update framework of matching

fastNLP/io/dataset_loader.py

@@ -269,7 +269,7 @@ class MatchingLoader(DataSetLoader):
     def _load(self, path: str) -> DataSet:
         raise NotImplementedError
 
-    def process(self, paths: Union[str, Dict[str, str]], **options) -> DataInfo:
+    def process(self, paths: Union[str, Dict[str, str]], input_field=None) -> DataInfo:
         if isinstance(paths, str):
             paths = {'train': paths}
 
@@ -289,6 +289,13 @@ class MatchingLoader(DataSetLoader):
                 raise RuntimeError(f'Your model is {self.data_format}, '
                                    f'Please choose from [esim, bert]')
 
+            if input_field is not None:
+                if isinstance(input_field, str):
+                    data.set_input(input_field)
+                elif isinstance(input_field, list):
+                    for field in input_field:
+                        data.set_input(field)
+
             data_set[n] = data
             print(f'successfully load {n} set!')
 
@@ -298,11 +305,11 @@ class MatchingLoader(DataSetLoader):
             raise RuntimeError(f'There is NOT label vocab attribute built!')
 
         if self.for_model != 'bert':
-            from fastNLP.modules.encoder.embedding import StaticEmbedding
-            embedding = StaticEmbedding(self.vocab, model_dir_or_name='en')
+            from fastNLP.modules.encoder.embedding import ElmoEmbedding
+            embedding = ElmoEmbedding(self.vocab, model_dir_or_name='en', requires_grad=True, layers='2')
 
         data_info = DataInfo(vocabs={'vocab': self.vocab, 'target_vocab': self.label_vocab},
-                             embeddings={'glove': embedding} if self.for_model != 'bert' else None,
+                             embeddings={'elmo': embedding} if self.for_model != 'bert' else None,
                              datasets=data_set)
 
         return data_info
@@ -338,15 +345,17 @@ class MatchingLoader(DataSetLoader):
             raw_ds.drop(lambda x: x[Const.TARGET] == '-')
 
         if not hasattr(self, 'vocab'):
-            self.vocab = Vocabulary().from_dataset(raw_ds, [Const.INPUTS(0), Const.INPUTS(1)])
+            self.vocab = Vocabulary().from_dataset(raw_ds, field_name=[Const.INPUTS(0), Const.INPUTS(1)])
         if not hasattr(self, 'label_vocab'):
             self.label_vocab = Vocabulary(padding=None, unknown=None).from_dataset(raw_ds, field_name=Const.TARGET)
 
         raw_ds.apply(lambda ins: [self.vocab.to_index(w) for w in ins[Const.INPUTS(0)]], new_field_name=Const.INPUTS(0))
         raw_ds.apply(lambda ins: [self.vocab.to_index(w) for w in ins[Const.INPUTS(1)]], new_field_name=Const.INPUTS(1))
-        raw_ds.apply(lambda ins: self.label_vocab.to_index(Const.TARGET), new_field_name=Const.TARGET)
+        raw_ds.apply(lambda ins: self.label_vocab.to_index(ins[Const.TARGET]), new_field_name=Const.TARGET)
+        raw_ds.apply(lambda ins: len(ins[Const.INPUTS(0)]), new_field_name=Const.INPUT_LENS(0))
+        raw_ds.apply(lambda ins: len(ins[Const.INPUTS(1)]), new_field_name=Const.INPUT_LENS(1))
 
-        raw_ds.set_input(Const.INPUTS(0), Const.INPUTS(1))
+        raw_ds.set_input(Const.INPUTS(0), Const.INPUTS(1), Const.INPUT_LENS(0), Const.INPUT_LENS(1))
         raw_ds.set_target(Const.TARGET)
 
         return raw_ds
@@ -405,6 +414,8 @@ class MatchingLoader(DataSetLoader):
         raw_ds.set_input(Const.INPUT, Const.INPUT_LENS(0), Const.INPUT_LENS(1))
         raw_ds.set_target(Const.TARGET)
 
+        return raw_ds
+
 
 class SNLILoader(JsonLoader):
     """

fastNLP/modules/encoder/bert.py

@@ -2,9 +2,9 @@
 import os
 from torch import nn
 import torch
-from ...core import Vocabulary
+from ...core.vocabulary import Vocabulary
 from ...io.file_utils import _get_base_url, cached_path
-from ._bert import _WordPieceBertModel
+from ._bert import _WordPieceBertModel, BertModel
 
 
 class BertWordPieceEncoder(nn.Module):

fastNLP/modules/encoder/embedding.py

@@ -152,6 +152,8 @@ class StaticEmbedding(TokenEmbedding):
 
