Merge pull request #18 from VowpalWabbit/renames

Renames

Author: olgavrou

README.md

@@ -257,27 +257,27 @@ The default `VwPolicy` is initialized with some default arguments. The default e
 
 The `VwPolicy`'s data files can be stored and examined or used to do [off policy evaluation](https://vowpalwabbit.org/docs/vowpal_wabbit/python/latest/tutorials/off_policy_evaluation.html) for hyper parameter tuning.
 
-The way to do this is to set a log file path to `vw_logs` on chain creation:
+The way to do this is to set a log file path to `rl_logs` on chain creation:
 
-`picker = learn_to_pick.PickBest.create(vw_logs=<path to log FILE>, [...])`
+`picker = learn_to_pick.PickBest.create(rl_logs=<path to log FILE>, [...])`
 
 ### Advanced featurization options
 
 #### auto_embed
 
 By default the input that is wrapped (`ToSelectFrom`, `BasedOn`) is not tampered with. This might not be sufficient featurization, so based on how complex the scenario is you can set auto-embeddings to ON
 
-`picker = learn_to_pick.PickBest.create(feature_embedder=learn_to_pick.PickBestFeatureEmbedder(auto_embed=True), [...])`
+`picker = learn_to_pick.PickBest.create(featurizer=learn_to_pick.PickBestFeaturizer(auto_embed=True), [...])`
 
 This will produce more complex embeddings and featurizations of the inputs, likely accelerating RL learning, albeit at the cost of increased runtime.
 
