Hugging Face datasets "trust_remote_code=True" required in CI

The dataset ships a Python loading script, and recent datasets versions refuse to execute repository code unless you explicitly pass trust_remote_code=True. In CI this fails hard because there is no interactive prompt to accept.

What this error means

load_dataset raises "ValueError: The repository for <name> contains custom code which must be executed to correctly load the dataset. ... Please pass the argument trust_remote_code=True to allow custom code to be run."

ValueError: The repository for my_dataset contains custom code which must be executed to
correctly load the dataset. You can inspect the repository content at https://hf.co/datasets/my_dataset.
Please pass the argument `trust_remote_code=True` to allow custom code to be run.

Common causes

The dataset uses a loading script

Script-based datasets run repository code to build splits; datasets now blocks that by default for safety.

No interactive confirmation is possible in CI

The prompt to allow code cannot be answered on a runner, so the call fails unless the flag is set in code.

How to fix it

Pass trust_remote_code after reviewing the script

1. Inspect the dataset repository script so you trust what it runs.
2. Pass trust_remote_code=True to load_dataset.
3. Pin the dataset revision so the code cannot change under you.

from datasets import load_dataset
load_dataset("my_dataset", trust_remote_code=True, revision="v1.0")

Prefer a script-free dataset

Where possible use a Parquet or data-file dataset that needs no code execution, avoiding the flag entirely.

load_dataset("parquet", data_files="data/*.parquet")

How to prevent it

• Review and pin the dataset revision before trusting its code.
• Prefer data-file datasets that require no trust_remote_code.
• Set the flag in code, not via an interactive prompt.

Related guides