Evo 2: Genome modeling and design across all domains of life

Evo 2 is a state of the art DNA language model for long context modeling and design. Evo 2 models DNA sequences at single-nucleotide resolution at up to 1 million base pair context length using the StripedHyena 2 architecture. Evo 2 was pretrained using Savanna. Evo 2 was trained autoregressively on OpenGenome2, a dataset containing 8.8 trillion tokens from all domains of life.

We describe Evo 2 in our paper: "Genome modeling and design across all domains of life with Evo 2".

[!NOTE]

• Evo 2 published: read more in Nature.
• Evo 2 20B released: 40B-level performance with double the speed, read more here.
• Light install for 7B models: option compatible with more hardware, see Installation.

Contents

• Setup
  • Requirements
  • Installation
  • Docker
• Usage
  • Checkpoints
  • Forward
  • Embeddings
  • Generation
• Notebooks
• Nvidia NIM
• Dataset
• Training and Finetuning
• Citation

Setup

This repo is for running Evo 2 locally for inference or generation, using our Vortex inference code. For training and finetuning, see the section here. You can run Evo 2 without any installation using the Nvidia Hosted API. You can also self-host an instance using Nvidia NIM. See the Nvidia NIM section for more information.

Requirements

Evo 2 is built on the Vortex inference repo, see the Vortex github for more details and Docker option.

System requirements

• [OS] Linux (official) or WSL2 (limited support)
• [Software]
  • CUDA: 12.1+ with compatible NVIDIA drivers
  • cuDNN: 9.3+
  • Compiler: GCC 9+ or Clang 10+ with C++17 support
  • Python 3.11 or 3.12
• Recommended Torch 2.6.x or 2.7.x

FP8 and Transformer Engine requirements

The 40B, 20B, and 1B models require FP8 via Transformer Engine for numerical accuracy and a Nvidia Hopper GPU. The 7B models can run in bfloat16 without Transformer Engine on any supported GPU.

| Model | FP8 (Transformer Engine) Required | |-------|-----------------------------------| | evo2_7b / evo2_7b_262k / evo2_7b_base | No | | evo2_20b | Yes | | evo2_40b / evo2_40b_base | Yes | | evo2_1b_base | Yes |

Always validate model outputs after configuration changes or on different hardware by using the tests.

Installation

Full install

Install Transformer Engine and Flash Attention first, then install Evo 2. We recommend using conda to install Transformer Engine:

conda install -c nvidia cuda-nvcc cuda-cudart-dev
conda install -c conda-forge transformer-engine-torch=2.3.0
pip install flash-attn==2.8.0.post2 --no-build-isolation
pip install evo2

Light install (7B models only, no Transformer Engine)

Evo 2 7B models can run without Transformer Engine or FP8-capable hardware. If you run into issues installing Flash Attention, see the Flash Attention GitHub for system requirements and troubleshooting.

# A compatible PyTorch must be installed before flash attention, for example: pip install torch==2.7.1 --index-url https://download.pytorch.org/whl/cu128
pip install flash-attn==2.8.0.post2 --no-build-isolation
pip install evo2

From source

git clone https://github.com/arcinstitute/evo2
cd evo2
pip install -e .

Verify installation

python -m evo2.test.test_evo2_generation --model_name evo2_7b  # or evo2_1b_base, evo2_20b, evo2_40b

Docker

Evo 2 can be run using Docker (shown below), Singularity, or Apptainer.

docker build -t evo2 .
docker run -it --rm --gpus '"device=0"' -v ./huggingface:/root/.cache/huggingface evo2 bash

Note: The volume mount (-v) preserves downloaded models between container runs and specifies where they are saved.

Once inside the container:

python -m evo2.test.test_evo2_generation --model_name evo2_7b

Usage

Checkpoints

We provide the following model checkpoints, hosted on HuggingFace: | Checkpoint Name | Description | |----------------------------------------|-------------| | evo2_40b | 40B parameter model with 1M context | | evo2_20b | 20B parameter model with 1M context | | evo2_7b | 7B parameter model with 1M context | | evo2_40b_base | 40B parameter model with 8K context | | evo2_7b_base | 7B parameter model with 8K context | | evo2_1b_base | Smaller 1B parameter model with 8K context | | evo2_7b_262k | 7B parameter model with 262K context | | evo2_7b_microviridae | 7B parameter base model fine-tuned on Microviridae genomes |

Note: The 40B model requires multiple H100 GPUs. Vortex automatically handles device placement, splitting the model across available CUDA devices.

