You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
# Gets the top 10 labels by feedbacks to augment the context that the LLM has, instead of just asking it to generate categories without knowing the most common labels
# Gets the top labels by feedbacks to augment the context that the LLM has, instead of just asking it to generate categories without knowing the most common labels
# If the LLM just forgets or adds extra primary labels, log it but still generate categories
195
-
iflen(label_groups) !=len(top_10_labels):
200
+
iflen(label_groups) !=len(top_labels):
196
201
logger.warning(
197
202
"Number of label groups does not match number of primary labels passed in Seer",
198
203
extra={
199
204
"label_groups": label_groups,
200
-
"top_10_labels": top_10_labels,
205
+
"top_labels": top_labels,
201
206
},
202
207
)
203
208
204
209
# If the LLM hallucinates primary label(s), log it but still generate categories
205
210
forlabel_groupinlabel_groups:
206
-
iflabel_group["primaryLabel"] notintop_10_labels:
211
+
iflabel_group["primaryLabel"] notintop_labels:
207
212
logger.warning(
208
213
"LLM hallucinated primary label",
209
214
extra={"label_group": label_group},
210
215
)
211
216
212
217
# Converts label_groups (which maps primary label to associated labels) to a list of lists, where the first element is the primary label and the rest are the associated labels
0 commit comments