Merge pull request #55 from nnick14/update_hf_tf

Update TensorFlow and Transformers versions

aspect_based_sentiment_analysis/models.py

@@ -1,9 +1,12 @@
 import logging
 from abc import ABC
 from abc import abstractmethod
-from typing import Tuple
+from typing import Tuple, Optional, Union
+
+import numpy as np
 
 import transformers
+from transformers.modeling_tf_utils import TFModelInputType
 import tensorflow as tf
 from tensorflow.keras import layers
 
@@ -120,9 +123,11 @@ class BertABSClassifier(ABSClassifier, transformers.TFBertPreTrainedModel):
     def __init__(self, config: BertABSCConfig, **kwargs):
         super().__init__(config, **kwargs)
         self.bert = transformers.TFBertMainLayer(
-            config, name="bert")
+            config, name="bert"
+        )
         initializer = transformers.modeling_tf_utils.get_initializer(
-            config.initializer_range)
+            config.initializer_range
+        )
         self.dropout = layers.Dropout(config.hidden_dropout_prob)
         self.classifier = layers.Dense(
             config.num_polarities,
@@ -131,21 +136,47 @@ def __init__(self, config: BertABSCConfig, **kwargs):
         )
 
     def call(
-            self,
-            token_ids: tf.Tensor,
-            attention_mask: tf.Tensor = None,
-            token_type_ids: tf.Tensor = None,
-            training: bool = False,
-            **bert_kwargs
+        self,
+        input_ids: Optional[TFModelInputType] = None,
+        attention_mask: Optional[Union[np.ndarray, tf.Tensor]] = None,
+        token_type_ids: Optional[Union[np.ndarray, tf.Tensor]] = None,
+        position_ids: Optional[Union[np.ndarray, tf.Tensor]] = None,
+        head_mask: Optional[Union[np.ndarray, tf.Tensor]] = None,
+        inputs_embeds: Optional[Union[np.ndarray, tf.Tensor]] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        training: Optional[bool] = False,
+        **kwargs,
     ) -> Tuple[tf.Tensor, Tuple[tf.Tensor, ...], Tuple[tf.Tensor, ...]]:
-        outputs = self.bert(
-            inputs=token_ids,
+        inputs = transformers.modeling_tf_utils.input_processing(
+            func=self.call,
+            config=self.config,
+            input_ids=input_ids,
             attention_mask=attention_mask,
             token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
             training=training,
-            **bert_kwargs
+            kwargs_call=kwargs,
+        )
+        outputs = self.bert(
+            input_ids=inputs["input_ids"],
+            attention_mask=inputs["attention_mask"],
+            token_type_ids=inputs["token_type_ids"],
+            position_ids=inputs["position_ids"],
+            head_mask=inputs["head_mask"],
+            inputs_embeds=inputs["inputs_embeds"],
+            output_attentions=inputs["output_attentions"],
+            output_hidden_states=inputs["output_hidden_states"],
+            return_dict=inputs["return_dict"],
+            training=inputs["training"],
         )
-        sequence_output, pooled_output, hidden_states, attentions = outputs
+        pooled_output = outputs[1]
         pooled_output = self.dropout(pooled_output, training=training)
         logits = self.classifier(pooled_output)
-        return logits, hidden_states, attentions
+        return logits, outputs.hidden_states, outputs.attentions

aspect_based_sentiment_analysis/pipelines.py

@@ -92,8 +92,7 @@ def preprocess(self, text: str, aspects: List[str]) -> Task:
         """
 
     @abstractmethod
-    def tokenize(self, examples: Iterable[Example]) -> Iterable[
-        TokenizedExample]:
+    def tokenize(self, examples: Iterable[Example]) -> Iterable[TokenizedExample]:
         """
         Tokenize the example. The model can not process the raw pair of two
         strings (text, aspect) directly.
@@ -113,8 +112,8 @@ def tokenize(self, examples: Iterable[Example]) -> Iterable[
     def encode(self, examples: Iterable[TokenizedExample]) -> InputBatch:
         """
         Encode tokenized examples. The input batch is a container of tensors
-        crucial for the model to make a prediction. The names are compatible 
-        with the *transformers* package. 
+        crucial for the model to make a prediction. The names are compatible
+        with the *transformers* package.
 
         Parameters
         ----------
@@ -250,7 +249,7 @@ def predict(self, input_batch: InputBatch) -> OutputBatch:
         # output including hidden states and attentions.
         with tf.GradientTape() as tape:
             logits, hidden_states, attentions = self.model.call(
-                token_ids=input_batch.token_ids,
+                input_ids=input_batch.token_ids,
                 attention_mask=input_batch.attention_mask,
                 token_type_ids=input_batch.token_type_ids
             )

aspect_based_sentiment_analysis/text_splitters.py

@@ -14,7 +14,7 @@ def sentencizer(name: str = 'en_core_web_sm') -> Callable[[str], List[str]]:
 
     def wrapper(text: str) -> List[str]:
         doc = nlp(text)
-        sentences = [sent.string.strip() for sent in doc.sents]
+        sentences = [str(sent).strip() for sent in doc.sents]
         return sentences
 
     return wrapper

aspect_based_sentiment_analysis/training/datasets.py

@@ -80,7 +80,7 @@ def preprocess_batch(
         encoded = self.tokenizer.batch_encode_plus(
             pairs,
             add_special_tokens=True,
-            pad_to_max_length=True,
+            padding=True,
             return_attention_masks=True,
             return_tensors='tf'
         )

environment.yml

@@ -1,15 +1,15 @@
 name: Aspect-Based-Sentiment-Analysis
 dependencies:
   - python=3.6.9
-  - scikit-learn=0.22
-  - pytest=5.3
-  - pytest-timeout=1.3
+  - scikit-learn=0.24.2
+  - pytest=6.2.4
+  - pytest-timeout=1.4.2
   - ipython=7.3
-  - spacy=2.3
-  - dataclasses=0.7
+  - spacy=3.1.0
+  - dataclasses=0.8
   - pip:
-      - tensorflow==2.2
-      - google-cloud-storage==1.26
-      - testfixtures==6.14
-      - transformers==2.5
-      - optuna==1.3
+      - tensorflow==2.5.0
+      - google-cloud-storage==1.40.0
+      - testfixtures==6.17.1
+      - transformers==4.8.2
+      - optuna==2.8.0

setup.py

@@ -24,8 +24,8 @@
     include_package_data=False,
     packages=find_packages(),
     install_requires=[
-        'tensorflow==2.4.0',
-        'transformers==2.5',
+        'tensorflow==2.5.0',
+        'transformers==4.8.2',
         'pytest',
         'scikit-learn',
         'ipython',