feat: add opt-in to skip NLTK downloads

src/aidial_rag_eval/dataframe/metrics.py

@@ -114,6 +114,7 @@ def calculate_generation_metrics(
     metric_binds: List[MetricBind],
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> pd.DataFrame:
     """
     Calculates RAG evaluation generation metrics from df_merged dataframe.
@@ -155,6 +156,7 @@ def calculate_generation_metrics(
                 llm=llm,
                 max_concurrency=max_concurrency,
                 show_progress_bar=show_progress_bar,
+                auto_download_nltk=auto_download_nltk,
             ).to_dict(orient="series")
         )
     df_metrics = pd.DataFrame(data=metric_results)
@@ -178,6 +180,7 @@ def create_generation_metrics_report(
     metric_binds: List[MetricBind],
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> pd.DataFrame:
     """
     Calculates RAG evaluation generation metrics from input dataframes.
@@ -220,6 +223,7 @@ def create_generation_metrics_report(
         metric_binds,
         max_concurrency,
         show_progress_bar,
+        auto_download_nltk,
     )
     return pd.merge(df_merged, df_metrics, left_index=True, right_index=True)
 
@@ -232,6 +236,7 @@ def create_rag_eval_metrics_report(
     metric_binds: Optional[List[MetricBind]] = None,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> pd.DataFrame:
     """
     Calculates RAG evaluation metrics from input dataframes.
@@ -282,5 +287,6 @@ def create_rag_eval_metrics_report(
         metric_binds,
         max_concurrency,
         show_progress_bar,
+        auto_download_nltk,
     )
     return pd.concat([df_merged, retrieval_metrics, generation_metrics], axis=1)

src/aidial_rag_eval/evaluate.py

@@ -21,6 +21,7 @@ def evaluate(
     metric_binds: Optional[List[MetricBind]] = None,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = False,
 ) -> Dataset:
     """
     Calculates RAG evaluation metrics from input
@@ -82,6 +83,7 @@ def evaluate(
         metric_binds=metric_binds,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
     aggregated_metrics = df_final.mean(numeric_only=True)
     assert isinstance(aggregated_metrics, pd.Series)

src/aidial_rag_eval/generation/inference.py

@@ -184,6 +184,7 @@ def segment_hypotheses(
     llm: BaseChatModel,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> List[SegmentedText]:
     """
     Function that segments hypotheses into hypothesis segments(roughly into
@@ -215,7 +216,8 @@ def segment_hypotheses(
     )
 
     segmented_hypotheses = [
-        SegmentedText.from_text(text=hypothesis) for hypothesis in hypotheses
+        SegmentedText.from_text(text=hypothesis, auto_download_nltk=auto_download_nltk)
+        for hypothesis in hypotheses
     ]
     if show_progress_bar:
         print("Converting hypothesis...")
@@ -280,6 +282,7 @@ def infer_statements(
     list_documents: Optional[List[Documents]] = None,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> List[List[Tuple[InferenceInputs, InferenceScore]]]:
     """
     Function that infers statements.
@@ -330,7 +333,10 @@ def infer_statements(
         document_names = [_join_documents(docs) for docs in list_documents]
     if questions is not None:
         segmented_questions = [
-            SegmentedText.from_text(text=question) for question in questions
+            SegmentedText.from_text(
+                text=question, auto_download_nltk=auto_download_nltk
+            )
+            for question in questions
         ]
         premises = [
             question_split.segments[-1] + "\n" + premise
@@ -362,6 +368,7 @@ def calculate_batch_inference(
     list_documents: Optional[List[Documents]] = None,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> List[InferenceReturn]:
     """
     Calculates pairwise the inference of a hypotheses from a premises.
@@ -404,6 +411,7 @@ def calculate_batch_inference(
         llm=llm,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
     statements: List[List[List[Statement]]] = extract_statements(
         list_of_hypothesis_segments=[
@@ -423,6 +431,7 @@ def calculate_batch_inference(
             list_documents=list_documents,
             max_concurrency=max_concurrency,
             show_progress_bar=show_progress_bar,
+            auto_download_nltk=auto_download_nltk,
         )
     )
 
