evo2-cli
A command-line tool for pulling DNA sequence embeddings out of Evo 2, the genomic foundation model. Point it at sequences, get back embedding vectors for downstream ML.
evo2-cli is a small command-line wrapper for extracting DNA embeddings from Evo 2, a large genomic foundation model. It turns “I have some sequences and want their learned representations” into a single command, instead of boilerplate model-loading code every time.
Why Embeddings
Foundation models like Evo 2 learn dense vector representations of DNA that capture sequence context far beyond k-mer counts. Those embeddings are the useful raw material for downstream tasks — variant effect prediction, sequence classification, clustering, similarity search — where you want the model’s representation as a feature rather than its raw generative output.
What It Does
The CLI handles the unglamorous parts: loading the model, batching sequences, running inference, and writing embedding vectors out in a form you can feed straight into a notebook or a training loop. The goal is to make Evo 2 embeddings a one-line building block in a bioinformatics workflow.
Status
Utility tool, portfolio project.