nemo_rl.data.datasets.response_datasets

Submodules

• nemo_rl.data.datasets.response_datasets.audiomcq
• nemo_rl.data.datasets.response_datasets.response_dataset
• nemo_rl.data.datasets.response_datasets.clevr
• nemo_rl.data.datasets.response_datasets.nemogym_dataset
• nemo_rl.data.datasets.response_datasets.squad
• nemo_rl.data.datasets.response_datasets.openmathinstruct2
• nemo_rl.data.datasets.response_datasets.tulu3
• nemo_rl.data.datasets.response_datasets.refcoco
• nemo_rl.data.datasets.response_datasets.aime
• nemo_rl.data.datasets.response_datasets.gsm8k
• nemo_rl.data.datasets.response_datasets.general_conversations_dataset
• nemo_rl.data.datasets.response_datasets.mmpr_tiny
• nemo_rl.data.datasets.response_datasets.helpsteer3
• nemo_rl.data.datasets.response_datasets.arrow_text_dataset
• nemo_rl.data.datasets.response_datasets.oai_format_dataset
• nemo_rl.data.datasets.response_datasets.daily_omni
• nemo_rl.data.datasets.response_datasets.intent
• nemo_rl.data.datasets.response_datasets.avqa
• nemo_rl.data.datasets.response_datasets.deepscaler
• nemo_rl.data.datasets.response_datasets.dapo_math
• nemo_rl.data.datasets.response_datasets.oasst
• nemo_rl.data.datasets.response_datasets.geometry3k
• nemo_rl.data.datasets.response_datasets.nemotron_cascade2_sft

Package Contents

Functions

load_response_dataset

Loads response dataset.

Data

DATASET_REGISTRY
__all__

API

nemo_rl.data.datasets.response_datasets.DATASET_REGISTRY

None

nemo_rl.data.datasets.response_datasets.load_response_dataset(data_config: nemo_rl.data.ResponseDatasetConfig)

Loads response dataset.

Resolution order for data_config["dataset_name"]:

1. If the name matches a key in DATASET_REGISTRY, use the built-in class.

2. Otherwise, if the name contains a ., treat it as a fully qualified dotted import path (e.g. my_pkg.my_module.MyDataset) and import the class dynamically. This lets users register custom datasets without editing nemo_rl.

3. Otherwise, raise ValueError with a helpful message.

nemo_rl.data.datasets.response_datasets.__all__

[‘AudioMCQDataset’, ‘ArrowTextDataset’, ‘AVQADataset’, ‘AIMEDataset’, ‘CLEVRCoGenTDataset’, ‘DailyOm…