Sp updates (#23)

* Upgrade SentencePiece and enable "max_sentence_length" option

* version bump

Co-authored-by: John Myers <[email protected]>

VERSION

@@ -1 +1 @@
-0.9.2
+0.9.3

examples/synthetic_records.ipynb

@@ -45,6 +45,7 @@
     "    max_lines=0, # use max_lines of training data. Set to 0 (zero) to on all lines in dataset\n",
     "    epochs=15, # 15-50 epochs with GPU for best performance\n",
     "    vocab_size=15000, # tokenizer model vocabulary size\n",
+    "    max_line_len=2048,  # the max line length SentencePiece will consider\n",
     "    character_coverage=1.0, # tokenizer model character coverage percent\n",
     "    gen_chars=0, # the maximum number of characters possible per-generated line of text\n",
     "    gen_lines=100, # the number of generated text lines\n",
@@ -61,7 +62,7 @@
     "    field_delimiter=\",\",  # if the training text is structured\n",
     "    # overwrite=True,  # enable this if you want to keep training models to the same checkpoint location\n",
     "    input_data_path=\"https://gretel-public-website.s3-us-west-2.amazonaws.com/datasets/uber_scooter_rides_1day.csv\" # filepath or S3\n",
-    ")"
+    ")\n"
    ]
   },
   {

setup.py

@@ -20,7 +20,7 @@
     packages=find_packages('src'),
     install_requires=[
         'tensorflow_privacy==0.2.2',
-        'sentencepiece==0.1.85',
+        'sentencepiece==0.1.91',
         'smart_open==1.10.0',
         'tqdm<5.0',
         'pandas==1.0.3',

src/gretel_synthetics/config.py

@@ -39,6 +39,7 @@ class _BaseConfig:
     rnn_units: int = 256
     dropout_rate: float = 0.2
     rnn_initializer: str = "glorot_uniform"
+    max_line_len: int = 2048
 
     # Input data configs
     field_delimiter: Optional[str] = None

src/gretel_synthetics/train.py

@@ -158,13 +158,24 @@ def _train_tokenizer(store: _BaseConfig) -> spm.SentencePieceProcessor:
     Trains SentencePiece tokenizer on training data
     """
     logging.info("Training SentencePiece tokenizer")
+    spm.SentencePieceTrainer.Train(
+        input=store.training_data,
+        model_prefix=store.tokenizer_prefix,
+        user_defined_symbols=["<n>", store.field_delimiter_token],
+        vocab_size=store.vocab_size,
+        hard_vocab_limit=False,
+        max_sentence_length=store.max_line_len,
+        character_coverage=store.character_coverage
+    )
+    """
     spm.SentencePieceTrainer.Train(
         f'--input={store.training_data} '
         f'--model_prefix={store.tokenizer_prefix} '
         f'--user_defined_symbols=<n>,{store.field_delimiter_token} '
         f'--vocab_size={store.vocab_size} '
         f'--hard_vocab_limit=false '
         f'--character_coverage={store.character_coverage}')
+    """
     _move_tokenizer_model(store)
 
     sp = spm.SentencePieceProcessor()

tests/test_config.py

@@ -48,6 +48,7 @@ def test_local_config_settings(mkdir):
         "gen_temp": 1.0,
         "gen_chars": 0,
         "gen_lines": 500,
+        "max_line_len": 2048,
         "save_all_checkpoints": True,
         "checkpoint_dir": "foo",
         "field_delimiter": None,