadattention models (required).
+        dim_feedforward: the dimension of the feedforward network model (default=2048).
+        dropout: the dropout value (default=0.1).
+
+    """
+
+    def __init__(self,
+                 d_model,
+                 nhead,
+                 dim_feedforward=2048,
+                 attention_dropout_rate=0.0,
+                 residual_dropout_rate=0.1):
+        super(TransformerDecoderLayer, self).__init__()
+        self.self_attn = MultiheadAttention(
+            d_model, nhead, dropout=attention_dropout_rate)
+        self.multihead_attn = MultiheadAttention(
+            d_model, nhead, dropout=attention_dropout_rate)
+
+        self.conv1 = Conv2D(
+            in_channels=d_model,
+            out_channels=dim_feedforward,
+            kernel_size=(1, 1))
+        self.conv2 = Conv2D(
+            in_channels=dim_feedforward,
+            out_channels=d_model,
+            kernel_size=(1, 1))
+
+        self.norm1 = LayerNorm(d_model)
+        self.norm2 = LayerNorm(d_model)
+        self.norm3 = LayerNorm(d_model)
+        self.dropout1 = Dropout(residual_dropout_rate)
+        self.dropout2 = Dropout(residual_dropout_rate)
+        self.dropout3 = Dropout(residual_dropout_rate)
+
+    def forward(self,
+                tgt,
+                memory,
+                tgt_mask=None,
+                memory_mask=None,
+                tgt_key_padding_mask=None,
+                memory_key_padding_mask=None):
+        """Pass the inputs (and mask) through the decoder layer.
+
+        Args:
+            tgt: the sequence to the decoder layer (required).
+            memory: the sequnce from the last layer of the encoder (required).
+            tgt_mask: the mask for the tgt sequence (optional).
+            memory_mask: the mask for the memory sequence (optional).
+            tgt_key_padding_mask: the mask for the tgt keys per batch (optional).
+            memory_key_padding_mask: the mask for the memory keys per batch (optional).
+
+        """
+        tgt2 = self.self_attn(
+            tgt,
+            tgt,
+            tgt,
+            attn_mask=tgt_mask,
+            key_padding_mask=tgt_key_padding_mask)[0]
+        tgt = tgt + self.dropout1(tgt2)
+        tgt = self.norm1(tgt)
+        tgt2 = self.multihead_attn(
+            tgt,
+            memory,
+            memory,
+            attn_mask=memory_mask,
+            key_padding_mask=memory_key_padding_mask)[0]
+        tgt = tgt + self.dropout2(tgt2)
+        tgt = self.norm2(tgt)
+
+        # default
+        tgt = tgt.transpose([1, 2, 0])
+        tgt = paddle.unsqueeze(tgt, 2)
+        tgt2 = self.conv2(F.relu(self.conv1(tgt)))
+        tgt2 = paddle.squeeze(tgt2, 2)
+        tgt2 = tgt2.transpose([2, 0, 1])
+        tgt = paddle.squeeze(tgt, 2)
+        tgt = tgt.transpose([2, 0, 1])
+
+        tgt = tgt + self.dropout3(tgt2)
+        tgt = self.norm3(tgt)
+        return tgt
+
+
+def _get_clones(module, N):
+    return LayerList([copy.deepcopy(module) for i in range(N)])
+
+
+class PositionalEncoding(nn.Layer):
+    """Inject some information about the relative or absolute position of the tokens
+        in the sequence. The positional encodings have the same dimension as
+        the embeddings, so that the two can be summed. Here, we use sine and cosine
+        functions of different frequencies.
+    .. math::
+        \text{PosEncoder}(pos, 2i) = sin(pos/10000^(2i/d_model))
+        \text{PosEncoder}(pos, 2i+1) = cos(pos/10000^(2i/d_model))
+        \text{where pos is the word position and i is the embed idx)
+    Args:
+        d_model: the embed dim (required).
+        dropout: the dropout value (default=0.1).
+        max_len: the max. length of the incoming sequence (default=5000).