     Example::
 
+        >>> embed = StaticEmbedding(vocab, model_dir_or_name='en-glove-6b-50')
+
 
     :param vocab: Vocabulary. 若该项为None则会读取所有的embedding。
     :param model_dir_or_name: 可以有两种方式调用预训练好的static embedding：第一种是传入embedding的文件名，第二种是传入embedding
@@ -311,8 +313,7 @@ class ElmoEmbedding(ContextualEmbedding):
 
     Example::
 
-        >>>
-        >>>
+        >>> embedding = ElmoEmbedding(vocab, model_dir_or_name='en', layers='2', requires_grad=True)
 
     :param vocab: 词表
     :param model_dir_or_name: 可以有两种方式调用预训练好的ELMo embedding：第一种是传入ELMo权重的文件名，第二种是传入ELMo版本的名称，
@@ -403,7 +404,7 @@ class BertEmbedding(ContextualEmbedding):
 
     Example::
 
-        >>>
+        >>> embedding = BertEmbedding(vocab, model_dir_or_name='en-base-uncased', requires_grad=False, layers='4,-2,-1')
 
 
     :param fastNLP.Vocabulary vocab: 词表
@@ -513,7 +514,7 @@ class CNNCharEmbedding(TokenEmbedding):
 
     Example::
 
-        >>>
+        >>> cnn_char_embed = CNNCharEmbedding(vocab)
 
 
     :param vocab: 词表
@@ -647,7 +648,7 @@ class LSTMCharEmbedding(TokenEmbedding):
 
     Example::
 
-        >>>
+        >>> lstm_char_embed = LSTMCharEmbedding(vocab)
 
     :param vocab: 词表
     :param embed_size: embedding的大小。默认值为50.

reproduction/matching/matching.py

@@ -2,31 +2,31 @@ import os
 
 import torch
 
-from fastNLP.core import Trainer, Tester, Adam, AccuracyMetric
+from fastNLP.core import Trainer, Tester, Adam, AccuracyMetric, Const
 
 from fastNLP.io.dataset_loader import MatchingLoader
 
 from reproduction.matching.model.bert import BertForNLI
+from reproduction.matching.model.esim import ESIMModel
 
 
-# bert_dirs = 'path/to/bert/dir'
-bert_dirs = '/remote-home/ygxu/BERT/BERT_English_uncased_L-12_H-768_A_12'
+bert_dirs = 'path/to/bert/dir'
 
 # load data set
-data_info = MatchingLoader(data_format='snli', for_model='bert', bert_dir=bert_dirs).process(
-    {#'train': './data/snli/snli_1.0_train.jsonl',
+# data_info = MatchingLoader(data_format='snli', for_model='bert', bert_dir=bert_dirs).process(...
+data_info = MatchingLoader(data_format='snli', for_model='esim').process(
+    {'train': './data/snli/snli_1.0_train.jsonl',
      'dev': './data/snli/snli_1.0_dev.jsonl',
-     'test': './data/snli/snli_1.0_test.jsonl'}
+     'test': './data/snli/snli_1.0_test.jsonl'},
+    input_field=[Const.TARGET]
 )
 
-print('successfully load data sets!')
+# model = BertForNLI(bert_dir=bert_dirs)
+model = ESIMModel(data_info.embeddings['elmo'],)
 
-
-model = BertForNLI(bert_dir=bert_dirs)
-
-trainer = Trainer(train_data=data_info.datasets['dev'], model=model,
-                  optimizer=Adam(lr=2e-5, model_params=model.parameters()),
-                  batch_size=torch.cuda.device_count() * 12, n_epochs=4, print_every=-1,
+trainer = Trainer(train_data=data_info.datasets['train'], model=model,
+                  optimizer=Adam(lr=1e-4, model_params=model.parameters()),
+                  batch_size=torch.cuda.device_count() * 24, n_epochs=20, print_every=-1,
                   dev_data=data_info.datasets['dev'],
                   metrics=AccuracyMetric(), metric_key='acc', device=[i for i in range(torch.cuda.device_count())],
                   check_code_level=-1)

