all publishers

nvidia-bionemo

@nvidia-bionemo source repo

93 published skills

  1. ▌
    Complexa Setup 3 · nvidia-bionemo bundle
    Agent runbook for first-time Proteina-Complexa setup, dotenv configuration, runtime activation, model-weight installation, GPU preflight, and environment validation. Use for "set up complexa", "install complexa", "complexa init", "complexa download", checkpoint installation or status, UV versus Docker selection, and fresh checkouts. SKILL.md invokes the first-party CLI; bundled executables safely inspect local readiness and write a private setup manifest.
    508 repo stars
  2. ▌
    Complexa Sweep 3 · nvidia-bionemo bundle
    Agent runbook for Proteina-Complexa parameter sweeps through the separately installed first-party CLI and config generator. Use for cartesian-product hyperparameter scans, sweeper YAML authoring, configuration comparisons, ablations, tuning, success-rate ranking, or Pareto searches over quality and wall-clock cost. SKILL.md performs the orchestration; bundled executables only inspect local readiness and write a run manifest.
    508 repo stars
  3. ▌
    Complexa Design 3 · nvidia-bionemo bundle
    Agent runbook for orchestrating an end-to-end protein, ligand-binder, or AME design through the separately installed first-party Proteina-Complexa CLI. Use for requests such as "design a binder", "run complexa design", "de novo binder", ligand binding, motif scaffolding, beam search, FK steering, MCTS, refolding, or reporting design success and diversity. SKILL.md selects and invokes the project CLI; bundled executables only inspect local readiness and write a run manifest, and do not implement model inference themselves.
    508 repo stars
  4. ▌
    Complexa Target 3 · nvidia-bionemo bundle
    Use this skill whenever the user wants to add, register, edit, list, show, or validate a Proteina-Complexa design target for any pipeline — protein binder (default), ligand binder, or AME / enzyme scaffolding. Triggers include "add a target", "define a new target for binder design", "register a hotspot", "set up a PDL1 binder target", "ligand binder pocket", "SMILES target", "AME task", "enzyme motif", "M0024_1nzy", "complexa target add", "complexa target show", "configure target X", "what targets are available", "where do hotspots live", "what does target_input mean", "chain-spec syntax", "binder length range", or any question about `configs/targets/{,ligand_}targets_dict.yaml` and `configs/design_tasks/ame_dict_v2.yaml`. Also covers `complexa validate target`. This is the only skill that touches the three targets dict files.
    508 repo stars
  5. ▌
    Complexa Evaluate Pdbs 3 · nvidia-bionemo bundle
    Agent runbook for evaluating an existing PDB directory through the separately installed first-party Proteina-Complexa CLI. Use to refold or score designs, compute interface pAE, i_pLDDT, scRMSD, motif RMSD, or designability, run complexa analysis, or assess third-party BindCraft, AlphaProteo, or RFdiffusion outputs. SKILL.md selects and invokes the evaluation configuration; bundled executables only inspect local readiness and write a run manifest.
    508 repo stars
  6. ▌
    Cuequivariance 2 · nvidia-bionemo bundle
    Define custom groups (Irrep subclasses), build segmented tensor products with CG coefficients, create equivariant polynomials and IrDictPolynomials, and use built-in descriptors (linear, tensor products, spherical harmonics). Use when working with cuequivariance group theory, irreps, or segmented polynomials.
    508 repo stars
  7. ▌
    Kermt Embed 2 · nvidia-bionemo bundle
    Extract per-molecule embeddings from any encoder-bearing KERMT checkpoint (grover_base / cmim / hybrid / finetuned). Writes one .npy per readout type (atom_from_atom, bond_from_atom, atom_from_bond, bond_from_bond) plus canonical_smiles.npy and validity.npy. Calls task/extract_embeddings.py (which featurizes SMILES on the fly — no pre-computed features needed).
    508 repo stars
  8. ▌
    Kermt Infer 2 · nvidia-bionemo bundle
    Run predictions with a finetuned KERMT checkpoint on a SMILES-only CSV. The skill validates that the input ckpt has task FFN heads (refuses pretrain ckpts with a redirect to kermt-finetune), validates the CSV, prepares the data (clean + rdkit_2d features), then launches main.py predict inside the kermt container (blocking, minutes-scale).
    508 repo stars
  9. ▌
    Kermt Setup 2 · nvidia-bionemo bundle
    Bootstrap the KERMT agent environment — verify host docker + nvidia-container-toolkit, build the kermt:latest image from the repo's Dockerfile if it doesn't yet exist, and run a GPU smoke test inside the container. Every other kermt-* skill depends on this; invoke it first.
    508 repo stars
  10. ▌
    Kermt Monitor 2 · nvidia-bionemo bundle
    Check progress for a detached KERMT run (pretrain, finetune, or any kermt_run_detached invocation). Reads run.json, queries docker for container state, tails the pretrain/finetune log, and parses progress lines (epoch, step, val loss).
    508 repo stars
  11. ▌
    Kermt Finetune 2 · nvidia-bionemo bundle
    Finetune a pretrained KERMT encoder on a labeled CSV. The skill validates the input checkpoint (must be a pretrain ckpt — grover_base / cmim / hybrid), validates the labeled CSV, prepares the data (clean + features + optional split), then launches main.py finetune inside the kermt container (detached for hours-scale runs). Hyperparameters come from agent/config/defaults_finetune.json with per-flag CLI override.
    508 repo stars
  12. ▌
    Evo2 Nim 2 · nvidia-bionemo bundle
    Generate and analyze DNA sequences using NVIDIA's Evo 2 BioNeMo NIM microservice. Use for Evo2/Evo 2, DNA generation, genomic sequence generation, hosted generation, local Docker deployment, local forward passes, layer outputs, logits, sampled probabilities, and BioNeMo NIM workflows.
    508 repo stars
  13. ▌
    Boltz2 Nim 2 · nvidia-bionemo bundle
    Use Boltz2 NIM for biomolecular structure prediction and binding affinity. Invoke for Boltz2, protein structures, protein-ligand/DNA/RNA complexes, SMILES or CCD ligands, pIC50/IC50 affinity scoring, mmCIF output, hosted NVIDIA API calls, or local Docker deployment.
