98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201 | class FaithfulnessEvaluator(BaseEvaluator):
"""
Faithfulness evaluator.
Evaluates whether a response is faithful to the contexts
(i.e. whether the response is supported by the contexts or hallucinated.)
This evaluator only considers the response string and the list of context strings.
Args:
raise_error(bool): Whether to raise an error when the response is invalid.
Defaults to False.
eval_template(Optional[Union[str, BasePromptTemplate]]):
The template to use for evaluation.
refine_template(Optional[Union[str, BasePromptTemplate]]):
The template to use for refining the evaluation.
"""
def __init__(
self,
llm: Optional[LLM] = None,
raise_error: bool = False,
eval_template: Optional[Union[str, BasePromptTemplate]] = None,
refine_template: Optional[Union[str, BasePromptTemplate]] = None,
) -> None:
"""Init params."""
self._llm = llm or Settings.llm
self._raise_error = raise_error
self._eval_template: BasePromptTemplate
if isinstance(eval_template, str):
self._eval_template = PromptTemplate(eval_template)
if isinstance(eval_template, BasePromptTemplate):
self._eval_template = eval_template
else:
model_name = self._llm.metadata.model_name
self._eval_template = TEMPLATES_CATALOG.get(
model_name, DEFAULT_EVAL_TEMPLATE
)
self._refine_template: BasePromptTemplate
if isinstance(refine_template, str):
self._refine_template = PromptTemplate(refine_template)
else:
self._refine_template = refine_template or DEFAULT_REFINE_TEMPLATE
def _get_prompts(self) -> PromptDictType:
"""Get prompts."""
return {
"eval_template": self._eval_template,
"refine_template": self._refine_template,
}
def _update_prompts(self, prompts: PromptDictType) -> None:
"""Update prompts."""
if "eval_template" in prompts:
self._eval_template = prompts["eval_template"]
if "refine_template" in prompts:
self._refine_template = prompts["refine_template"]
async def aevaluate(
self,
query: str | None = None,
response: str | None = None,
contexts: Sequence[str] | None = None,
sleep_time_in_seconds: int = 0,
**kwargs: Any,
) -> EvaluationResult:
"""Evaluate whether the response is faithful to the contexts."""
del kwargs # Unused
await asyncio.sleep(sleep_time_in_seconds)
if contexts is None or response is None:
raise ValueError("contexts and response must be provided")
docs = [Document(text=context) for context in contexts]
index = SummaryIndex.from_documents(docs)
query_engine = index.as_query_engine(
llm=self._llm,
text_qa_template=self._eval_template,
refine_template=self._refine_template,
)
response_obj = await query_engine.aquery(response)
raw_response_txt = str(response_obj)
if "yes" in raw_response_txt.lower():
passing = True
else:
passing = False
if self._raise_error:
raise ValueError("The response is invalid")
return EvaluationResult(
query=query,
response=response,
contexts=contexts,
passing=passing,
score=1.0 if passing else 0.0,
feedback=raw_response_txt,
)
|