+    Examples:
+        >>> pos_encoder = PositionalEncoding(d_model)
+    """
+
+    def __init__(self, dropout, dim, max_len=5000):
+        super(PositionalEncoding, self).__init__()
+        self.dropout = nn.Dropout(p=dropout)
+
+        pe = paddle.zeros([max_len, dim])
+        position = paddle.arange(0, max_len, dtype=paddle.float32).unsqueeze(1)
+        div_term = paddle.exp(
+            paddle.arange(0, dim, 2).astype('float32') *
+            (-math.log(10000.0) / dim))
+        pe[:, 0::2] = paddle.sin(position * div_term)
+        pe[:, 1::2] = paddle.cos(position * div_term)
+        pe = pe.unsqueeze(0)
+        pe = pe.transpose([1, 0, 2])
+        self.register_buffer('pe', pe)
+
+    def forward(self, x):
+        """Inputs of forward function
+        Args:
+            x: the sequence fed to the positional encoder model (required).
+        Shape:
+            x: [sequence length, batch size, embed dim]
+            output: [sequence length, batch size, embed dim]
+        Examples:
+            >>> output = pos_encoder(x)
+        """
+        x = x + self.pe[:x.shape[0], :]
+        return self.dropout(x)
+
+
+class PositionalEncoding_2d(nn.Layer):
+    """Inject some information about the relative or absolute position of the tokens
+        in the sequence. The positional encodings have the same dimension as
+        the embeddings, so that the two can be summed. Here, we use sine and cosine
+        functions of different frequencies.
+    .. math::
+        \text{PosEncoder}(pos, 2i) = sin(pos/10000^(2i/d_model))
+        \text{PosEncoder}(pos, 2i+1) = cos(pos/10000^(2i/d_model))
+        \text{where pos is the word position and i is the embed idx)
+    Args:
+        d_model: the embed dim (required).
+        dropout: the dropout value (default=0.1).
+        max_len: the max. length of the incoming sequence (default=5000).
+    Examples:
+        >>> pos_encoder = PositionalEncoding(d_model)
+    """
+
+    def __init__(self, dropout, dim, max_len=5000):
+        super(PositionalEncoding_2d, self).__init__()
+        self.dropout = nn.Dropout(p=dropout)
+
+        pe = paddle.zeros([max_len, dim])
+        position = paddle.arange(0, max_len, dtype=paddle.float32).unsqueeze(1)
+        div_term = paddle.exp(
+            paddle.arange(0, dim, 2).astype('float32') *
+            (-math.log(10000.0) / dim))
+        pe[:, 0::2] = paddle.sin(position * div_term)
+        pe[:, 1::2] = paddle.cos(position * div_term)
+        pe = pe.unsqueeze(0).transpose([1, 0, 2])
+        self.register_buffer('pe', pe)
+
+        self.avg_pool_1 = nn.AdaptiveAvgPool2D((1, 1))
+        self.linear1 = nn.Linear(dim, dim)
+        self.linear1.weight.data.fill_(1.)
+        self.avg_pool_2 = nn.AdaptiveAvgPool2D((1, 1))
+        self.linear2 = nn.Linear(dim, dim)
+        self.linear2.weight.data.fill_(1.)
+
+    def forward(self, x):
+        """Inputs of forward function
+        Args:
+            x: the sequence fed to the positional encoder model (required).
+        Shape:
+            x: [sequence length, batch size, embed dim]
+            output: [sequence length, batch size, embed dim]
+        Examples:
+            >>> output = pos_encoder(x)
+        """
+        w_pe = self.pe[:x.shape[-1], :]
+        w1 = self.linear1(self.avg_pool_1(x).squeeze()).unsqueeze(0)
+        w_pe = w_pe * w1
+        w_pe = w_pe.transpose([1, 2, 0])
+        w_pe = w_pe.unsqueeze(2)
+
+        h_pe = self.pe[:x.shape[-2], :]
+        w2 = self.linear2(self.avg_pool_2(x).squeeze()).unsqueeze(0)
+        h_pe = h_pe * w2
+        h_pe = h_pe.transpose([1, 2, 0])
+        h_pe = h_pe.unsqueeze(3)
+
+        x = x + w_pe + h_pe
+        x = x.reshape(
+            [x.shape[0], x.shape[1], x.shape[2] * x.shape[3]]).transpose(
