add kuleshov recipe; rename attack_methods --> search_methods

textattack/__init__.py

@@ -2,13 +2,13 @@ name = "textattack"
 
 from . import attack_recipes
 from . import attack_results
-from . import attack_methods
 from . import constraints
 from . import datasets
 from . import goal_functions
 from . import goal_function_results
 from . import loggers
 from . import models
+from . import search_methods
 from . import shared 
 from . import tokenizers
 from . import transformations

textattack/attack_recipes/__init__.py

@@ -2,6 +2,7 @@ from .alzantot_2018 import Alzantot2018
 from .alzantot_2018_adjusted import Alzantot2018Adjusted
 from .deepwordbug_gao_2018 import DeepWordBugGao2018
 from .hotflip_ebrahimi_2017 import HotFlipEbrahimi2017
+from .kuleshov_2017 import Kuleshov2017
 from .seq2sick_cheng_2018_blackbox import Seq2SickCheng2018BlackBox
 from .textfooler_jin_2019 import TextFoolerJin2019
 from .textfooler_jin_2019_adjusted import TextFoolerJin2019Adjusted

textattack/attack_recipes/alzantot_2018.py

@@ -30,7 +30,7 @@ def Alzantot2018(model):
     # Maximum words perturbed percentage of 20%
     #
     constraints.append(
-            WordsPerturbed(max_percent=20)
+            WordsPerturbed(max_percent=0.2)
     )
     #
     # Maximum word embedding euclidean distance of 0.5.

textattack/attack_recipes/deepwordbug_gao_2018.py

@@ -8,9 +8,9 @@
     
 """
 
-from textattack.attack_methods import GreedyWordSwapWIR
 from textattack.constraints.overlap import LevenshteinEditDistance
 from textattack.goal_functions import UntargetedClassification
+from textattack.search_methods import GreedyWordSwapWIR
 from textattack.transformations import CompositeTransformation
 from textattack.transformations.black_box import \
     WordSwapNeighboringCharacterSwap, \

textattack/attack_recipes/hotflip_ebrahimi_2017.py

@@ -11,12 +11,12 @@
     paper).
 """
 
-from textattack.attack_methods import BeamSearch
+from textattack.goal_functions import UntargetedClassification
 from textattack.constraints.grammaticality import PartOfSpeech
 from textattack.constraints.overlap import WordsPerturbed
 from textattack.constraints.semantics import WordEmbeddingDistance
+from textattack.search_methods import BeamSearch
 from textattack.transformations.white_box import GradientBasedWordSwap
-from textattack.goal_functions import UntargetedClassification
 
 def HotFlipEbrahimi2017(model):
     #

textattack/attack_recipes/kuleshov_2017.py

@@ -8,8 +8,8 @@
 """
 
 from textattack.constraints.overlap import WordsPerturbed
-from textattack.constraints.grammaticality.language_models import Google1BillionWordsLanguageModel
+from textattack.constraints.grammaticality.language_models import GPT2
-from textattack.constraints.semantics import WordEmbeddingDistance
+from textattack.constraints.semantics import ThoughtVector
 from textattack.goal_functions import UntargetedClassification
 from textattack.search_methods import GreedyWordSwap
 from textattack.transformations.black_box import WordSwapEmbedding

textattack/attack_recipes/seq2sick_cheng_2018_blackbox.py

@@ -12,9 +12,9 @@
     
 """
 
-from textattack.attack_methods import GreedyWordSwapWIR
 from textattack.constraints.overlap import LevenshteinEditDistance
 from textattack.goal_functions import NonOverlappingOutput
+from textattack.search_methods import GreedyWordSwapWIR
 from textattack.transformations.black_box import WordSwapEmbedding
 
 def Seq2SickCheng2018BlackBox(model, goal_function='non_overlapping'):

textattack/attack_recipes/textfooler_jin_2019.py

@@ -8,12 +8,12 @@
     
 """
 
-from textattack.attack_methods import GreedyWordSwapWIR
+from textattack.goal_functions import UntargetedClassification
 from textattack.constraints.grammaticality import PartOfSpeech
 from textattack.constraints.semantics import WordEmbeddingDistance
 from textattack.constraints.semantics.sentence_encoders import UniversalSentenceEncoder
+from textattack.search_methods import GreedyWordSwapWIR
 from textattack.transformations.black_box import WordSwapEmbedding
-from textattack.goal_functions import UntargetedClassification
 
 def TextFoolerJin2019(model):
     #

textattack/attack_recipes/textfooler_jin_2019_adjusted.py

@@ -8,12 +8,12 @@
     
 """
 
-from textattack.attack_methods import GreedyWordSwapWIR
 from textattack.constraints.semantics import WordEmbeddingDistance
 from textattack.constraints.semantics.sentence_encoders import UniversalSentenceEncoder, BERT
 from textattack.constraints.grammaticality import PartOfSpeech, LanguageTool
-from textattack.transformations.black_box import WordSwapEmbedding
 from textattack.goal_functions import UntargetedClassification
+from textattack.search_methods import GreedyWordSwapWIR
+from textattack.transformations.black_box import WordSwapEmbedding
 
 def TextFoolerJin2019Adjusted(model, SE_thresh=0.98, sentence_encoder='bert'):
     #

textattack/constraints/grammaticality/language_models/language_model_constraint.py

