Roland Tannous e087a4a72b fix: improve Pass 2 prompt to correctly split INPUT/OUTPUT columns ... The LLM was putting all columns in user_template (e.g. summarization dataset had both document AND summary as user input). Fixed by: - Reframed system message: explicitly states user=INPUT, assistant=OUTPUT - Added 4 concrete correct examples (summarization, NLI, translation, QA) showing exactly how to split columns - Added "NEVER put the output/target column in the user template" rule - Added sanity check: if assistant_template has no column placeholders, reject the result and fall back to simple classification		2026-03-10 16:47:50 +00:00
..
__init__.py	feat: add ShareGPT+image VLM format support and improve image column detection	2026-03-10 01:27:36 +00:00
chat_templates.py	Add AGPL-3.0 SPDX headers to all source files	2026-03-09 20:17:45 +00:00
data_collators.py	Add AGPL-3.0 SPDX headers to all source files	2026-03-09 20:17:45 +00:00
dataset_utils.py	fix: harden template mapping for complex column types and curly braces	2026-03-10 15:43:35 +00:00
format_conversion.py	feat: add LLM-assisted dataset detection using ephemeral GGUF helper	2026-03-10 09:20:45 +00:00
format_detection.py	fix: probe image column candidates when multiple exist	2026-03-10 01:38:33 +00:00
llm_assist.py	fix: improve Pass 2 prompt to correctly split INPUT/OUTPUT columns	2026-03-10 16:47:50 +00:00
model_mappings.py	Add AGPL-3.0 SPDX headers to all source files	2026-03-09 20:17:45 +00:00
vlm_processing.py	debug: switch to print() for subprocess visibility	2026-03-10 12:49:01 +00:00