-By default, [sbert.net's sentence_transformers's ](https://www.sbert.net/docs/pretrained_models.html#model-overview) `all-mpnet-base-v2` model will be used for these embeddings but you can set a different embeddings model by initializing feature_embedder with a different model. You could also set an entirely different embeddings encoding object, as long as it has an `encode()` function that returns a list of the encodings.
+By default, [sbert.net's sentence_transformers's ](https://www.sbert.net/docs/pretrained_models.html#model-overview) `all-mpnet-base-v2` model will be used for these embeddings but you can set a different embeddings model by initializing featurizer with a different model. You could also set an entirely different embeddings encoding object, as long as it has an `encode()` function that returns a list of the encodings.
 
 ```python
 from sentence_transformers import SentenceTransformer
 
 picker = learn_to_pick.PickBest.create(
-    feature_embedder=learn_to_pick.PickBestFeatureEmbedder(
+    featurizer=learn_to_pick.PickBestFeaturizer(
         auto_embed=True,
         model=SentenceTransformer("all-mpnet-base-v2")
     ),

notebooks/readme.ipynb

@@ -87,7 +87,7 @@
     "\n",
     "# with delayed score\n",
     "\n",
-    "picker = PickBest.create(selection_scorer=None)\n",
+    "picker = learn_to_pick.PickBest.create(selection_scorer=None)\n",
     "result = picker.run(pick = learn_to_pick.ToSelectFrom([\"option1\", \"option2\"]),\n",
     "                    criteria = learn_to_pick.BasedOn(\"some criteria\")\n",
     ")\n",
@@ -297,7 +297,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "picker = learn_to_pick.PickBest.create(vw_logs='./logs', selection_scorer=None)"
+    "picker = learn_to_pick.PickBest.create(rl_logs='./logs', selection_scorer=None)"
    ]
   },
   {
@@ -319,7 +319,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "picker = learn_to_pick.PickBest.create(vw_logs='./logs', selection_scorer=None, feature_embedder=learn_to_pick.PickBestFeatureEmbedder(auto_embed=True))\n",
+    "picker = learn_to_pick.PickBest.create(rl_logs='./logs', selection_scorer=None, featurizer=learn_to_pick.PickBestFeaturizer(auto_embed=True))\n",
     "result = picker.run(pick = learn_to_pick.ToSelectFrom([\"option1\", \"option2\"]),\n",
     "                    criteria = learn_to_pick.BasedOn(\"age:32\")\n",
     ")\n",
@@ -355,11 +355,11 @@
     "from sentence_transformers import SentenceTransformer\n",
     "\n",
     "picker = learn_to_pick.PickBest.create(\n",
-    "    feature_embedder=learn_to_pick.PickBestFeatureEmbedder(\n",
+    "    featurizer=learn_to_pick.PickBestFeaturizer(\n",
     "        auto_embed=True,\n",
     "        model=SentenceTransformer(\"multi-qa-mpnet-base-dot-v1\")\n",
     "    ),\n",
-    "    vw_logs='./logs',\n",
+    "    rl_logs='./logs',\n",
     "    selection_scorer=None\n",
     ")\n",
     "\n",
@@ -378,10 +378,10 @@
    "outputs": [],
    "source": [
     "picker = learn_to_pick.PickBest.create(\n",
-    "    feature_embedder=learn_to_pick.PickBestFeatureEmbedder(\n",
+    "    featurizer=learn_to_pick.PickBestFeaturizer(\n",
     "        auto_embed=False,\n",
     "    ),\n",
-    "    vw_logs='./logs',\n",
+    "    rl_logs='./logs',\n",
     "    selection_scorer=None\n",
     ")\n",
     "\n",

src/learn_to_pick/__init__.py

@@ -4,7 +4,7 @@
     AutoSelectionScorer,
     BasedOn,
     Embed,
-    Embedder,
+    Featurizer,
     ModelRepository,
     Policy,
     SelectionScorer,
@@ -17,7 +17,7 @@
 from learn_to_pick.pick_best import (
     PickBest,
     PickBestEvent,
-    PickBestFeatureEmbedder,
+    PickBestFeaturizer,
     PickBestRandomPolicy,
     PickBestSelected,
 )
@@ -41,14 +41,14 @@ def configure_logger() -> None:
     "PickBest",
     "PickBestEvent",
     "PickBestSelected",
-    "PickBestFeatureEmbedder",
+    "PickBestFeaturizer",
     "PickBestRandomPolicy",
     "Embed",
     "BasedOn",
     "ToSelectFrom",
     "SelectionScorer",
     "AutoSelectionScorer",
-    "Embedder",
+    "Featurizer",
     "ModelRepository",
     "Policy",
     "VwPolicy",

src/learn_to_pick/base.py

@@ -181,7 +181,7 @@ def __init__(
         self,
         model_repo: ModelRepository,
         vw_cmd: List[str],
-        feature_embedder: Embedder,
+        featurizer: Featurizer,
         vw_logger: VwLogger,
         *args: Any,
         **kwargs: Any,
@@ -190,35 +190,35 @@ def __init__(
         self.model_repo = model_repo
         self.vw_cmd = vw_cmd
         self.workspace = self.model_repo.load(vw_cmd)
-        self.feature_embedder = feature_embedder
+        self.featurizer = featurizer
         self.vw_logger = vw_logger
 
     def predict(self, event: TEvent) -> Any:
         import vowpal_wabbit_next as vw
 
         text_parser = vw.TextFormatParser(self.workspace)
         return self.workspace.predict_one(
-            parse_lines(text_parser, self.feature_embedder.format(event))
+            parse_lines(text_parser, self.featurizer.format(event))
         )
 
     def learn(self, event: TEvent) -> None:
         import vowpal_wabbit_next as vw
 