@@ -34,7 +34,7 @@ class LanguageModelConstraint(Constraint):
         x_adv_prob = self.get_log_prob_at_index(x_adv, i)
         if self.max_log_prob_diff is None:
             x_prob, x_adv_prob = math.log(p1), math.log(p2)
-        return (x_prob - x_adv_prob).abs() <= self.max_log_prob_diff
+        return abs(x_prob - x_adv_prob) <= self.max_log_prob_diff
     
     def extra_repr_keys(self):
         return ['max_log_prob_diff']

textattack/constraints/overlap/words_perturbed.py

@@ -1,3 +1,4 @@
+import math
 from textattack.constraints import Constraint
 
 class WordsPerturbed(Constraint):
@@ -6,6 +7,8 @@ class WordsPerturbed(Constraint):
     def __init__(self, max_num_words=None, max_percent=None):
         if (max_num_words is None) and (max_percent is None):
             raise ValueError('must set either max perc or max num words')
+        if max_percent and not (0 <= max_percent <= 1):
+            raise ValueError('max perc must be between 0 and 1')
         self.max_num_words = max_num_words
         self.max_percent = max_percent
     
@@ -16,7 +19,7 @@ class WordsPerturbed(Constraint):
         num_words_diff = len(x_adv.all_words_diff(original_text))
         if self.max_percent:
             min_num_words = min(len(x_adv.words), len(original_text.words))
-            max_words_perturbed = round(min_num_words * (self.max_percent / 100))
+            max_words_perturbed = math.ceil(min_num_words * (self.max_percent))
             max_percent_met = num_words_diff <= max_words_perturbed
         else:
             max_percent_met = True

textattack/constraints/semantics/thought_vector.py

@@ -17,6 +17,7 @@ class ThoughtVector(Constraint):
     """
     def __init__(self, embedding_type='paragramcf', max_mse_dist=None, min_cos_sim=None):
         self.word_embedding = WordEmbedding(embedding_type)
+        self.embedding_type = embedding_type
         
         if (max_mse_dist or min_cos_sim) is None:
             raise ValueError('Must set max_mse_dist or min_cos_sim')
@@ -26,9 +27,18 @@ class ThoughtVector(Constraint):
     
     @functools.lru_cache(maxsize=2**10)
     def _get_thought_vector(self, tokenized_text):
-        return torch.sum([self.word_embedding[word] for word in tokenized_text.words])
+        """ Sums the embeddings of all the words in `tokenized_text` into a
+            "thought vector".
+        """
+        embeddings = []
+        for word in tokenized_text.words:
+            embedding = self.word_embedding[word]
+            if embedding is not None: # out-of-vocab words do not have embeddings
+                embeddings.append(embedding)
+        embeddings = torch.tensor(embeddings)
+        return torch.sum(embeddings, dim=0)
     
-    def __call__(self, x, x_adv):
+    def __call__(self, x, x_adv, original_text=None):
         """ Returns true if (x, x_adv) are closer than `self.min_cos_sim`
             and `self.max_mse_dist`. """
         
@@ -47,7 +57,7 @@ class ThoughtVector(Constraint):
                 return False
         # Check MSE distance.
         if self.max_mse_dist:
-            mse_dist = torch.sum((e1 - e2) ** 2)
+            mse_dist = torch.sum((thought_vector_1 - thought_vector_2) ** 2)
             if mse_dist > self.max_mse_dist:
                 return False
         return True

textattack/goal_functions/classification/untargeted_classification.py

@@ -10,8 +10,9 @@ class UntargetedClassification(ClassificationGoalFunction):
             below this score. Otherwise, goal is to change the overall predicted
             class.
     """
-    def __init__(self, target_max_score=None):
+    def __init__(self, *args, target_max_score=None, **kwargs):
         self.target_max_score = target_max_score
+        super().__init__(*args, **kwargs)
     
     def _is_goal_complete(self, model_output, ground_truth_output):
         if self.target_max_score:

textattack/search_methods/attack.py

@@ -76,7 +76,7 @@ class Attack:
         transformations = original_transformations[:]
         for C in self.constraints:
             if len(transformations) == 0: break
-            transformations = C.call_many(text, transformations, original_text)
+            transformations = C.call_many(text, transformations, original_text=original_text)
         # Default to false for all original transformations.
         for original_transformation in original_transformations:
             self.constraints_cache[original_transformation] = False

textattack/shared/utils.py

@@ -216,7 +216,7 @@ def html_table_from_rows(rows, title=None, header=None, style_dict=None):
     return table_html
 
 def has_letter(word):
-    """ Returns true if `word` contains at least one character in [A-Za-z].
+    """ Returns true if `word` contains at least one character in [A-Za-z]. """
     for c in word:
         if c.isalpha(): return True
     return False

textattack/shared/word_embedding.py

@@ -48,4 +48,13 @@ class WordEmbedding:
             self.cos_sim_mat = {}
     
     def __getitem__(self, index):
+        """ Gets a word embedding by word or ID. 
+            
+            If word or ID not found, returns None.
+        """
+        if isinstance(index, str):
+            try:
+                index = self.word2index[index]
+            except KeyError:
+                return None
         return self.embeddings[index]