Write text parser information to model.info.txt

Author: wpm

bisemantic/classifier.py

@@ -17,7 +17,7 @@
 from keras.models import load_model
 
 from bisemantic import logger
-from bisemantic.data import TextPairEmbeddingGenerator, embedding_size, label
+from bisemantic.data import TextPairEmbeddingGenerator, embedding_size, label, text_parser_info
 
 
 class TextPairClassifier(object):
@@ -58,7 +58,7 @@ def train(cls, training_data, bidirectional, lstm_units, epochs, dropout=None, m
         if model_directory is not None:
             os.makedirs(model_directory)
             with open(cls._info_filename(model_directory), "w") as f:
-                f.write("%s\n" % model)
+                f.write("%s\n%s\n" % (text_parser_info(), model))
         return cls._train(epochs, model, model_directory, training, validation_data)
 
     @classmethod

bisemantic/data.py

@@ -258,6 +258,11 @@ def embedding_size():
     return _load_text_parser().vocab.vectors_length
 
 
+def text_parser_info():
+    _load_text_parser()
+    return _text_parser_description()
+
+
 # Singleton instance of text tokenizer and embedder.
 text_parser = None
 
@@ -266,5 +271,10 @@ def _load_text_parser():
     global text_parser
     if text_parser is None:
         text_parser = spacy.load("en", tagger=None, parser=None, entity=None)
-        logger.info("%s: %s" % (text_parser.meta["name"], text_parser.meta["description"]))
+        logger.info(_text_parser_description())
     return text_parser
+
+
+def _text_parser_description():
+    # This assumes the text parser has already been loaded.
+    return "%s: %s Embedding size %d" % (text_parser.meta["name"], text_parser.meta["description"], embedding_size())