#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/llm-fine-tuning/dataset-preparation-curation/the-sharegpt-format

░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░█████████████████████████████████████████████████████████████████████████████████████████████

#####   ######  #####    ###    #   #  ###  #   #  ######
##  ##  ##      ##  ##  ## ##   #   #   #   #   #  ##
#####   ####    #####   #   #   #   #   #   #   #  ####
##  #   ##      ##      ## ##    # #    #    # #   ##
##   #  ######  ##       ###      #    ###    #    ######

$ curl repovive.com/roadmaps/llm-fine-tuning/dataset-preparation-curation/the-sharegpt-format

Repovive

The ShareGPT Format - Dataset Preparation & Curation | LLM Fine-Tuning | Repovive

Repovive.

LLM Fine-Tuning10 sections · 313 units1h total

Open in Course

The ShareGPT Format

ShareGPT format represents multi-turn conversations:

[
  {"from": "human", "value": "..."},
  {"from": "gpt", "value": "..."},
  {"from": "human", "value": "..."},
  {"from": "gpt", "value": "..."}
]

Use this when you're training conversational behavior. The model learns to continue the conversation pattern, not just answer single prompts.