    508 repo stars
  14. ▌
    Genmol Nim 2 · nvidia-bionemo bundle
    Generate novel drug-like molecules using the GenMol NIM microservice. Use for de novo generation, scaffold decoration, motif extension, lead optimization, SAFE notation, QED or LogP ranking, hosted NVIDIA API calls, or local Docker deployment. GenMol takes SAFE notation in the smiles field, not ordinary SMILES.
    508 repo stars
  15. ▌
    Molmim Nim 2 · nvidia-bionemo bundle
    Use this skill for MolMIM, NVIDIA's BioNeMo NIM microservice for small-molecule latent-space generation and optimization. Invoke for MolMIM, molecular embeddings, hidden states, latent decoding, sampling around a seed SMILES, CMA-ES guided molecule generation, QED or plogP optimization, hosted NVIDIA API calls, or local Docker deployment.
    508 repo stars
  16. ▌
    Parabricks 2 · nvidia-bionemo bundle
    Route NVIDIA Parabricks pbrun tools, assess GPU/runtime readiness, and provide version-aware command guidance for FASTQ/BAM processing, RNA-seq, variant calling, BAM QC, and GVCF workflows. Do NOT use for inspecting or accelerating whole pipelines — use genomics-workflow-acceleration.
    508 repo stars
  17. ▌
    Kermt Pretrain Scratch 2 · nvidia-bionemo bundle
    Pretrain a fresh KERMT model from scratch on a user-provided corpus. Builds a new vocabulary from the corpus, instantiates the model architecture from defaults, and launches pretrain_ddp.py inside the kermt container (detached for long runs). Unlike kermt-continue-pretrain, no starting checkpoint is loaded — the model is randomly initialized.
    508 repo stars
  18. ▌
    Kermt Add Cmim Pretrain 2 · nvidia-bionemo bundle
    Convert a grover_base checkpoint (encoder-only or encoder + vocab heads) into a hybrid checkpoint by adding a randomly-initialized cMIM decoder + latent_dist, then continue pretraining on the user's corpus as hybrid (vocab + contrast). Effectively kermt-continue-pretrain with a one-time ckpt-conversion step prepended.
    508 repo stars
  19. ▌
    Kermt Continue Pretrain 2 · nvidia-bionemo bundle
    Continue pretraining from an existing KERMT checkpoint. The skill validates the user's checkpoint and pretrain CSV, prepares the data into shard/vocab/features form, then launches pretrain_ddp.py inside the kermt container (detached for long runs). Auto-dispatches `--pretrain_mode` based on the checkpoint type (grover_base vocab-only, cmim, or hybrid).
    508 repo stars
  20. ▌
    Diffdock Nim 2 · nvidia-bionemo bundle
    Run DiffDock molecular docking via NVIDIA NIM to predict small-molecule binding poses against protein targets. Use for DiffDock, molecular docking, ligand docking, blind docking, SMILES or SDF ligands, ranked poses, confidence scores, hosted NVIDIA API, or local Docker deployment.
    508 repo stars
  21. ▌
    Openfold2 Nim 2 · nvidia-bionemo bundle
    Use this skill for OpenFold2, NVIDIA's BioNeMo NIM microservice for monomer protein structure prediction. Invoke whenever the user mentions OpenFold2, AlphaFold2-like monomer folding, protein sequence-to-structure prediction, A3M MSAs, mmCIF templates, hosted NVIDIA API calls, or local Docker deployment.
    508 repo stars
  22. ▌
    Openfold3 Nim 2 · nvidia-bionemo bundle
    Use this skill for OpenFold3, NVIDIA's BioNeMo NIM microservice for biomolecular structure prediction. Invoke whenever the user mentions OpenFold3 or needs protein, protein-ligand, protein-DNA/RNA, or multi-chain complex prediction with the hosted NVIDIA API or local Docker NIM. Covers endpoint choice, auth, request payloads, output artifacts, confidence scores, and local container setup.
    508 repo stars
  23. ▌
    Msa Search Nim 2 · nvidia-bionemo bundle
    Generate multiple sequence alignments (MSAs) for protein sequences using the ColabFold MSA-Search NIM. Use for homolog search, UniRef30/ColabFold env searches, A3M or FASTA alignments, paired MSA search for complexes, PDB70 structural templates, hosted NVIDIA API calls, or local Docker deployment. For local deployment, download the databases in parallel with aria2c and launch via NIM_MODEL_NAME (the recommended default fast path, ~14 min vs over 80 min for the built-in downloader); a plain docker run uses the slow built-in downloader.
    508 repo stars
  24. ▌
    Nvmolkit Usage 2 · nvidia-bionemo bundle
    Write code that calls the installed nvMolKit Python API for GPU-accelerated, batched RDKit-style operations - Morgan fingerprints, Tanimoto/cosine similarity, ETKDG conformer embedding, MMFF/UFF optimization, TFD, conformer RMSD, Butina clustering, and substructure search. Use when the user is importing `nvmolkit.*`, debugging an `nvmolkit` call, choosing between nvMolKit and RDKit for a batched cheminformatics workflow, or wiring nvMolKit results into a torch/numpy pipeline. Out of scope: building nvMolKit from source.
    508 repo stars
  25. ▌
    Complexa Setup 2 · nvidia-bionemo bundle
    First-time setup, environment configuration, and model-weight installation for Proteina-Complexa. Reach for this skill whenever the user says "set up complexa", "install complexa", "configure my environment file", "first-time setup", "what models do I have installed", "what's in my environment file", "download model weights", "download Complexa / AF2 / RF3 / ProteinMPNN / LigandMPNN / ESM2 / ESMFold checkpoints", "preflight my GPU", "verify environment", "complexa init", "complexa download", "complexa download --status", "complexa validate env", or any time a fresh checkout needs to be made runnable. This is the first skill to run on a new clone — it drives `complexa init`, `complexa download`, and `complexa validate env` end-to-end, edits the required `.env` keys, picks the right runtime (UV vs Docker), and emits a replayable setup artifact.
    508 repo stars
  26. ▌
    Complexa Sweep 2 · nvidia-bionemo bundle
    Use this skill whenever the user wants to run a parameter sweep over a Proteina-Complexa design pipeline — cartesian-product hyperparameter scans, Pareto search over generation/reward/evaluation knobs, or any "compare configurations" workflow. Trigger phrases include "sweep beam width", "sweep nsteps", "hyperparameter sweep", "parameter scan", "scan beam_width and temperature", "compare configurations", "find the best generation params", "what's the optimal nsteps", "Pareto search for binder quality vs wall-clock", "complexa sweep", "tune Complexa", "ablate the reward weights", "configs/sweeps", "--sweeper", "run beam_width.yaml". This is the only skill that owns sweeper YAML authoring, cartesian-product expansion, and per-config result ranking.
    508 repo stars
  27. ▌
    Proteinmpnn Nim 2 · nvidia-bionemo bundle
    Run ProteinMPNN inverse folding via NVIDIA NIM to design protein sequences for a target backbone. Use for ProteinMPNN, inverse folding, sequence design, backbone redesign, fixed chains/residues, omit_AAs, sampling temperature, soluble model, hosted NVIDIA API, local Docker, PDB input, and multi-FASTA output.
    508 repo stars
  28. ▌
    Rfdiffusion Nim 2 · nvidia-bionemo bundle
    Run RFDiffusion protein backbone design via NVIDIA NIM. Use for de novo protein backbones, motif scaffolding, binder design, hotspot residues, contigs syntax, diffusion steps, hosted NVIDIA API calls, local Docker deployment, and PDB backbone outputs for ProteinMPNN sequence design.
    508 repo stars
  29. ▌
    Complexa Design 2 · nvidia-bionemo bundle
    End-to-end Proteina-Complexa design pipeline driver. Use when the user wants to "design a binder", "design binders for X", "run complexa design", "de novo binder", "PDL1 binder", "TrkA binder", "design proteins for target", "protein binder design", "ligand binder", "design a small-molecule binder", "ATP-binding protein", "AME motif scaffolding", "scaffold a motif near a ligand", "motif + ligand design", "enzyme scaffolding", "flow matching protein design", "beam-search binder", "FK steering", "MCTS protein design", or "refold with AF2 / RF3 / ESMFold", or wants success rates, interface pAE, scRMSD, or FoldSeek diversity from one command. Drives the complexa design pipeline from target picking to manifest emission and reports how many designs passed.
    508 repo stars
  30. ▌
    Complexa Target 2 · nvidia-bionemo bundle
    Use this skill whenever the user wants to add, register, edit, list, show, or validate a Proteina-Complexa design target for any pipeline — protein binder (default), ligand binder, or AME / enzyme scaffolding. Triggers include "add a target", "define a new target for binder design", "register a hotspot", "set up a PDL1 binder target", "ligand binder pocket", "SMILES target", "AME task", "enzyme motif", "M0024_1nzy", "complexa target add", "complexa target show", "configure target X", "what targets are available", "where do hotspots live", "what does target_input mean", "chain-spec syntax", "binder length range", or any question about `configs/targets/{,ligand_}targets_dict.yaml` and `configs/design_tasks/ame_dict_v2.yaml`. Also covers `complexa validate target`. This is the only skill that touches the three targets dict files.
    508 repo stars
  31. ▌
    Complexa Evaluate Pdbs 2 · nvidia-bionemo bundle
    Standalone evaluation of an existing PDB directory with Proteina-Complexa. Use when the user wants to "evaluate PDB files", "re-fold these designs", "compute interface pAE", "compute i_pLDDT for a folder", "run AF2 / RF3 / ESMFold on my designs", "score binder candidates", "designability of this folder", "scRMSD for designs", "motif RMSD for these PDBs", "complexa analysis", "complexa evaluate from a PDB directory", "evaluate from pdb dir", or to score third-party outputs (BindCraft, AlphaProteo, RFdiffusion, hand-curated decoys). Picks the correct evaluate_*.yaml config, wires ++dataset.pdb_dir and the folding backend, runs the evaluate → analyze chain, parses the result CSV, and reports pass-rates against the right result_type thresholds.
    508 repo stars
  32. ▌
    Drug Discovery Pipeline 2 · nvidia-bionemo bundle
    NOTE: molecule and target inputs and your NGC_API_KEY are transmitted to external NVIDIA-hosted API endpoints on every call. Use local NIM containers for confidential or proprietary data. Run a complete computational drug discovery pipeline using NVIDIA BioNeMo NIMs: generate drug-like molecules with GenMol, dock them to a protein target with DiffDock, then predict binding affinity with Boltz2. Use this skill whenever the user wants to generate and screen small molecule drug candidates, perform hit discovery, optimize leads against a protein target, or do virtual screening combining molecule generation, docking, and affinity prediction. Triggers on: drug discovery pipeline, hit discovery, lead optimization, virtual screening, molecule generation, molecular docking, binding affinity, GenMol, DiffDock, Boltz2, SMILES, SAFE notation, NIM microservice. This is a multi-step pipeline composing three BioNeMo NIMs.
    508 repo stars
  33. ▌
    Protein Binder Design 2 · nvidia-bionemo bundle
    Orchestrate an end-to-end de novo protein binder design campaign against a protein target by composing BioNeMo NIM skills. Use for binder design, minibinder design, de novo binders, RFdiffusion + ProteinMPNN + Boltz2/OpenFold3 pipelines, epitope/hotspot-targeted design, in-silico binder validation, and ranking designs by interface confidence.
    508 repo stars
  34. ▌
    Complexa Binder Design 2 · nvidia-bionemo bundle
    Run a complete protein binder design campaign with NVIDIA Proteina-Complexa: resolve a target structure and hotspots from a name/sequence/PDB, co-design binder sequence+structure with reward-guided test-time search (best-of-n, beam search, FK steering, MCTS), select with the internal AF2 reward gate, then INDEPENDENTLY validate each binder by refolding the complex with Boltz2 (default) or OpenFold3 and rank on interface confidence, pLDDT, ipSAE, apo/holo stability, and hotspot contact. Use whenever the user wants de novo binders against a named target, sequence, or PDB, hotspot/epitope-targeted design, Proteina-Complexa / Complexa, or ranked validated binders from one request. Sibling of protein-binder-design (RFdiffusion + ProteinMPNN); this skill uses Proteina-Complexa.
    508 repo stars
  35. ▌
    Genomics Workflow Acceleration 2 · nvidia-bionemo bundle
    Use when accelerating existing genomics workflows with NVIDIA Parabricks, improving runtime or price/performance, converting pipeline steps to GPUs, or comparing CPU and GPU workflow outputs. Adds optional GPU steps in-place with runtime toggles (default off). Do NOT use for individual pbrun command routing — use parabricks.
    508 repo stars
  36. ▌
    Msa Structure Prediction Pipeline 2 · nvidia-bionemo bundle
    NOTE: your protein sequence and the retrieved MSA alignment are transmitted to external NVIDIA-hosted APIs (health.api.nvidia.com) on every call. Use local NIM containers for confidential or proprietary sequences. Run a complete protein structure prediction pipeline using NVIDIA BioNeMo NIMs: search for MSA alignments with MSA-Search (ColabFold), then predict the structure with OpenFold3 using the retrieved alignments. Use this skill whenever the user wants to predict a protein structure with maximum accuracy using MSA context, run the full AlphaFold3-style pipeline, generate MSA-informed structure predictions, or improve structure prediction accuracy by providing evolutionary information. Triggers on: MSA structure prediction pipeline, structure prediction pipeline, MSA-informed prediction, OpenFold3, ColabFold MSA, AlphaFold3 pipeline, protein structure, homology search, a3m alignment, UniRef30, NIM microservice. This pipeline chains MSA-Search and OpenFold3.
    508 repo stars
  37. ▌
    Cuequivariance · nvidia-bionemo bundle
    Define custom groups (Irrep subclasses), build segmented tensor products with CG coefficients, create equivariant polynomials and IrDictPolynomials, and use built-in descriptors (linear, tensor products, spherical harmonics). Use when working with cuequivariance group theory, irreps, or segmented polynomials.
    508 repo stars
  38. ▌
    Kermt Embed · nvidia-bionemo bundle
    Extract per-molecule embeddings from any encoder-bearing KERMT checkpoint (grover_base / cmim / hybrid / finetuned). Writes one .npy per readout type (atom_from_atom, bond_from_atom, atom_from_bond, bond_from_bond) plus canonical_smiles.npy and validity.npy. Calls task/extract_embeddings.py (which featurizes SMILES on the fly — no pre-computed features needed).
    508 repo stars
  39. ▌
    Kermt Infer · nvidia-bionemo bundle
    Run predictions with a finetuned KERMT checkpoint on a SMILES-only CSV. The skill validates that the input ckpt has task FFN heads (refuses pretrain ckpts with a redirect to kermt-finetune), validates the CSV, prepares the data (clean + rdkit_2d features), then launches main.py predict inside the kermt container (blocking, minutes-scale).
    508 repo stars
  40. ▌
    Kermt Setup · nvidia-bionemo bundle
    Bootstrap the KERMT agent environment — verify host docker + nvidia-container-toolkit, build the kermt:latest image from the repo's Dockerfile if it doesn't yet exist, and run a GPU smoke test inside the container. Every other kermt-* skill depends on this; invoke it first.
    508 repo stars
  41. ▌
    Kermt Monitor · nvidia-bionemo bundle
    Check progress for a detached KERMT run (pretrain, finetune, or any kermt_run_detached invocation). Reads run.json, queries docker for container state, tails the pretrain/finetune log, and parses progress lines (epoch, step, val loss).
    508 repo stars
  42. ▌
    Kermt Finetune · nvidia-bionemo bundle
    Finetune a pretrained KERMT encoder on a labeled CSV. The skill validates the input checkpoint (must be a pretrain ckpt — grover_base / cmim / hybrid), validates the labeled CSV, prepares the data (clean + features + optional split), then launches main.py finetune inside the kermt container (detached for hours-scale runs). Hyperparameters come from agent/config/defaults_finetune.json with per-flag CLI override.
    508 repo stars
  43. ▌
    Kermt Pretrain Scratch · nvidia-bionemo bundle
    Pretrain a fresh KERMT model from scratch on a user-provided corpus. Builds a new vocabulary from the corpus, instantiates the model architecture from defaults, and launches pretrain_ddp.py inside the kermt container (detached for long runs). Unlike kermt-continue-pretrain, no starting checkpoint is loaded — the model is randomly initialized.
    508 repo stars
  44. ▌
    Kermt Add Cmim Pretrain · nvidia-bionemo bundle
    Convert a grover_base checkpoint (encoder-only or encoder + vocab heads) into a hybrid checkpoint by adding a randomly-initialized cMIM decoder + latent_dist, then continue pretraining on the user's corpus as hybrid (vocab + contrast). Effectively kermt-continue-pretrain with a one-time ckpt-conversion step prepended.
    508 repo stars
  45. ▌
    Kermt Continue Pretrain · nvidia-bionemo bundle
    Continue pretraining from an existing KERMT checkpoint. The skill validates the user's checkpoint and pretrain CSV, prepares the data into shard/vocab/features form, then launches pretrain_ddp.py inside the kermt container (detached for long runs). Auto-dispatches `--pretrain_mode` based on the checkpoint type (grover_base vocab-only, cmim, or hybrid).
    508 repo stars
  46. ▌
    Test · nvidia-bionemo bundle
    Write and run multi-rank pytest parity tests that prove a CP implementation is numerically equivalent to its serial reference, using mp.spawn / spawn_multiprocessing as in the Boltz-CP test framework. Covers unit, layer-integration, module-integration, and workflow-integration levels; handles missing test data by asking the user or synthesizing random features in the correct format; and enforces the anti-vacuous rules (serial ground truth, explicit random grad_output, fp64 default tolerances, sharding-active and replicated-identical assertions). Also covers a second, non-parity class — property / invariant tests (e.g. test_rng_entropy) that assert a structural invariant such as RNG entropy when there is no serial value oracle. Use to verify any output of shard_data_feats or dtensor_modules.
    508 repo stars
  47. ▌
    Benchmark · nvidia-bionemo bundle
    Benchmark a CP inference or training workflow to find the maximum token count that fits at a given CP size and to record end-to-end walltime. Sweeps input size (tokens N, atoms, MSA depth S) over real or synthesized data, ramps N until OOM to report the max, times fwd (inference) or fwd+bwd+step (training) with warmup and median-of-repeats, captures peak memory, and writes a results table to docs/cp_benchmark.md. Use once a CP workflow runs end-to-end, to quantify the memory headroom CP buys and the walltime cost.
    508 repo stars
  48. ▌
    Build Infra · nvidia-bionemo bundle
    Probe and establish the distributed test infrastructure for CP development. Inventories local GPUs (count, model, memory, NVLink topology), checks the software stack (Python, PyTorch+CUDA, NCCL, torch.distributed), and runs shipped smoke tests for batch_isend_irecv P2P, all_gather, all_reduce, reduce_scatter, and a DTensor round-trip. Maps the GPU count to whether 2D CP is testable (2D needs >=4 GPUs; full integration wants 8). When insufficient local GPUs are available, generates a SLURM submission template and guides the user to provide cluster access. Records docs/cp_infra.md. Use after learn_context, before shard_data_feats / dtensor_modules / test.
    508 repo stars
  49. ▌
    Mem Profile · nvidia-bionemo bundle
    Memory-profile a context-parallel (CP) inference (or training) workflow with the PyTorch CUDA caching-allocator history, then attribute the top-N memory peaks to specific modules and lines of code. Wraps the end-to-end forward in torch.cuda.memory._record_memory_history() + _dump_snapshot() (the same mechanism as Boltz2's CUDAMemoryProfile Lightning callback) under a torchrun launcher that writes one snapshot per rank, then runs a stdlib analyzer (mem_profile_analysis.py) that replays the allocation timeline, finds the distinct peaks, and emits a markdown report with clickable file:line links to the call sites holding memory at each peak — sorted by peak, then by contribution. Use once a CP workflow runs end-to-end and you need to find the largest token count that fits and which module is the memory bottleneck.
    508 repo stars
  50. ▌
    Nsys Profile · nvidia-bionemo bundle
    Profile a context-parallel (CP) inference (or training) workflow with NVIDIA Nsight Systems (nsys). Locates or installs the nsys binary, mocks random features + small Glorot-init weights (or loads a real checkpoint), wraps the end-to-end forward in a torchrun launcher, runs nsys with the right CLI options (cuda/nvtx trace, --pytorch functions-trace/autograd, cudabacktrace, python-backtrace), and auto-annotates high-level module names (TriMul, Pairformer, trunk, diffusion, MSA) via a custom --python-functions-trace JSON — then reads the kernel/NVTX breakdown (Cannon-ring SendRecv vs GEMM, trunk vs diffusion vs confidence). Use once a CP workflow runs end-to-end and you need to see where GPU time and communication go.
    508 repo stars
  51. ▌
    Dispatch Work · nvidia-bionemo bundle
    Orchestrate a team of agents to integrate many CP modules and tests in parallel. Builds the work-list from the module map, derives a data-flow dependency graph, topologically sorts it into GPU-slot-bounded waves, and spawns a coder+reviewer pair per work item. Reviewers enforce serial-as-ground-truth, no-vacuous-tests, and the fold-cp hard rules; coupled up/down-stream items cross-review each other's input/output placement contracts before finalizing. Use when the scope is several CP modules and their tests at once — not for a single module (use dtensor_modules + test directly for that).
    508 repo stars
  52. ▌
    Learn Context · nvidia-bionemo bundle
    Explore a custom co-folding / structure-prediction model to understand its inference and training workflows, entry points, training framework (PyTorch Lightning vs DeepSpeed), data-feature format, and featurization pipeline — then map those components onto the existing Boltz-CP context-parallel reference so downstream CP integration reuses what already exists. Produces docs/current_code_structure.md. Use FIRST, before any other fold-cp skill, or whenever that doc is missing/stale. Runs interactively (Q&A with the user) or autonomously on request.
    508 repo stars
  53. ▌
    Dist Lifecycle · nvidia-bionemo
    Stand up the distributed-model lifecycle for a CP training/inference run: device placement before DTensor wrapping, the all-trainable-params-are-DTensors invariant (with placeholder/freeze for unimplemented modules), checkpoint save (DTensor → plain) and load (realign via the live state_dict template + redistribute optimizer state to parameter placements), resume RNG seed offset, and DTensor-safe EMA. Use after dtensor_modules + build_infra, when wiring the ported CP modules into a real trainer/predictor (Lightning or custom) and you need checkpoints, resume, or EMA to work — the gap between "modules pass parity" and "training runs and resumes".
    508 repo stars
  54. ▌
    Dtensor Modules · nvidia-bionemo bundle
    Implement DTensor-based context-parallel model modules that mirror a serial reference. Maps each serial layer/module to its Boltz-CP counterpart with exact input/output shapes, placements, collectives, and backward memory budget; writes the CP implementation following the autograd.Function conventions (explicit collectives, no implicit DTensor ops, promote_types, explicit from_local shape/stride, attribute-name + registration-order mirroring); and verifies it against the serial forward+backward via the test skill. Use when porting a custom model's trunk / attention / triangle / OPM / diffusion / confidence / loss modules to CP.
    508 repo stars
  55. ▌
    Shard Data Feats · nvidia-bionemo bundle
    Implement distributed data-feature sharding for CP: assign DTensor placements to every model feature, build the placement-definition dictionary, implement atom-feature pack/pad/scatter, per-shard and cross-axis divisibility padding, the DTensor Dataset / DataLoader / DataModule and collate, and the parity tests that prove the sharded features reassemble to the serial features. Traces the user's featurization from the inference/training drivers and mirrors the Boltz-CP data pipeline. Use after learn_context + build_infra, when wiring a custom model's features into a context-parallel mesh.
    508 repo stars
  56. ▌
    Cpize Model Workflow · nvidia-bionemo bundle
    Orchestrate the END-TO-END integration of context parallelism into a custom co-folding / structure-prediction model: turn the whole effort into a prioritized, dependency-sorted worklist, then drive it phase by phase — map the model (learn_context), verify infra (build_infra), shard data features (shard_data_feats), port modules (dtensor_modules, fanned out via dispatch_work), prove correctness (test), wire the trainer/predictor (dist_lifecycle), and finally profile memory and compute (mem_profile, nsys_profile) and benchmark. Use as the front door when the user wants to "CP-ify my model" as a whole program of work — it prioritizes, sequences, and gates the long task list and delegates each task to the right fold-cp skill. Not for a single module (use dtensor_modules + test directly).
    508 repo stars
  57. ▌
    Bionemo Phage Generation · nvidia-bionemo bundle
    Use when starting, planning, or resuming BioNeMo and Evo 2 bacteriophage genome generation or design for phage therapy research, including host-specific candidates for antibiotic-resistant infections and antimicrobial resistance (AMR); locates or acquires a compatible recipe checkout before implementation-specific work.
    508 repo stars
  58. ▌
    Bionemo Phage Design · nvidia-bionemo bundle
    Use when planning or running an Evo 2 bacteriophage genome-design project for phage therapy research, including host-specific candidates for antibiotic-resistant infections and antimicrobial resistance (AMR); coordinates evidence review, genome collection, SFT, GDPO reinforcement learning, checkpoint operations, safety QC, generation, and final screening.
    508 repo stars
  59. ▌
    Bionemo Phage Design Prepare Sft · nvidia-bionemo bundle
    Use when phage genomes must be deduplicated, clustered, split without near-duplicate leakage, and converted into explicit train, validation, and test inputs for Evo 2 supervised fine-tuning.
    508 repo stars
  60. ▌
    Bionemo Phage Design Adapt Execution · nvidia-bionemo bundle
    Use when a phage-design workflow must discover or adapt to local GPU, SSH, Slurm, Lepton, manual, or unfamiliar execution infrastructure and produce durable launch, monitoring, resume, or handoff commands.
    508 repo stars
  61. ▌
    Bionemo Phage Design Collect Genomes · nvidia-bionemo bundle
    Use when an Evo 2 phage SFT project needs a reproducible Microviridae or new phage genome collection from NCBI, paper-linked repositories, supplements, or other public biological databases.
    508 repo stars
  62. ▌
    Bionemo Phage Design Operate Nemo Rl · nvidia-bionemo bundle
    Use when launching, monitoring, resuming, relaunching, or selecting checkpoints from a NeMo-RL Evo2 phage optimization run.
    508 repo stars
  63. ▌
    Bionemo Phage Design Research Evidence · nvidia-bionemo bundle
    Use when a phage-design decision needs literature, database, dataset, gene-essentiality, synteny, viability, bootability, host, threshold, or model evidence.
    508 repo stars
  64. ▌
    Bionemo Phage Design Plan Rl Objectives · nvidia-bionemo bundle
    Use when converting a phage-design goal into target-specific RL rewards, validation criteria, and final QC filters, especially for a new reference phage or altered objective.
    508 repo stars
  65. ▌
    Bionemo Phage Design Generate And Screen · nvidia-bionemo bundle
    Use when producing, deduplicating, hard-QC screening, clustering, ranking, or selecting final phage designs from a chosen RL checkpoint.
    508 repo stars
  66. ▌
    Bionemo Phage Design Operate Mbridge Sft · nvidia-bionemo bundle
    Use when launching, monitoring, stopping, resuming, or relaunching Evo 2 phage SFT with Megatron Bridge, or when selecting its best validation-loss checkpoint across local, SSH, scheduler, or cloud execution.
    508 repo stars
  67. ▌
    Bionemo Phage Design Calibrate Rl Sampling · nvidia-bionemo bundle
    Use after selecting an Evo 2 phage SFT checkpoint and defining RL objectives to calibrate prompt serialization, temperature, prefix-length distribution, and fixed validation sampling.
    508 repo stars
  68. ▌
    Bionemo Phage Design Implement Rl Objectives · nvidia-bionemo bundle
    Use when adding or changing Evo2 phage RL metrics, reward functions, filter logic, or validation criteria after an objective plan has been approved.
    508 repo stars
  69. ▌
    Bionemo Phage Design Publish Stage Artifacts · nvidia-bionemo bundle
    Use when a phage-design user requests publication, backup, or an intermediate snapshot of selected checkpoints, validation generations, logs, results, or final deliverables to object, cloud, mounted, or network storage.
    508 repo stars
  70. ▌
    Diffdock Nim · nvidia-bionemo bundle
    Run DiffDock molecular docking via NVIDIA NIM to predict small-molecule binding poses against protein targets. Use for DiffDock, molecular docking, ligand docking, blind docking, SMILES or SDF ligands, ranked poses, confidence scores, hosted NVIDIA API, or local Docker deployment.
    508 repo stars
  71. ▌
    Openfold2 Nim · nvidia-bionemo bundle
    Use this skill for OpenFold2, NVIDIA's BioNeMo NIM microservice for monomer protein structure prediction. Invoke whenever the user mentions OpenFold2, AlphaFold2-like monomer folding, protein sequence-to-structure prediction, A3M MSAs, mmCIF templates, hosted NVIDIA API calls, or local Docker deployment.
    508 repo stars
  72. ▌
    Openfold3 Nim · nvidia-bionemo bundle
    Use this skill for OpenFold3, NVIDIA's BioNeMo NIM microservice for biomolecular structure prediction. Invoke whenever the user mentions OpenFold3 or needs protein, protein-ligand, protein-DNA/RNA, or multi-chain complex prediction with the hosted NVIDIA API or local Docker NIM. Covers endpoint choice, auth, request payloads, output artifacts, confidence scores, and local container setup.
    508 repo stars
  73. ▌
    Parabricks · nvidia-bionemo bundle
    Route NVIDIA Parabricks pbrun tools, assess GPU/runtime readiness, and provide version-aware command guidance for FASTQ/BAM processing, RNA-seq, variant calling, BAM QC, and GVCF workflows. Do NOT use for inspecting or accelerating whole pipelines — use genomics-workflow-acceleration.
    508 repo stars
  74. ▌
    Msa Search Nim · nvidia-bionemo bundle
    Generate multiple sequence alignments (MSAs) for protein sequences using the ColabFold MSA-Search NIM. Use for homolog search, UniRef30/ColabFold env searches, A3M or FASTA alignments, paired MSA search for complexes, PDB70 structural templates, hosted NVIDIA API calls, or local Docker deployment. For local deployment, download the databases in parallel with aria2c and launch via NIM_MODEL_NAME (the recommended default fast path, ~14 min vs over 80 min for the built-in downloader); a plain docker run uses the slow built-in downloader.
    508 repo stars
  75. ▌
    Proteinmpnn Nim · nvidia-bionemo bundle
    Run ProteinMPNN inverse folding via NVIDIA NIM to design protein sequences for a target backbone. Use for ProteinMPNN, inverse folding, sequence design, backbone redesign, fixed chains/residues, omit_AAs, sampling temperature, soluble model, hosted NVIDIA API, local Docker, PDB input, and multi-FASTA output.
    508 repo stars
  76. ▌
    Rfdiffusion Nim · nvidia-bionemo bundle
    Run RFDiffusion protein backbone design via NVIDIA NIM. Use for de novo protein backbones, motif scaffolding, binder design, hotspot residues, contigs syntax, diffusion steps, hosted NVIDIA API calls, local Docker deployment, and PDB backbone outputs for ProteinMPNN sequence design.
    508 repo stars
  77. ▌
    Genomics Workflow Acceleration · nvidia-bionemo bundle
    Use when accelerating existing genomics workflows with NVIDIA Parabricks, improving runtime or price/performance, converting pipeline steps to GPUs, or comparing CPU and GPU workflow outputs. Adds optional GPU steps in-place with runtime toggles (default off). Do NOT use for individual pbrun command routing — use parabricks.
    508 repo stars
  78. ▌
    Drug Discovery Pipeline · nvidia-bionemo bundle
    NOTE: molecule and target inputs and your NGC_API_KEY are transmitted to external NVIDIA-hosted API endpoints on every call. Use local NIM containers for confidential or proprietary data. Run a complete computational drug discovery pipeline using NVIDIA BioNeMo NIMs: generate drug-like molecules with GenMol, dock them to a protein target with DiffDock, then predict binding affinity with Boltz2. Use this skill whenever the user wants to generate and screen small molecule drug candidates, perform hit discovery, optimize leads against a protein target, or do virtual screening combining molecule generation, docking, and affinity prediction. Triggers on: drug discovery pipeline, hit discovery, lead optimization, virtual screening, molecule generation, molecular docking, binding affinity, GenMol, DiffDock, Boltz2, SMILES, SAFE notation, NIM microservice. This is a multi-step pipeline composing three BioNeMo NIMs.
    508 repo stars
  79. ▌
    Complexa Setup · nvidia-bionemo bundle
    First-time setup, environment configuration, and model-weight installation for Proteina-Complexa. Reach for this skill whenever the user says "set up complexa", "install complexa", "configure my environment file", "first-time setup", "what models do I have installed", "what's in my environment file", "download model weights", "download Complexa / AF2 / RF3 / ProteinMPNN / LigandMPNN / ESM2 / ESMFold checkpoints", "preflight my GPU", "verify environment", "complexa init", "complexa download", "complexa download --status", "complexa validate env", or any time a fresh checkout needs to be made runnable. This is the first skill to run on a new clone — it drives `complexa init`, `complexa download`, and `complexa validate env` end-to-end, edits the required `.env` keys, picks the right runtime (UV vs Docker), and emits a replayable setup artifact.
    508 repo stars
  80. ▌
    Complexa Sweep · nvidia-bionemo bundle
    Use this skill whenever the user wants to run a parameter sweep over a Proteina-Complexa design pipeline — cartesian-product hyperparameter scans, Pareto search over generation/reward/evaluation knobs, or any "compare configurations" workflow. Trigger phrases include "sweep beam width", "sweep nsteps", "hyperparameter sweep", "parameter scan", "scan beam_width and temperature", "compare configurations", "find the best generation params", "what's the optimal nsteps", "Pareto search for binder quality vs wall-clock", "complexa sweep", "tune Complexa", "ablate the reward weights", "configs/sweeps", "--sweeper", "run beam_width.yaml". This is the only skill that owns sweeper YAML authoring, cartesian-product expansion, and per-config result ranking.
    508 repo stars
  81. ▌
    Complexa Design · nvidia-bionemo bundle
    End-to-end Proteina-Complexa design pipeline driver. Use when the user wants to "design a binder", "design binders for X", "run complexa design", "de novo binder", "PDL1 binder", "TrkA binder", "design proteins for target", "protein binder design", "ligand binder", "design a small-molecule binder", "ATP-binding protein", "AME motif scaffolding", "scaffold a motif near a ligand", "motif + ligand design", "enzyme scaffolding", "flow matching protein design", "beam-search binder", "FK steering", "MCTS protein design", or "refold with AF2 / RF3 / ESMFold", or wants success rates, interface pAE, scRMSD, or FoldSeek diversity from one command. Drives the complexa design pipeline from target picking to manifest emission and reports how many designs passed.
    508 repo stars
  82. ▌
    Complexa Target · nvidia-bionemo bundle
    Use this skill whenever the user wants to add, register, edit, list, show, or validate a Proteina-Complexa design target for any pipeline — protein binder (default), ligand binder, or AME / enzyme scaffolding. Triggers include "add a target", "define a new target for binder design", "register a hotspot", "set up a PDL1 binder target", "ligand binder pocket", "SMILES target", "AME task", "enzyme motif", "M0024_1nzy", "complexa target add", "complexa target show", "configure target X", "what targets are available", "where do hotspots live", "what does target_input mean", "chain-spec syntax", "binder length range", or any question about `configs/targets/{,ligand_}targets_dict.yaml` and `configs/design_tasks/ame_dict_v2.yaml`. Also covers `complexa validate target`. This is the only skill that touches the three targets dict files.
    508 repo stars
  83. ▌
    Msa Structure Prediction Pipeline · nvidia-bionemo bundle
    NOTE: your protein sequence and the retrieved MSA alignment are transmitted to external NVIDIA-hosted APIs (health.api.nvidia.com) on every call. Use local NIM containers for confidential or proprietary sequences. Run a complete protein structure prediction pipeline using NVIDIA BioNeMo NIMs: search for MSA alignments with MSA-Search (ColabFold), then predict the structure with OpenFold3 using the retrieved alignments. Use this skill whenever the user wants to predict a protein structure with maximum accuracy using MSA context, run the full AlphaFold3-style pipeline, generate MSA-informed structure predictions, or improve structure prediction accuracy by providing evolutionary information. Triggers on: MSA structure prediction pipeline, structure prediction pipeline, MSA-informed prediction, OpenFold3, ColabFold MSA, AlphaFold3 pipeline, protein structure, homology search, a3m alignment, UniRef30, NIM microservice. This pipeline chains MSA-Search and OpenFold3.
