500 Internal Server Error
Gitea Version: 1.24.6
+
+
+ 
@@ -64,6 +64,7 @@ The first five are for classification and entailment attacks:
- **tf-adjusted**: TextFooler attack with constraint thresholds adjusted based on human evaluation and grammaticality enforced.
- **alz-adjusted**: Alzantot's attack adjusted to follow the same constraints as tf-adjusted such that the only difference is the search method.
- **deepwordbug**: Replace-1 scoring and multi-transformation character-swap attack (["Black-box Generation of Adversarial Text Sequences to Evade Deep Learning Classifiers"](https://arxiv.org/abs/1801.04354)).
+- **hotflip**: Beam search and gradient-based word swap (["HotFlip: White-Box Adversarial Examples for Text Classification"](https://arxiv.org/abs/1712.06751)
The final is for translation attacks:
- **seq2sick**: Greedy attack with goal of changing every word in the output translation. Currently implemented as black-box with plans to change to white-box as done in paper (["Seq2Sick: Evaluating the Robustness of Sequence-to-Sequence Models with Adversarial Examples"](https://arxiv.org/abs/1803.01128)).
@@ -78,17 +79,17 @@ To allow for word replacement after a sequence has been tokenized, we include a
TextAttack is model-agnostic! Anything that overrides `__call__`, takes in `TokenizedText`, and correctly formats output works. However, TextAttack provides pre-trained models and samples for the following datasets:
-Classification:
+#### Classification:
* AG News dataset topic classification
* IMDB dataset sentiment classification
* Movie Review dataset sentiment classification
* Yelp dataset sentiment classification
-Entailment:
+#### Entailment:
* SNLI datastet
* MNLI dataset (matched & unmatched)
-Translation:
+#### Translation:
* newstest2013 English to German dataset
### Attacks
diff --git a/textattack/attack_methods/beam_search.py b/textattack/attack_methods/beam_search.py
index a403070e..99bedadd 100644
--- a/textattack/attack_methods/beam_search.py
+++ b/textattack/attack_methods/beam_search.py
@@ -58,7 +58,6 @@ class BeamSearch(Attack):
# in descending order and filling the beam from there.
best_indices = -scores.argsort()[:self.beam_width]
beam = [potential_next_beam[i] for i in best_indices]
-
if best_result is None:
return FailedAttackResult(original_result, best_result)
diff --git a/textattack/attack_recipes/ebrahimi_2017_hotflip.py b/textattack/attack_recipes/ebrahimi_2017_hotflip.py
index 6f1b35c9..8938672d 100644
--- a/textattack/attack_recipes/ebrahimi_2017_hotflip.py
+++ b/textattack/attack_recipes/ebrahimi_2017_hotflip.py
@@ -23,7 +23,7 @@ def Ebrahimi2017HotFlip(model):
# "HotFlip ... uses the gradient with respect to a one-hot input
# representation to efficiently estimate which individual change has the
# highest estimated loss."
- transformation = GradientBasedWordSwap(model, replace_stopwords=False)
+ transformation = GradientBasedWordSwap(model, top_n=1, replace_stopwords=False)
constraints = []
#
# 0. "We were able to create only 41 examples (2% of the correctly-
diff --git a/textattack/constraints/semantics/word_embedding_distance.py b/textattack/constraints/semantics/word_embedding_distance.py
index 33a59b85..ad10f905 100644
--- a/textattack/constraints/semantics/word_embedding_distance.py
+++ b/textattack/constraints/semantics/word_embedding_distance.py
@@ -116,6 +116,8 @@ class WordEmbeddingDistance(Constraint):
x_adv = x_adv.words[i]
except AttributeError:
raise AttributeError('Cannot apply word embedding distance constraint without `modified_word_index`')
+ except IndexError:
+ raise IndexError(f'Could not find word at index {i} with x {x} x_adv {x_adv}.')
if not self.cased:
# If embedding vocabulary is all lowercase, lowercase words.
Gitea Version: 1.24.6