Forward

Evo 2 can be used to score the likelihoods across a DNA sequence.

import torch
from evo2 import Evo2

evo2_model = Evo2('evo2_7b')

sequence = 'ACGT'
input_ids = torch.tensor(
    evo2_model.tokenizer.tokenize(sequence),
    dtype=torch.int,
).unsqueeze(0).to('cuda:0')

outputs, _ = evo2_model(input_ids)
logits = outputs[0]

print('Logits: ', logits)
print('Shape (batch, length, vocab): ', logits.shape)

Embeddings

Evo 2 embeddings can be saved for use downstream. We find that intermediate embeddings work better than final embeddings, see our paper for details.

import torch
from evo2 import Evo2

evo2_model = Evo2('evo2_7b')

sequence = 'ACGT'
input_ids = torch.tensor(
    evo2_model.tokenizer.tokenize(sequence),
    dtype=torch.int,
).unsqueeze(0).to('cuda:0')

layer_name = 'blocks.28.mlp.l3'

outputs, embeddings = evo2_model(input_ids, return_embeddings=True, layer_names=[layer_name])

print('Embeddings shape: ', embeddings[layer_name].shape)

Generation

Evo 2 can generate DNA sequences based on prompts.

from evo2 import Evo2

evo2_model = Evo2('evo2_7b')

output = evo2_model.generate(prompt_seqs=["ACGT"], n_tokens=400, temperature=1.0, top_k=4)

print(output.sequences[0])

Notebooks

We provide example notebooks.

The BRCA1 scoring notebook shows zero-shot BRCA1 variant effect prediction. This example includes a walkthrough of:

• Performing zero-shot BRCA1 variant effect predictions using Evo 2
• Reference vs alternative allele normalization

The generation notebook shows DNA sequence completion with Evo 2. This example shows:

• DNA prompt based generation and 'DNA autocompletion'
• How to get and prompt using phylogenetic species tags for generation

The exon classifier notebook demonstrates exon classification using Evo 2 embeddings. This example shows:

• Running the Evo 2 based exon classifier
• Performance metrics and visualization

The sparse autoencoder (SAE) notebook explores interpretable features learned by Evo 2. This example includes:

• Running and visualizing Evo 2 SAE features
• Demonstrating SAE features on a part of the E. coli genome

Nvidia NIM

Evo 2 is available on Nvidia NIM and hosted API.

• Documentation
• Quickstart

The quickstart guides users through running Evo 2 on the NVIDIA NIM using a python or shell client after starting NIM. An example python client script is shown below. This is the same way you would interact with the Nvidia hosted API.

#!/usr/bin/env python3
import requests
import os
import json
from pathlib import Path

key = os.getenv("NVCF_RUN_KEY") or input("Paste the Run Key: ")

r = requests.post(
    url=os.getenv("URL", "https://health.api.nvidia.com/v1/biology/arc/evo2-40b/generate"),
    headers={"Authorization": f"Bearer {key}"},
    json={
        "sequence": "ACTGACTGACTGACTG",
        "num_tokens": 8,
        "top_k": 1,
        "enable_sampled_probs": True,
    },
)

if "application/json" in r.headers.get("Content-Type", ""):
    print(r, "Saving to output.json:\n", r.text[:200], "...")
    Path("output.json").write_text(r.text)
elif "application/zip" in r.headers.get("Content-Type", ""):
    print(r, "Saving large response to data.zip")
    Path("data.zip").write_bytes(r.content)
else:
    print(r, r.headers, r.content)

Very long sequences

You can use Savanna or Nvidia BioNemo for embedding long sequences. Vortex can currently compute over very long sequences via teacher prompting, however please note that forward pas