Jm/py 3.6 (#19)

* Modify TYPE_CHECKING annotations for py3.6 support

* Install dataclasses for CI/CD for py3.6 usage

* README update

* Add dataclasses to setup as an extra for 3.6 support

Co-authored-by: John Myers <john@gretel.ai>

.github/workflows/main.yml

@@ -10,9 +10,10 @@ jobs:
     - name: Setup Python
       uses: actions/setup-python@v1
       with:
-        python-version: 3.7
+        python-version: 3.6
     - name: Install
       run: |
+        pip install dataclasses==0.7
         pip install -e .
         pip install -r test-requirements.txt
     - name: Lint

README.md

@@ -5,7 +5,7 @@
 [![Documentation Status](https://readthedocs.org/projects/gretel-synthetics/badge/?version=stable)](https://gretel-synthetics.readthedocs.io/en/stable/?badge=stable)
 
 
-This code has been developed and tested on Python 3.7. Python 3.8 is currently unsupported.
+This code has been developed and tested on Python 3.7. Python 3.8 is currently unsupported. While not developed on Python 3.6, this code will run in Google Colab, which currently uses 3.6. If you wish to use Python 3.6, out side of Google Colab, you may install with the `3.6` extras: `pip install gretel-synthetics[tf,3.6]`, for example.
 
 This package allows developers to quickly get immersed with synthetic data generation through the use of neural networks. The more complex pieces of working with libraries like Tensorflow and differential privacy are bundled into friendly Python classes and functions.
 

setup.py

@@ -27,6 +27,7 @@
         'numpy==1.18.3'
     ],
     extras_require={
-        'tf': ['tensorflow==2.1.0']
+        'tf': ['tensorflow==2.1.0'],
+        '3.6': ['dataclasses==0.7']
     }
 )

src/gretel_synthetics/generate.py

@@ -6,7 +6,6 @@
     - Created a config
     - Trained a model
 """
-from __future__ import annotations
 import logging
 import sentencepiece as spm
 import tensorflow as tf
@@ -69,14 +68,14 @@ def values_as_list(self) -> List[str]:
 )
 
 
-def _load_tokenizer(store: _BaseConfig) -> spm.SentencePieceProcessor:
+def _load_tokenizer(store: "_BaseConfig") -> spm.SentencePieceProcessor:
     logging.info("Loading SentencePiece tokenizer")
     sp = spm.SentencePieceProcessor()
     sp.Load(store.tokenizer_model)
     return sp
 
 
-def _prepare_model(sp: spm, batch_size: int, store: _BaseConfig) -> tf.keras.Sequential:
+def _prepare_model(sp: spm, batch_size: int, store: "_BaseConfig") -> tf.keras.Sequential:
     model = _build_sequential_model(
         vocab_size=len(sp), batch_size=batch_size, store=store
     )
@@ -91,14 +90,14 @@ def _prepare_model(sp: spm, batch_size: int, store: _BaseConfig) -> tf.keras.Seq
     return model
 
 
-def _load_model(store: _BaseConfig) -> Tuple[spm.SentencePieceProcessor, tf.keras.Sequential]:
+def _load_model(store: "_BaseConfig") -> Tuple[spm.SentencePieceProcessor, tf.keras.Sequential]:
     sp = _load_tokenizer(store)
     model = _prepare_model(sp, 1, store)
     return sp, model
 
 
 def generate_text(
-    store: _BaseConfig, start_string: str = "<n>", line_validator: callable = None
+    store: "_BaseConfig", start_string: str = "<n>", line_validator: callable = None
 ):
     """A generator that will load a model and start creating records.
 
@@ -169,7 +168,7 @@ def _predict_chars(
     model: tf.keras.Sequential,
     sp: spm.SentencePieceProcessor,
     start_string: str,
-    store: _BaseConfig,
+    store: "_BaseConfig",
 ) -> str:
     """
     Evaluation step (generating text using the learned model).