+                [2, 0, 1])
+
+        return self.dropout(x)
+
+
+class Embeddings(nn.Layer):
+    def __init__(self, d_model, vocab, padding_idx, scale_embedding):
+        super(Embeddings, self).__init__()
+        self.embedding = nn.Embedding(vocab, d_model, padding_idx=padding_idx)
+        w0 = np.random.normal(0.0, d_model**-0.5,
+                              (vocab, d_model)).astype(np.float32)
+        self.embedding.weight.set_value(w0)
+        self.d_model = d_model
+        self.scale_embedding = scale_embedding
+
+    def forward(self, x):
+        if self.scale_embedding:
+            x = self.embedding(x)
+            return x * math.sqrt(self.d_model)
+        return self.embedding(x)
+
+
+class Beam():
+    ''' Beam search '''
+
+    def __init__(self, size, device=False):
+
+        self.size = size
+        self._done = False
+        # The score for each translation on the beam.
+        self.scores = paddle.zeros((size, ), dtype=paddle.float32)
+        self.all_scores = []
+        # The backpointers at each time-step.
+        self.prev_ks = []
+        # The outputs at each time-step.
+        self.next_ys = [paddle.full((size, ), 0, dtype=paddle.int64)]
+        self.next_ys[0][0] = 2
+
+    def get_current_state(self):
+        "Get the outputs for the current timestep."
+        return self.get_tentative_hypothesis()
+
+    def get_current_origin(self):
+        "Get the backpointers for the current timestep."
+        return self.prev_ks[-1]
+
+    @property
+    def done(self):
+        return self._done
+
+    def advance(self, word_prob):
+        "Update beam status and check if finished or not."
+        num_words = word_prob.shape[1]
+
+        # Sum the previous scores.
+        if len(self.prev_ks) > 0:
+            beam_lk = word_prob + self.scores.unsqueeze(1).expand_as(word_prob)
+        else:
+            beam_lk = word_prob[0]
+
+        flat_beam_lk = beam_lk.reshape([-1])
+        best_scores, best_scores_id = flat_beam_lk.topk(self.size, 0, True,
+                                                        True)  # 1st sort
+        self.all_scores.append(self.scores)
+        self.scores = best_scores
+        # bestScoresId is flattened as a (beam x word) array,
+        # so we need to calculate which word and beam each score came from
+        prev_k = best_scores_id // num_words
+        self.prev_ks.append(prev_k)
+        self.next_ys.append(best_scores_id - prev_k * num_words)
+        # End condition is when top-of-beam is EOS.
+        if self.next_ys[-1][0] == 3:
+            self._done = True
+            self.all_scores.append(self.scores)
+
+        return self._done
+
+    def sort_scores(self):
+        "Sort the scores."
+        return self.scores, paddle.to_tensor(
+            [i for i in range(self.scores.shape[0])], dtype='int32')
+
+    def get_the_best_score_and_idx(self):
+        "Get the score of the best in the beam."
+        scores, ids = self.sort_scores()
+        return scores[1], ids[1]
+
+    def get_tentative_hypothesis(self):
+        "Get the decoded sequence for the current timestep."
+        if len(self.next_ys) == 1:
+            dec_seq = self.next_ys[0].unsqueeze(1)
+        else:
+            _, keys = self.sort_scores()
+            hyps = [self.get_hypothesis(k) for k in keys]
+            hyps = [[2] + h for h in hyps]
+            dec_seq = paddle.to_tensor(hyps, dtype='int64')
+        return dec_seq
+
+    def get_hypothesis(self, k):
+        """ Walk back to construct the full hypothesis. """
+        hyp = []
+        for j in range(len(self.prev_ks) - 1, -1, -1):
+            hyp.append(self.next_ys[j + 1][k])
+            k = self.prev_ks[j][k]
+        return list(map(lambda x: x.item(), hyp[::-1]))

