GenerateTextEmbeddingDataPromptTemplateDict
TextPrompt
dictionary containing text embedding tasks
generation, query, positive and hard negative samples generation,
from the “Improving Text Embeddings with Large Language Models” paper.
Parameters:
- GENERATE_TASKS (TextPrompt): A prompt to generate a list
- of: obj:
num_tasks
synthetic text_embedding tasks. - ASSISTANT_PROMPT (TextPrompt): A system prompt for the AI assistant to generate synthetic :obj:
user_query
, :obj:positive document
, - and: obj:
hard_negative_document
for a specific :obj:task
with specified parameters including :obj:query_type
, :obj:query_length
, :obj:clarity
, :obj:num_words
, :obj:language
and :obj:difficulty
.