@@ -459,6 +468,7 @@ def calculate_inference(
     documents: Optional[Documents] = None,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> InferenceReturn:
     """
     Calculates the inference of a hypothesis from a premise.
@@ -504,5 +514,6 @@ def calculate_inference(
         list_documents=list_documents,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
     return inference_returns[0]

src/aidial_rag_eval/generation/metric_binds.py

@@ -32,6 +32,7 @@ def _wrapped_dataframe_inference(
     document_column: Optional[str] = None,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> pd.DataFrame:
     inference_returns = calculate_batch_inference(
         premises=_get_column_as_list_str(df_merged, premise_column),
@@ -49,6 +50,7 @@ def _wrapped_dataframe_inference(
         ),
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
     return pd.DataFrame(
         [vars(inference_return) for inference_return in inference_returns]
@@ -62,20 +64,22 @@ def _wrapped_dataframe_refusal(
     prefix: str,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> pd.DataFrame:
     refusal_returns = calculate_batch_refusal(
         answers=_get_column_as_list_str(df_merged, answer_column),
         llm=llm,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
     return pd.DataFrame([vars(refusal) for refusal in refusal_returns]).add_prefix(
         prefix
     )
 
 
 def context_to_answer_inference(
-    df_merged, llm, max_concurrency, show_progress_bar, **kwargs
+    df_merged, llm, max_concurrency, show_progress_bar, auto_download_nltk, **kwargs
 ) -> pd.DataFrame:
     return _wrapped_dataframe_inference(
         df_merged=df_merged,
@@ -93,11 +97,12 @@ def context_to_answer_inference(
         document_column=MergedColumns.DOCUMENTS,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
 
 
 def answer_to_ground_truth_inference(
-    df_merged, llm, max_concurrency, show_progress_bar, **kwargs
+    df_merged, llm, max_concurrency, show_progress_bar, auto_download_nltk, **kwargs
 ) -> pd.DataFrame:
     return _wrapped_dataframe_inference(
         df_merged=df_merged,
@@ -109,11 +114,12 @@ def answer_to_ground_truth_inference(
         document_column=MergedColumns.DOCUMENTS,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
 
 
 def ground_truth_to_answer_inference(
-    df_merged, llm, max_concurrency, show_progress_bar, **kwargs
+    df_merged, llm, max_concurrency, show_progress_bar, auto_download_nltk, **kwargs
 ) -> pd.DataFrame:
     return _wrapped_dataframe_inference(
         df_merged=df_merged,
@@ -125,11 +131,12 @@ def ground_truth_to_answer_inference(
         document_column=MergedColumns.DOCUMENTS,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
 
 
 def answer_refusal(
-    df_merged, llm, max_concurrency, show_progress_bar, **kwargs
+    df_merged, llm, max_concurrency, show_progress_bar, auto_download_nltk, **kwargs
 ) -> pd.DataFrame:
     return _wrapped_dataframe_refusal(
         df_merged=df_merged,
@@ -138,11 +145,12 @@ def answer_refusal(
         prefix=ANSWER_REFUSAL_PREFIX,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
 
 
 def ground_truth_refusal(
-    df_merged, llm, max_concurrency, show_progress_bar, **kwargs
+    df_merged, llm, max_concurrency, show_progress_bar, auto_download_nltk, **kwargs
 ) -> pd.DataFrame:
     return _wrapped_dataframe_refusal(
         df_merged=df_merged,
@@ -151,6 +159,7 @@ def ground_truth_refusal(
         prefix=GT_ANSWER_REFUSAL_PREFIX,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
 
 

src/aidial_rag_eval/generation/refusal.py

@@ -15,6 +15,7 @@ def calculate_batch_refusal(
     llm: BaseChatModel,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> List[RefusalReturn]:
     """
     Checks if the answers are answer refusal.
@@ -40,7 +41,10 @@ def calculate_batch_refusal(
         Returns the list of the answer refusals.
     """
     detector = LLMRefusalDetector(llm, max_concurrency)
-    answers_split = [SegmentedText.from_text(text=answer) for answer in answers]
+    answers_split = [
+        SegmentedText.from_text(text=answer, auto_download_nltk=auto_download_nltk)
+        for answer in answers
+    ]
     # As a heuristic, we send only the first 3 segments in the prompt.
     # We believe that if there are 3 whole segments with information
     # that is not related to refusal to answer,
@@ -60,6 +64,7 @@ def calculate_refusal(
     llm: BaseChatModel,
     max_concurrency: int = 8,
     show_progress_bar: bool = True,
+    auto_download_nltk: bool = True,
 ) -> RefusalReturn:
     """
     Checks if the answer is answer refusal.
@@ -89,5 +94,6 @@ def calculate_refusal(
         llm=llm,
         max_concurrency=max_concurrency,
         show_progress_bar=show_progress_bar,
+        auto_download_nltk=auto_download_nltk,
     )
     return refusal_returns[0]

src/aidial_rag_eval/generation/utils/segmented_text.py

@@ -80,8 +80,9 @@ def __init__(
         self.delimiters = delimiters.copy()
 
     @classmethod
-    def from_text(cls, text: Text) -> "SegmentedText":
-        nltk.download("punkt_tab", quiet=True)
+    def from_text(cls, text: Text, auto_download_nltk: bool = True) -> "SegmentedText":
+        if auto_download_nltk:
+            nltk.download("punkt_tab", quiet=True)
 
         max_len = 500
         min_len = 10