Tiny Opengenome2: A Subset of 1M+ DNA Sequences for ML

A curated subset of the OpenGenome2 dataset contains over 1 million DNA sequences with over 185 billion base pair tokens across 16 biological categories. It is designed to replicate the distribution used to train the Evo2 model but with fewer examples. The dataset was created by andrewdalpino and was last updated on April 20, 2026.

Use Cases

• Knowledge distillation for genomic language models based on the dataset's replication of the Evo2 training distribution.
• Rapid prototyping of sequence-based machine learning models using a smaller, more manageable subset of data.
• Academic research and benchmarking in computational genomics leveraging the 16 broad biological categories.

Strengths

• Contains over 1 million DNA sequences.
• Includes over 185 billion base pair tokens.
• Covers a broad spectrum of biological life across 16 categories.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Curated from the OpenGenome2 dataset.

Freshness

Last updated 2026-04-20 23:04:57; freshness should be verified.