-        vw_ex = self.feature_embedder.format(event)
+        vw_ex = self.featurizer.format(event)
         text_parser = vw.TextFormatParser(self.workspace)
         multi_ex = parse_lines(text_parser, vw_ex)
         self.workspace.learn_one(multi_ex)
 
     def log(self, event: TEvent) -> None:
         if self.vw_logger.logging_enabled():
-            vw_ex = self.feature_embedder.format(event)
+            vw_ex = self.featurizer.format(event)
             self.vw_logger.log(vw_ex)
 
     def save(self) -> None:
         self.model_repo.save(self.workspace)
 
 
-class Embedder(Generic[TEvent], ABC):
+class Featurizer(Generic[TEvent], ABC):
     def __init__(self, *args: Any, **kwargs: Any):
         pass
 
@@ -325,12 +325,12 @@ class RLLoop(Generic[TEvent]):
         - metrics (Optional[Union[MetricsTrackerRollingWindow, MetricsTrackerAverage]]): Tracker for metrics, can be set to None.
 
     Initialization Attributes:
-        - feature_embedder (Embedder): Embedder used for the `BasedOn` and `ToSelectFrom` inputs.
+        - featurizer (Featurizer): Featurizer used for the `BasedOn` and `ToSelectFrom` inputs.
         - model_save_dir (str, optional): Directory for saving the VW model. Default is the current directory.
         - reset_model (bool): If set to True, the model starts training from scratch. Default is False.
         - vw_cmd (List[str], optional): Command line arguments for the VW model.
         - policy (Type[VwPolicy]): Policy used by the chain.
-        - vw_logs (Optional[Union[str, os.PathLike]]): Path for the VW logs.
+        - rl_logs (Optional[Union[str, os.PathLike]]): Path for the VW logs.
         - metrics_step (int): Step for the metrics tracker. Default is -1. If set without metrics_window_size, average metrics will be tracked, otherwise rolling window metrics will be tracked.
         - metrics_window_size (int): Window size for the metrics tracker. Default is -1. If set, rolling window metrics will be tracked.
 

src/learn_to_pick/pick_best.py

@@ -42,9 +42,9 @@ def __init__(
         self.based_on = based_on
 
 
-class PickBestFeatureEmbedder(base.Embedder[PickBestEvent]):
+class PickBestFeaturizer(base.Featurizer[PickBestEvent]):
     """
-    Text Embedder class that embeds the `BasedOn` and `ToSelectFrom` inputs into a format that can be used by the learning policy
+    Text Featurizer class that embeds the `BasedOn` and `ToSelectFrom` inputs into a format that can be used by the learning policy
 
     Attributes:
         model name (Any, optional): The type of embeddings to be used for feature representation. Defaults to BERT SentenceTransformer.
@@ -259,7 +259,7 @@ class PickBest(base.RLLoop[PickBestEvent]):
         RLLoop
 
     Attributes:
-        feature_embedder (PickBestFeatureEmbedder, optional): Is an advanced attribute. Responsible for embedding the `BasedOn` and `ToSelectFrom` inputs. If omitted, a default embedder is utilized.
+        featurizer (PickBestFeaturizer, optional): Is an advanced attribute. Responsible for embedding the `BasedOn` and `ToSelectFrom` inputs. If omitted, a default embedder is utilized.
     """
 
     def _call_before_predict(self, inputs: Dict[str, Any]) -> PickBestEvent:
@@ -341,11 +341,11 @@ def create(
             selection_scorer = base.AutoSelectionScorer(llm=llm)
 
         policy_args = {
-            "feature_embedder": kwargs.pop("feature_embedder", None),
+            "featurizer": kwargs.pop("featurizer", None),
             "vw_cmd": kwargs.pop("vw_cmd", None),
             "model_save_dir": kwargs.pop("model_save_dir", None),
             "reset_model": kwargs.pop("reset_model", None),
-            "vw_logs": kwargs.pop("vw_logs", None),
+            "rl_logs": kwargs.pop("rl_logs", None),
         }
 
         if policy and any(policy_args.values()):
@@ -366,14 +366,14 @@ def create(
 
     @staticmethod
     def create_policy(
-        feature_embedder: Optional[base.Embedder] = None,
+        featurizer: Optional[base.Featurizer] = None,
         vw_cmd: Optional[List[str]] = None,
         model_save_dir: str = "./",
         reset_model: bool = False,
-        vw_logs: Optional[Union[str, os.PathLike]] = None
+        rl_logs: Optional[Union[str, os.PathLike]] = None
     ):
-        if not feature_embedder:
-            feature_embedder = PickBestFeatureEmbedder(auto_embed=False)
+        if not featurizer:
+            featurizer = PickBestFeaturizer(auto_embed=False)
 
         vw_cmd = vw_cmd or []
         interactions = []
@@ -391,7 +391,7 @@ def create_policy(
                 "--quiet",
             ]
 
-        if feature_embedder.auto_embed:
+        if featurizer.auto_embed:
             interactions += [
                 "--interactions=@#",
                 "--ignore_linear=@",
@@ -405,8 +405,8 @@ def create_policy(
                 model_save_dir, with_history=True, reset=reset_model
             ),
             vw_cmd=vw_cmd,
-            feature_embedder=feature_embedder,
-            vw_logger=base.VwLogger(vw_logs),
+            featurizer=featurizer,
+            vw_logger=base.VwLogger(rl_logs),
         )
 
     def _default_policy(self):