From 82f969d666f8085de426b798bd6a97bfc990d7d2 Mon Sep 17 00:00:00 2001 From: extio-de Date: Sun, 3 Nov 2024 22:53:35 +0100 Subject: [PATCH] Fix paragraphToContextualPrompt response post-processing; --- .../preparer/Text2Jsonl2WithContextualPrompts.java | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/src/main/java/de/extio/lmdatasetprep/preparer/Text2Jsonl2WithContextualPrompts.java b/src/main/java/de/extio/lmdatasetprep/preparer/Text2Jsonl2WithContextualPrompts.java index db03fd0..52135b8 100644 --- a/src/main/java/de/extio/lmdatasetprep/preparer/Text2Jsonl2WithContextualPrompts.java +++ b/src/main/java/de/extio/lmdatasetprep/preparer/Text2Jsonl2WithContextualPrompts.java @@ -102,9 +102,13 @@ public class Text2Jsonl2WithContextualPrompts implements Consumer { String prompt = completion.response(); final int colon = prompt.indexOf(':'); - if (colon > -1 && colon > prompt.length() + 1) { + if (colon > -1 && colon < prompt.length() + 1) { prompt = prompt.substring(colon + 1); } + final int preamble = prompt.indexOf("\n\n"); + if (preamble > -1 && preamble < prompt.length() + 2) { + prompt = prompt.substring(preamble + 2); + } prompt = prompt.trim(); final var qaLine = new QaLine(prompt, paragraph);