Merge pull request rllm-org#179 from UKGovernmentBEIS/bugfix/sevenllm-dependencies

Bugfix: Import SEvenLLM dependencies locally

Author: MattFisher

src/inspect_evals/sevenllm/scorers.py

@@ -1,10 +1,7 @@
 import re
 
-import jieba  # type: ignore
 from inspect_ai.scorer import Score, Scorer, Target, mean, scorer, stderr
 from inspect_ai.solver import TaskState
-from rouge import Rouge  # type: ignore
-from sentence_transformers import SentenceTransformer, util
 
 SENTENCE_TRANSFORMER_REPO = (
     "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
@@ -24,6 +21,9 @@ def split_sentences(text: str, is_zh: bool) -> list[str]:
 
 def tokenize_sentence(sentence: str, is_zh: bool) -> str:
     """Tokenize sentences (for ROUGE-L scoring)."""
+    # Import jieba here to avoid loading it in the global scope
+    import jieba  # type: ignore
+
     return " ".join(jieba.cut(sentence)) if is_zh else sentence
 
 
@@ -58,6 +58,9 @@ def rouge_l_scorer(is_zh: bool) -> Scorer:
         and provides the result alongside the model-generated output and a
         detailed explanation.
     """
+    # Import rouge here to avoid loading it in the global scope
+    from rouge import Rouge  # type: ignore
+
     rouge = Rouge()
 
     async def score(state: TaskState, target: Target) -> Score:
@@ -111,6 +114,9 @@ def semantic_similarity_scorer() -> Scorer:
         Scorer: A scoring function that computes semantic similarity and provides
         both the similarity score and a detailed explanation of the result.
     """
+    # Import SentenceTransformer here to avoid loading it in the global scope
+    from sentence_transformers import SentenceTransformer, util  # type: ignore
+
     model = SentenceTransformer(SENTENCE_TRANSFORMER_REPO)
 
     async def score(state: TaskState, target: Target) -> Score: