model: Added ModelNotTrained Error

Fixes: #125

Author: yashlamba

dffml/model/model.py

@@ -22,6 +22,10 @@
 from ..util.entrypoint import Entrypoint, base_entry_point
 
 
+class ModelNotTrained(Exception):
+    pass
+
+
 class ModelConfig(BaseConfig, NamedTuple):
     directory: str
 

model/scikit/dffml_model_scikit/scikit_base.py

@@ -5,6 +5,7 @@
 """
 import os
 import json
+import math
 import hashlib
 from pathlib import Path
 from typing import AsyncIterator, Tuple, Any, NamedTuple
@@ -16,7 +17,7 @@
 from dffml.repo import Repo
 from dffml.source.source import Sources
 from dffml.accuracy import Accuracy
-from dffml.model.model import ModelConfig, ModelContext, Model
+from dffml.model.model import ModelConfig, ModelContext, Model, ModelNotTrained
 
 
 class ScikitConfig(ModelConfig, NamedTuple):
@@ -33,7 +34,7 @@ def __init__(self, parent, features):
 
     @property
     def confidence(self):
-        return self.parent.saved.get(self._features_hash, None)
+        return self.parent.saved.get(self._features_hash, float("nan"))
 
     @confidence.setter
     def confidence(self, confidence):
@@ -60,7 +61,7 @@ async def __aenter__(self):
         return self
 
     async def __aexit__(self, exc_type, exc_value, traceback):
-        joblib.dump(self.clf, self._filename())
+        pass
 
     async def train(self, sources: Sources):
         data = []
@@ -77,6 +78,8 @@ async def train(self, sources: Sources):
         joblib.dump(self.clf, self._filename())
 
     async def accuracy(self, sources: Sources) -> Accuracy:
+        if not os.path.isfile(self._filename()):
+            raise ModelNotTrained("Train model before assessing for accuracy.")
         data = []
         async for repo in sources.with_features(self.features):
             feature_data = repo.features(
@@ -91,9 +94,11 @@ async def accuracy(self, sources: Sources) -> Accuracy:
         self.logger.debug("Model Accuracy: {}".format(self.confidence))
         return self.confidence
 
-    async def predict(self, repos: AsyncIterator[Repo]) -> AsyncIterator[Repo]:
-        if self.confidence is None:
-            raise ValueError("Model Not Trained")
+    async def predict(
+        self, repos: AsyncIterator[Repo]
+    ) -> AsyncIterator[Tuple[Repo, Any, float]]:
+        if not os.path.isfile(self._filename()):
+            raise ModelNotTrained("Train model before prediction.")
         async for repo in repos:
             feature_data = repo.features(self.features)
             df = pd.DataFrame(feature_data, index=[0])

model/scratch/dffml_model_scratch/slr.py

@@ -15,7 +15,7 @@
 from dffml.source.source import Sources
 from dffml.feature import Features
 from dffml.accuracy import Accuracy
-from dffml.model.model import ModelConfig, ModelContext, Model
+from dffml.model.model import ModelConfig, ModelContext, Model, ModelNotTrained
 from dffml.util.entrypoint import entry_point
 from dffml.util.cli.arg import Arg
 
@@ -110,11 +110,15 @@ async def train(self, sources: Sources):
 
     async def accuracy(self, sources: Sources) -> Accuracy:
         if self.regression_line is None:
-            raise ValueError("Model Not Trained")
+            raise ModelNotTrained("Train model before assessing for accuracy.")
         accuracy_value = self.regression_line[2]
         return Accuracy(accuracy_value)
 
-    async def predict(self, repos: AsyncIterator[Repo]) -> AsyncIterator[Repo]:
+    async def predict(
+        self, repos: AsyncIterator[Repo]
+    ) -> AsyncIterator[Tuple[Repo, Any, float]]:
+        if self.regression_line is None:
+            raise ModelNotTrained("Train model before prediction.")
         async for repo in repos:
             feature_data = repo.features(self.features)
             repo.predicted(

model/tensorflow/dffml_model_tensorflow/dnnc.py

@@ -16,7 +16,7 @@
 from dffml.repo import Repo
 from dffml.feature import Feature, Features
 from dffml.source.source import Sources
-from dffml.model.model import ModelConfig, ModelContext, Model
+from dffml.model.model import ModelConfig, ModelContext, Model, ModelNotTrained
 from dffml.accuracy import Accuracy
 from dffml.util.entrypoint import entry_point
 from dffml.base import BaseConfig
@@ -300,7 +300,7 @@ async def accuracy(self, sources: Sources) -> Accuracy:
         as test data.
         """
         if not os.path.isdir(self.model_dir_path):
-            raise NotADirectoryError("Model not trained")
+            raise ModelNotTrained("Train model before assessing for accuracy.")
         input_fn = await self.accuracy_input_fn(
             sources, batch_size=20, shuffle=False, epochs=1
         )
@@ -312,7 +312,7 @@ async def predict(self, repos: AsyncIterator[Repo]) -> AsyncIterator[Repo]:
         Uses trained data to make a prediction about the quality of a repo.
         """
         if not os.path.isdir(self.model_dir_path):
-            raise NotADirectoryError("Model not trained")
+            raise ModelNotTrained("Train model before prediction.")
         # Create the input function
         input_fn, predict = await self.predict_input_fn(repos)
         # Makes predictions on classifications