persona-data 0.6.0

Shared dataset loading and prompt formatting for implicit-personalization projects

persona-data

Shared dataset loading, prompt formatting, and environment utilities for the implicit-personalization projects.

What's in the box

• SynthPersonaDataset — persona profiles plus QA pairs (docs)
• PersonaGuessDataset — turn-based persona games (docs)
• NemotronPersonasFranceDataset / NemotronPersonasUSADataset — NVIDIA persona-only datasets (docs)
• Roleplay and multiple-choice prompt helpers (docs)
• Environment helpers: set_seed, get_device, get_artifacts_dir

Installation

Add as a uv git source in your project's pyproject.toml:

[project]
dependencies = ["persona-data"]

[tool.uv.sources]
persona-data = { git = "ssh://[e-mail]/implicit-personalization/persona-data.git" }

For local development alongside other repos:

[tool.uv.sources]
persona-data = { path = "../persona-data", editable = true }

Then uv sync.

Testing

uv run --with pytest pytest tests/test_datasets.py

The release workflow also runs tests/smoke_test.py against the built wheel and source distribution.

Package layout

src/persona_data/
├── synth_persona.py       # SynthPersonaDataset, PersonaDataset, PersonaData, QAPair, Statement
├── persona_guess.py       # PersonaGuessDataset, GameRecord, Turn
├── nemotron_personas.py   # NemotronPersonasFranceDataset, NemotronPersonasUSADataset
├── prompts.py             # format_prompt, format_mc_question, format_messages
└── environment.py         # set_seed, get_device, get_artifacts_dir

Quick start

from persona_data.synth_persona import SynthPersonaDataset
from persona_data.prompts import format_messages, format_prompt

dataset = SynthPersonaDataset()
persona = dataset[0]

system_prompt = format_prompt(persona, "biography")
messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": "Where did you grow up?"},
]
full_prompt, response_start_idx = format_messages(
    messages, tokenizer, add_generation_prompt=True
)

# Leakage-aware train/test split: FRQs for train, shared MCQs for test.
train_qa, test_qa = dataset.train_test_split(persona.id)

See the docs for full APIs.

Used by

• persona-vectors — activation extraction and steering
• cues_attribution — section-level ablation attribution
• persona-2-lora — LoRA-based persona internalization