ppocr/postprocess/__init__.py

@@ -24,18 +24,16 @@ __all__ = ['build_post_process']
 from .db_postprocess import DBPostProcess, DistillationDBPostProcess
 from .east_postprocess import EASTPostProcess
 from .sast_postprocess import SASTPostProcess
-from .rec_postprocess import CTCLabelDecode, AttnLabelDecode, SRNLabelDecode, DistillationCTCLabelDecode, \
+from .rec_postprocess import CTCLabelDecode, AttnLabelDecode, SRNLabelDecode, DistillationCTCLabelDecode, NRTRLabelDecode, \
     TableLabelDecode
 from .cls_postprocess import ClsPostProcess
 from .pg_postprocess import PGPostProcess
 
-
 def build_post_process(config, global_config=None):
     support_dict = [
         'DBPostProcess', 'EASTPostProcess', 'SASTPostProcess', 'CTCLabelDecode',
         'AttnLabelDecode', 'ClsPostProcess', 'SRNLabelDecode', 'PGPostProcess',
-        'DistillationCTCLabelDecode', 'TableLabelDecode',
-        'DistillationDBPostProcess'
+        'DistillationCTCLabelDecode', 'NRTRLabelDecode', 'TableLabelDecode', 'DistillationDBPostProcess'
     ]
 
     config = copy.deepcopy(config)

ppocr/postprocess/rec_postprocess.py

@@ -156,6 +156,69 @@ class DistillationCTCLabelDecode(CTCLabelDecode):
         return output
 
 
+class NRTRLabelDecode(BaseRecLabelDecode):
+    """ Convert between text-label and text-index """
+
+    def __init__(self,
+                 character_dict_path=None,
+                 character_type='EN_symbol',
+                 use_space_char=True,
+                 **kwargs):
+        super(NRTRLabelDecode, self).__init__(character_dict_path,
+                                             character_type, use_space_char)
+
+    def __call__(self, preds, label=None, *args, **kwargs):
+        if preds.dtype == paddle.int64:
+            if isinstance(preds, paddle.Tensor):
+                preds = preds.numpy()
+            if preds[0][0]==2:
+                preds_idx = preds[:,1:]
+            else:
+                preds_idx = preds
+
+            text = self.decode(preds_idx)
+            if label is None:
+                return text
+            label = self.decode(label[:,1:])
+        else:
+            if isinstance(preds, paddle.Tensor):
+                preds = preds.numpy()
+            preds_idx = preds.argmax(axis=2)
+            preds_prob = preds.max(axis=2)
+            text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False)
+            if label is None:
+                return text
+            label = self.decode(label[:,1:])
+        return text, label
+
+    def add_special_char(self, dict_character):
+        dict_character = ['blank','<unk>','<s>','</s>'] + dict_character
+        return dict_character
+    
+    def decode(self, text_index, text_prob=None, is_remove_duplicate=False):
+        """ convert text-index into text-label. """
+        result_list = []
+        batch_size = len(text_index)
+        for batch_idx in range(batch_size):
+            char_list = []
+            conf_list = []
+            for idx in range(len(text_index[batch_idx])):
+                if text_index[batch_idx][idx] == 3: # end
+                    break
+                try:
+                    char_list.append(self.character[int(text_index[batch_idx][idx])])
+                except:
+                    continue
+                if text_prob is not None:
+                    conf_list.append(text_prob[batch_idx][idx])
+                else:
+                    conf_list.append(1)
+            text = ''.join(char_list)
+            result_list.append((text.lower(), np.mean(conf_list)))
+        return result_list
+
+
+
 class AttnLabelDecode(BaseRecLabelDecode):
     """ Convert between text-label and text-index """
 
@@ -193,8 +256,7 @@ class AttnLabelDecode(BaseRecLabelDecode):
                     if idx > 0 and text_index[batch_idx][idx - 1] == text_index[
                             batch_idx][idx]:
                         continue
-                char_list.append(self.character[int(text_index[batch_idx][
-                    idx])])
+                char_list.append(self.character[int(text_index[batch_idx][idx])])
                 if text_prob is not None:
                     conf_list.append(text_prob[batch_idx][idx])
                 else:

tools/program.py

@@ -186,9 +186,11 @@ def train(config,
     model.train()
 
     use_srn = config['Architecture']['algorithm'] == "SRN"
-    try: 
+    use_nrtr = config['Architecture']['algorithm'] == "NRTR"
+
+    try:
         model_type = config['Architecture']['model_type']
-    except: 
+    except:
         model_type = None
 
     if 'start_epoch' in best_model_dict:
@@ -213,7 +215,7 @@ def train(config,
             images = batch[0]
             if use_srn:
                 model_average = True
-            if use_srn or model_type == 'table':
+            if use_srn or model_type == 'table' or use_nrtr:
                 preds = model(images, data=batch[1:])
             else:
                 preds = model(images)
@@ -398,7 +400,7 @@ def preprocess(is_train=False):
     alg = config['Architecture']['algorithm']
     assert alg in [
         'EAST', 'DB', 'SAST', 'Rosetta', 'CRNN', 'STARNet', 'RARE', 'SRN',
-        'CLS', 'PGNet', 'Distillation', 'TableAttn'
+        'CLS', 'PGNet', 'Distillation', 'NRTR', 'TableAttn'
     ]
 
     device = 'gpu:{}'.format(dist.ParallelEnv().dev_id) if use_gpu else 'cpu'