Training Models

All actively supported model training workflows live in bionemo-recipes.

Where to look

Use the README in the relevant recipe or model directory as the source of truth for setup and execution:

Workflow	Location
ESM-2 native PyTorch + TE	bionemo-recipes/recipes/esm2_native_te
ESM-2 with Accelerate + TE	bionemo-recipes/recipes/esm2_accelerate_te
Geneformer + TE	bionemo-recipes/recipes/geneformer_native_te_mfsdp_fp8
Evo2 (Megatron Bridge)	bionemo-recipes/recipes/evo2_megatron
TE-optimized AMPLIFY model	bionemo-recipes/models/amplify
TE-optimized ESM-2 model	bionemo-recipes/models/esm2
TE-optimized Geneformer model	bionemo-recipes/models/geneformer

Local workflow

For most training workflows:

1. Open the recipes devcontainer or use a compatible CUDA/PyTorch environment.
2. cd into the model or recipe directory you want to work on.
3. Install dependencies according to that directory's README.
4. Run the documented training, fine-tuning, or notebook workflow from that directory.

Examples:

cd bionemo-recipes/recipes/esm2_native_te
pip install -r requirements.txt
python train_ddp.py

cd bionemo-recipes/recipes/evo2_megatron
bash .ci_build.sh
source ./.ci_test_env.sh
train_evo2 --help

Shared framework packages

Some recipes depend on reusable libraries under sub-packages. Install them into your active environment with editable installs when developing locally:

uv pip install -e ./sub-packages/bionemo-core
uv pip install -e ./sub-packages/bionemo-scdl
uv pip install -e "./sub-packages/bionemo-recipeutils[basecamp]"

Data and checkpoints

The download_bionemo_data CLI remains the standard way to fetch supported BioNeMo datasets and checkpoints:

download_bionemo_data --list-resources

Set DATA_SOURCE=ngc for public resources, or DATA_SOURCE=pbss for internal NVIDIA workflows where applicable.