Instruction Tuning

Instruction tuning formats for supervised fine-tuning.

alpaca

instruction; input(optional)

data.jsonl

{"instruction": "...", "input": "...", "output": "..."}

jeopardy

question and answer

data.jsonl

{"question": "...", "category": "...", "answer": "..."}

oasst

instruction

data.jsonl

{"INSTRUCTION": "...", "RESPONSE": "..."}

gpteacher

instruction; input(optional)

data.jsonl

{"instruction": "...", "input": "...", "response": "..."}

reflection

instruction with reflect; input(optional)

data.jsonl

{"instruction": "...", "input": "...", "output": "...", "reflection": "...", "corrected": "..."}

explainchoice

question, choices, (solution OR explanation)

data.jsonl

{"question": "...", "choices": ["..."], "solution": "...", "explanation": "..."}

concisechoice

question, choices, (solution OR explanation)

data.jsonl

{"question": "...", "choices": ["..."], "solution": "...", "explanation": "..."}

summarizetldr

article and summary

data.jsonl

{"article": "...", "summary": "..."}

alpaca_chat

basic instruct for alpaca chat

data.jsonl

{"instruction": "...", "input": "...", "response": "..."}

alpaca_chat.load_qa

question and answer for alpaca chat

data.jsonl

{"question": "...", "answer": "..."}

alpaca_chat.load_concise

question and answer for alpaca chat, for concise answers

data.jsonl

{"instruction": "...", "input": "...", "response": "..."}

alpaca_chat.load_camel_ai

question and answer for alpaca chat, for load_camel_ai

data.jsonl

{"message_1": "...", "message_2": "..."}

alpaca_w_system.load_open_orca

support for open orca datasets with included system prompts, instruct

data.jsonl

{"system_prompt": "...", "question": "...", "response": "..."}

context_qa

in context question answering from an article

data.jsonl

{"article": "...", "question": "...", "answer": "..."}

context_qa.load_v2

in context question answering (alternate)

data.jsonl

{"context": "...", "question": "...", "answer": "..."}

context_qa.load_404

in context question answering from an article, with default response for no answer from context

data.jsonl

{"article": "...", "unanswerable_question": "..."}

creative_acr.load_answer

instruction and revision

data.jsonl

{"instruction": "...", "revision": "..."}

creative_acr.load_critique

critique

data.jsonl

{"scores": "...", "critiques": "...", "instruction": "...", "answer": "..."}

creative_acr.load_revise

critique and revise

data.jsonl

{"scores": "...", "critiques": "...", "instruction": "...", "answer": "...", "revision": "..."}

metharme

instruction, adds additional eos tokens

data.jsonl

{"prompt": "...", "generation": "..."}

How to add custom prompt format

For a dataset that is preprocessed for instruction purposes:

data.jsonl

{"input": "...", "output": "..."}

You can use this example in your YAML config:

config.yaml

datasets:
  - path: repo
    type:
      system_prompt: ""
      field_system: system
      field_instruction: input
      field_output: output
      format: "[INST] {instruction} [/INST]"
      no_input_format: "[INST] {instruction} [/INST]"

See full config options under here.