Merge pull request #11 from zalandoresearch/GH-10-text-classification

GH-10: Added text classification model and trainer

.gitignore

@@ -0,0 +1,104 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/

flair/embeddings.py

@@ -1,16 +1,16 @@
+import os
 import pickle
 import re
-import os
-from abc import ABC, abstractmethod
-from typing import List, Dict, Tuple
+from abc import abstractmethod
+from typing import List
 
 import gensim
 import numpy as np
 import torch
 
-from .file_utils import cached_path
-from .language_model import RNNModel
+from flair.models.language_model import RNNModel
 from .data import Dictionary, Token, Sentence, TaggedCorpus
+from .file_utils import cached_path
 
 
 class TextEmbeddings(torch.nn.Module):
@@ -50,7 +50,7 @@ def embed(self, sentences: List[Sentence]) -> List[Sentence]:
         return sentences
 
     @abstractmethod
-    def _add_embeddings_internal(self, sentences: List[Sentence]):
+    def _add_embeddings_internal(self, sentences: List[Sentence]) -> List[Sentence]:
         """Private method for adding embeddings to all words in a list of sentences."""
         pass
 
@@ -84,14 +84,14 @@ def embed(self, sentences: List[Sentence], static_embeddings: bool = True):
             embedding.embed(sentences)
 
     @property
-    def embedding_type(self):
+    def embedding_type(self) -> str:
         return self.__embedding_type
 
     @property
     def embedding_length(self) -> int:
         return self.__embedding_length
 
-    def _add_embeddings_internal(self, sentences: List[Sentence]):
+    def _add_embeddings_internal(self, sentences: List[Sentence]) -> List[Sentence]:
 
         for embedding in self.embeddings:
             embedding._add_embeddings_internal(sentences)
@@ -413,6 +413,7 @@ def _add_embeddings_internal(self, sentences: List[Sentence]) -> List[Sentence]:
 
 
 class OnePassStoreEmbeddings(TextEmbeddings):
+
     def __init__(self, embedding_stack: StackedEmbeddings, corpus: TaggedCorpus, detach: bool = True):
         super().__init__()
 
@@ -564,14 +565,22 @@ def _add_embeddings_internal(self, sentences: List[Sentence]):
 class TextLSTMEmbedder(TextEmbeddings):
 
     def __init__(self, word_embeddings: List[TextEmbeddings], hidden_states=128, num_layers=1,
-                 reproject_words: bool = True):
-        """The constructor takes a list of embeddings to be combined."""
+                 reproject_words: bool = True, bidirectional: bool = True):
+        """The constructor takes a list of embeddings to be combined.
+        :param word_embeddings: a list of word embeddings
+        :param hidden_states: the number of hidden states in the lstm
+        :param num_layers: the number of layers for the lstm
+        :param reproject_words: boolean value, indicating whether to reproject the word embedding in a separate linear
+        layer before putting them into the lstm or not
+        :param bidirectional: boolean value, indicating whether to use a bidirectional lstm or not
+        """
         super().__init__()
 
         # self.embeddings: StackedEmbeddings = StackedEmbeddings(embeddings=word_embeddings)
         self.embeddings: List[TextEmbeddings] = word_embeddings
 
         self.reproject_words = reproject_words
+        self.bidirectional = bidirectional
 
         self.length_of_all_word_embeddings = 0
         for word_embedding in self.embeddings:
@@ -580,14 +589,16 @@ def __init__(self, word_embeddings: List[TextEmbeddings], hidden_states=128, num
         self.name = 'text_lstm'
         self.static_embeddings = False
 
-        # self.__embedding_length: int = hidden_states
-        self.__embedding_length: int = hidden_states * 2
+        if self.bidirectional:
+            self.__embedding_length: int = hidden_states * 2
+        else:
+            self.__embedding_length: int = hidden_states
 
         # bidirectional LSTM on top of embedding layer
         self.word_reprojection_map = torch.nn.Linear(self.length_of_all_word_embeddings,
                                                      self.length_of_all_word_embeddings)
         self.rnn = torch.nn.LSTM(self.length_of_all_word_embeddings, hidden_states, num_layers=num_layers,
-                                 bidirectional=True)
+                                 bidirectional=self.bidirectional)
         self.dropout = torch.nn.Dropout(0.5)
 
     @property

flair/language_model.py → flair/models/language_model.py

@@ -3,7 +3,7 @@
 import math
 from torch.autograd import Variable
 from typing import Dict, List
-from .data import Dictionary
+from flair.data import Dictionary
 
 
 class RNNModel(nn.Module):

flair/tagging_model.py → flair/models/tagging_model.py

@@ -7,8 +7,8 @@
 import numpy as np
 
 from flair.file_utils import cached_path
-from .data import Dictionary, Sentence, Token
-from .embeddings import TextEmbeddings
+from flair.data import Dictionary, Sentence, Token
+from flair.embeddings import TextEmbeddings
 
 from typing import List, Tuple, Union