fix spacy example

examples/models/sklearn_spacy_text/.gitignore

@@ -0,0 +1 @@
+*.model

examples/models/sklearn_spacy_text/Dockerfile

@@ -0,0 +1,4 @@
+FROM seldonio/seldon-core-s2i-python3:1.1.1-rc
+
+RUN pip install spacy
+RUN python -m spacy download en_core_web_sm

examples/models/sklearn_spacy_text/ml_utils.py

@@ -1,15 +1,14 @@
 import spacy
-from spacy.cli import download
-import re 
+import re
 import numpy as np
 from sklearn.base import TransformerMixin
 from html.parser import HTMLParser
 import dill
 import sys, os
 
-download("en_core_web_sm")
-nlp = spacy.load('en_core_web_sm', parser=False, entity=False)
-    
+nlp = spacy.load("en_core_web_sm", parser=False, entity=False)
+
+
 class SpacyTokenTransformer(TransformerMixin):
     __symbols = set("!$%^&*()_+|~-=`{}[]:\";'<>?,./-")
 
@@ -20,29 +19,30 @@ def transform(self, X, **kwargs):
 
     def fit(self, X, y=None, **fit_params):
         return self
-    
+
     @staticmethod
     def transform_to_token(text):
         str_text = str(text)
-        doc = nlp(str_text, disable=['parser', 'tagger', 'ner'])
+        doc = nlp(str_text, disable=["parser", "tagger", "ner"])
         tokens = []
         for token in doc:
             if token.like_url:
                 clean_token = "URL"
             else:
                 clean_token = token.lemma_.lower().strip()
-                if len(clean_token) < 1 or clean_token in \
-                        SpacyTokenTransformer.__symbols: 
+                if (
+                    len(clean_token) < 1
+                    or clean_token in SpacyTokenTransformer.__symbols
+                ):
                     continue
             tokens.append(clean_token)
         return tokens
 
+
 class CleanTextTransformer(TransformerMixin):
     __html_parser = HTMLParser()
-    __uplus_pattern = \
-        re.compile("\<[uU]\+(?P<digit>[a-zA-Z0-9]+)\>")
-    __markup_link_pattern = \
-        re.compile("\[(.*)\]\((.*)\)")
+    __uplus_pattern = re.compile("\<[uU]\+(?P<digit>[a-zA-Z0-9]+)\>")
+    __markup_link_pattern = re.compile("\[(.*)\]\((.*)\)")
 
     def transform(self, X, **kwargs):
         f = np.vectorize(CleanTextTransformer.transform_clean_text)
@@ -51,25 +51,25 @@ def transform(self, X, **kwargs):
 
     def fit(self, X, y=None, **fit_params):
         return self
-    
+
     @staticmethod
     def transform_clean_text(raw_text):
         try:
             decoded = raw_text.encode("ISO-8859-1").decode("utf-8")
         except:
             decoded = raw_text.encode("ISO-8859-1").decode("cp1252")
-        html_unescaped = CleanTextTransformer.\
-            __html_parser.unescape(decoded) 
+        html_unescaped = CleanTextTransformer.__html_parser.unescape(decoded)
         html_unescaped = re.sub(r"\r\n", " ", html_unescaped)
         html_unescaped = re.sub(r"\r\r\n", " ", html_unescaped)
         html_unescaped = re.sub(r"\r", " ", html_unescaped)
         html_unescaped = html_unescaped.replace("&gt;", " > ")
         html_unescaped = html_unescaped.replace("&lt;", " < ")
         html_unescaped = html_unescaped.replace("--", " - ")
         html_unescaped = CleanTextTransformer.__uplus_pattern.sub(
-            " U\g<digit> ", html_unescaped)
+            " U\g<digit> ", html_unescaped
+        )
         html_unescaped = CleanTextTransformer.__markup_link_pattern.sub(
-            " \1 \2 ", html_unescaped)
+            " \1 \2 ", html_unescaped
+        )
         html_unescaped = html_unescaped.replace("\\", "")
         return html_unescaped
-

examples/models/sklearn_spacy_text/requirements.txt

@@ -1,5 +1,5 @@
-scipy>= 0.13.3
-scikit-learn>=0.18
-spacy==2.0.18
-dill==0.2.9
-seldon-core==0.2.7
+seldon-core>=1.1
+scipy>=1.0
+scikit-learn>=0.20
+spacy>=2.2
+dill>=0.3