Dust - spolu > toolbot-repro

Teach LLM to teach itself new tasks by teaching it to generate few shot examples from high-level tasks descriptions and prompting itself with them (a reproduction of @chillzaza_'s ToolBot).

input

dataset:

dev

data

dataset:

examples

code

code :

llm

model:

openai

code-davinci-002

temperature:

max tokens:

stop:

INSTRUCTION

advanced

few-shot

introduction:

Generate 4 QUERY/RESPONSE examples for each of the high-level INSTRUCTION provided:

examples:

INSTRUCTION: ${PP_EXAMPLES.instruction}
${PP_EXAMPLES.examples}

count:

prompt:

code

code :

code

code :

map

from:

repeat:

llm

model:

openai

code-davinci-002

temperature:

max tokens:

stop:

QUERY

advanced

prompt:

code

code :

reduce