Merge pull request #476 from QData/api-tests

Fix bugs and update training test

tests/test_command_line/test_train.py

@@ -5,7 +5,7 @@
 
 
 def test_train_tiny():
-    command = "textattack train --model-name-or-path lstm --attack deepwordbug --dataset glue^cola --model-max-length 32 --num-epochs 1 --num-train-adv-examples 200"
+    command = "textattack train --model-name-or-path lstm --attack deepwordbug --dataset glue^cola --model-max-length 32 --num-epochs 2 --num-clean-epochs 1 --num-train-adv-examples 200"
 
     # Run command and validate outputs.
     result = run_command_and_get_result(command)

textattack/models/tokenizers/glove_tokenizer.py

@@ -128,6 +128,7 @@ def __init__(
         self.pad_token_id = pad_token_id
         self.oov_token_id = unk_token_id
         self.convert_id_to_word = self.id_to_token
+        self.model_max_length = max_length
         # Set defaults.
         self.enable_padding(length=max_length, pad_id=pad_token_id)
         self.enable_truncation(max_length=max_length)

textattack/trainer.py

@@ -539,7 +539,7 @@ def train(self):
         else:
             train_batch_size = self.training_args.per_device_train_batch_size
 
-        if self.training_args.attack is None:
+        if self.attack is None:
             num_clean_epochs = self.training_args.num_epochs
         else:
             num_clean_epochs = self.training_args.num_clean_epochs
@@ -831,12 +831,15 @@ def _write_readme(self, best_eval_score, best_eval_score_epoch, train_batch_size
             and self.training_args.model_max_length
         ):
             model_max_length = self.training_args.model_max_length
-        elif isinstance(self.model_wrapper.model, transformers.PreTrainedModel):
-            model_max_length = self.model_wrapper.model.config.max_position_embeddings
         elif isinstance(
-            self.model_wrapper.model, (LSTMForClassification, WordCNNForClassification)
+            self.model_wrapper.model,
+            (
+                transformers.PreTrainedModel,
+                LSTMForClassification,
+                WordCNNForClassification,
+            ),
         ):
-            model_max_length = self.model_wrapper.model.max_length
+            model_max_length = self.model_wrapper.tokenizer.model_max_length
         else:
             model_max_length = None