reproduction/matching/model/esim.py

@@ -0,0 +1,182 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from torch.nn import CrossEntropyLoss
+
+from fastNLP.models import BaseModel
+from fastNLP.modules.encoder.embedding import TokenEmbedding
+from fastNLP.modules.encoder.lstm import LSTM
+from fastNLP.core.const import Const
+from fastNLP.core.utils import seq_len_to_mask
+
+
+class ESIMModel(BaseModel):
+    def __init__(self, init_embedding: TokenEmbedding, hidden_size=None, num_labels=3, dropout_rate=0.3,
+                 dropout_embed=0.1):
+        super(ESIMModel, self).__init__()
+
+        self.embedding = init_embedding
+        self.dropout_embed = EmbedDropout(p=dropout_embed)
+        if hidden_size is None:
+            hidden_size = self.embedding.embed_size
+        self.rnn = BiRNN(self.embedding.embed_size, hidden_size, dropout_rate=dropout_rate)
+        # self.rnn = LSTM(self.embedding.embed_size, hidden_size, dropout=dropout_rate, bidirectional=True)
+
+        self.interfere = nn.Sequential(nn.Dropout(p=dropout_rate),
+                                       nn.Linear(8 * hidden_size, hidden_size),
+                                       nn.ReLU())
+        nn.init.xavier_uniform_(self.interfere[1].weight.data)
+        self.bi_attention = SoftmaxAttention()
+
+        self.rnn_high = BiRNN(self.embedding.embed_size, hidden_size, dropout_rate=dropout_rate)
+        # self.rnn_high = LSTM(hidden_size, hidden_size, dropout=dropout_rate, bidirectional=True)
+
+        self.classifier = nn.Sequential(nn.Dropout(p=dropout_rate),
+                                        nn.Linear(8 * hidden_size, hidden_size),
+                                        nn.Tanh(),
+                                        nn.Dropout(p=dropout_rate),
+                                        nn.Linear(hidden_size, num_labels))
+        nn.init.xavier_uniform_(self.classifier[1].weight.data)
+        nn.init.xavier_uniform_(self.classifier[4].weight.data)
+
+    def forward(self, words1, words2, seq_len1, seq_len2, target=None):
+        mask1 = seq_len_to_mask(seq_len1)
+        mask2 = seq_len_to_mask(seq_len2)
+        a0 = self.embedding(words1)  # B * len * emb_dim
+        b0 = self.embedding(words2)
+        a0, b0 = self.dropout_embed(a0), self.dropout_embed(b0)
+        a = self.rnn(a0, mask1.byte())  # a: [B, PL, 2 * H]
+        b = self.rnn(b0, mask2.byte())
+
+        ai, bi = self.bi_attention(a, mask1, b, mask2)
+
+        a_ = torch.cat((a, ai, a - ai, a * ai), dim=2)  # ma: [B, PL, 8 * H]
+        b_ = torch.cat((b, bi, b - bi, b * bi), dim=2)
+        a_f = self.interfere(a_)
+        b_f = self.interfere(b_)
+
+        a_h = self.rnn_high(a_f, mask1.byte())  # ma: [B, PL, 2 * H]
+        b_h = self.rnn_high(b_f, mask2.byte())
+
+        a_avg = self.mean_pooling(a_h, mask1, dim=1)
+        a_max, _ = self.max_pooling(a_h, mask1, dim=1)
+        b_avg = self.mean_pooling(b_h, mask2, dim=1)
+        b_max, _ = self.max_pooling(b_h, mask2, dim=1)
+
+        out = torch.cat((a_avg, a_max, b_avg, b_max), dim=1)  # v: [B, 8 * H]
+        logits = torch.tanh(self.classifier(out))
+
+        if target is not None:
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(logits, target)
+
+            return {Const.LOSS: loss, Const.OUTPUT: logits}
+        else:
+            return {Const.OUTPUT: logits}
+
+    def predict(self, **kwargs):
+        return self.forward(**kwargs)
+
+    # input [batch_size, len , hidden]
+    # mask  [batch_size, len] (111...00)
+    @staticmethod
+    def mean_pooling(input, mask, dim=1):
+        masks = mask.view(mask.size(0), mask.size(1), -1).float()
+        return torch.sum(input * masks, dim=dim) / torch.sum(masks, dim=1)
+
+    @staticmethod
+    def max_pooling(input, mask, dim=1):
+        my_inf = 10e12
+        masks = mask.view(mask.size(0), mask.size(1), -1)
+        masks = masks.expand(-1, -1, input.size(2)).float()
+        return torch.max(input + masks.le(0.5).float() * -my_inf, dim=dim)
+
+
+class EmbedDropout(nn.Dropout):
+
+    def forward(self, sequences_batch):
+        ones = sequences_batch.data.new_ones(sequences_batch.shape[0], sequences_batch.shape[-1])
+        dropout_mask = nn.functional.dropout(ones, self.p, self.training, inplace=False)
+        return dropout_mask.unsqueeze(1) * sequences_batch
+
+
+class BiRNN(nn.Module):
+    def __init__(self, input_size, hidden_size, dropout_rate=0.3):
+        super(BiRNN, self).__init__()
+        self.dropout_rate = dropout_rate
+        self.rnn = nn.LSTM(input_size, hidden_size,
+                           num_layers=1,
+                           bidirectional=True,
+                           batch_first=True)
+
+    def forward(self, x, x_mask):
+        # Sort x
+        lengths = x_mask.data.eq(1).long().sum(1).squeeze()
+        _, idx_sort = torch.sort(lengths, dim=0, descending=True)
+        _, idx_unsort = torch.sort(idx_sort, dim=0)
+        lengths = list(lengths[idx_sort])
+
+        x = x.index_select(0, idx_sort)
+        # Pack it up
+        rnn_input = nn.utils.rnn.pack_padded_sequence(x, lengths, batch_first=True)
+        # Apply dropout to input
+        if self.dropout_rate > 0:
+            dropout_input = F.dropout(rnn_input.data, p=self.dropout_rate, training=self.training)
+            rnn_input = nn.utils.rnn.PackedSequence(dropout_input, rnn_input.batch_sizes)
+        output = self.rnn(rnn_input)[0]
+        # Unpack everything
+        output = nn.utils.rnn.pad_packed_sequence(output, batch_first=True)[0]
+        output = output.index_select(0, idx_unsort)
+        if output.size(1) != x_mask.size(1):
+            padding = torch.zeros(output.size(0),
+                                  x_mask.size(1) - output.size(1),
+                                  output.size(2)).type(output.data.type())
+            output = torch.cat([output, padding], 1)
+        return output
+
+
+def masked_softmax(tensor, mask):
+    tensor_shape = tensor.size()
+    reshaped_tensor = tensor.view(-1, tensor_shape[-1])
+
+    # Reshape the mask so it matches the size of the input tensor.
+    while mask.dim() < tensor.dim():
+        mask = mask.unsqueeze(1)
+    mask = mask.expand_as(tensor).contiguous().float()
+    reshaped_mask = mask.view(-1, mask.size()[-1])
+    result = F.softmax(reshaped_tensor * reshaped_mask, dim=-1)
+    result = result * reshaped_mask
+    # 1e-13 is added to avoid divisions by zero.
+    result = result / (result.sum(dim=-1, keepdim=True) + 1e-13)
+    return result.view(*tensor_shape)
+
+
+def weighted_sum(tensor, weights, mask):
+    w_sum = weights.bmm(tensor)
+    while mask.dim() < w_sum.dim():
+        mask = mask.unsqueeze(1)
+    mask = mask.transpose(-1, -2)
+    mask = mask.expand_as(w_sum).contiguous().float()
+    return w_sum * mask
+
+
+class SoftmaxAttention(nn.Module):
+
+    def forward(self, premise_batch, premise_mask, hypothesis_batch, hypothesis_mask):
+        similarity_matrix = premise_batch.bmm(hypothesis_batch.transpose(2, 1)
+                                              .contiguous())
+
+        prem_hyp_attn = masked_softmax(similarity_matrix, hypothesis_mask)
+        hyp_prem_attn = masked_softmax(similarity_matrix.transpose(1, 2)
+                                       .contiguous(),
+                                       premise_mask)
+
+        attended_premises = weighted_sum(hypothesis_batch,
+                                         prem_hyp_attn,
+                                         premise_mask)
+        attended_hypotheses = weighted_sum(premise_batch,
+                                           hyp_prem_attn,
+                                           hypothesis_mask)
+
+        return attended_premises, attended_hypotheses