    508 repo stars
  84. ▌
    Protein Binder Design · nvidia-bionemo bundle
    Orchestrate an end-to-end de novo protein binder design campaign against a protein target by composing BioNeMo NIM skills. Use for binder design, minibinder design, de novo binders, RFdiffusion + ProteinMPNN + Boltz2/OpenFold3 pipelines, epitope/hotspot-targeted design, in-silico binder validation, and ranking designs by interface confidence.
    508 repo stars
  85. ▌
    Complexa Binder Design · nvidia-bionemo bundle
    Run a complete protein binder design campaign with NVIDIA Proteina-Complexa: resolve a target structure and hotspots from a name/sequence/PDB, co-design binder sequence+structure with reward-guided test-time search (best-of-n, beam search, FK steering, MCTS), select with the internal AF2 reward gate, then INDEPENDENTLY validate each binder by refolding the complex with Boltz2 (default) or OpenFold3 and rank on interface confidence, pLDDT, ipSAE, apo/holo stability, and hotspot contact. Use whenever the user wants de novo binders against a named target, sequence, or PDB, hotspot/epitope-targeted design, Proteina-Complexa / Complexa, or ranked validated binders from one request. Sibling of protein-binder-design (RFdiffusion + ProteinMPNN); this skill uses Proteina-Complexa.
    508 repo stars
  86. ▌
    Complexa Evaluate Pdbs · nvidia-bionemo bundle
    Standalone evaluation of an existing PDB directory with Proteina-Complexa. Use when the user wants to "evaluate PDB files", "re-fold these designs", "compute interface pAE", "compute i_pLDDT for a folder", "run AF2 / RF3 / ESMFold on my designs", "score binder candidates", "designability of this folder", "scRMSD for designs", "motif RMSD for these PDBs", "complexa analysis", "complexa evaluate from a PDB directory", "evaluate from pdb dir", or to score third-party outputs (BindCraft, AlphaProteo, RFdiffusion, hand-curated decoys). Picks the correct evaluate_*.yaml config, wires ++dataset.pdb_dir and the folding backend, runs the evaluate → analyze chain, parses the result CSV, and reports pass-rates against the right result_type thresholds.
    508 repo stars
  87. ▌
    Evo2 Nim · nvidia-bionemo bundle
    Generate and analyze DNA sequences using NVIDIA's Evo 2 BioNeMo NIM microservice. Use for Evo2/Evo 2, DNA generation, genomic sequence generation, hosted generation, local Docker deployment, local forward passes, layer outputs, logits, sampled probabilities, and BioNeMo NIM workflows.
    508 repo stars
  88. ▌
    Boltz2 Nim · nvidia-bionemo bundle
    Use Boltz2 NIM for biomolecular structure prediction and binding affinity. Invoke for Boltz2, protein structures, protein-ligand/DNA/RNA complexes, SMILES or CCD ligands, pIC50/IC50 affinity scoring, mmCIF output, hosted NVIDIA API calls, or local Docker deployment.
    508 repo stars
  89. ▌
    Genmol Nim · nvidia-bionemo bundle
    Generate novel drug-like molecules using the GenMol NIM microservice. Use for de novo generation, scaffold decoration, motif extension, lead optimization, SAFE notation, QED or LogP ranking, hosted NVIDIA API calls, or local Docker deployment. GenMol takes SAFE notation in the smiles field, not ordinary SMILES.
    508 repo stars
  90. ▌
    Molmim Nim · nvidia-bionemo bundle
    Use this skill for MolMIM, NVIDIA's BioNeMo NIM microservice for small-molecule latent-space generation and optimization. Invoke for MolMIM, molecular embeddings, hidden states, latent decoding, sampling around a seed SMILES, CMA-ES guided molecule generation, QED or plogP optimization, hosted NVIDIA API calls, or local Docker deployment.
    508 repo stars
  91. ▌
    Nvmolkit Usage · nvidia-bionemo bundle
    Write code that calls the installed nvMolKit Python API for GPU-accelerated, batched RDKit-style operations - Morgan fingerprints, Tanimoto/cosine similarity, ETKDG conformer embedding, MMFF/UFF optimization, TFD, conformer RMSD, Butina clustering, substructure search, and maximum common substructure (MCS) search. Use when the user is importing `nvmolkit.*`, debugging an `nvmolkit` call, choosing between nvMolKit and RDKit for a batched cheminformatics workflow, or wiring nvMolKit results into a torch/numpy pipeline. Out of scope: building nvMolKit from source.
    508 repo stars
  92. ▌
    Nvqsp · nvidia-bionemo bundle
    Build and run GPU-accelerated Quantitative Systems Pharmacology (QSP) and PBPK population ODE simulations with the nvQSP library, and differentiate them. nvQSP provides a sparse polynomial RODAS4 stiff solver (prebuilt) and a model-specialized dense adaptive TSIT5 solver, both with first-class gradients. Use this whenever the user wants to simulate compartmental PK/PBPK models, run virtual-population or batch patient simulations, do population PK variability studies, fit or train model parameters against data, solve stiff pharmacology ODEs on an NVIDIA GPU, or mentions nvQSP, `nvqsp`, `sparse.solve`, `gradients.solve`, `tsit5`, RODAS4, TSIT5, parameter sensitivities, or translating a pharmacology model into A0/A1/A2 coefficient form. Trigger this even when the user just describes a compartmental model ("two-compartment IV infusion across 1000 patients") without naming the library, since encoding the model into nvQSP's sparse-tensor form and getting the dosing/CSR conventions right is the hard part.
    508 repo stars
  93. ▌
    Hyenand Retrofit · nvidia-bionemo bundle
    Replace attention in a PyTorch model with HyenaND from the nvSubquadratic library. Covers 1D / 2D / 3D hosts (ViT, U-Net, diffusion, causal LM, hierarchical encoders). Trigger when the user wants a subquadratic alternative to attention, ports a model to HyenaND, swaps `nn.MultiheadAttention` or `F.scaled_dot_product_attention` for a Hyena mixer, builds a striped Hyena LM, or asks "how do I use nvSubquadratic with my model." Phrases like "make my ViT subquadratic," "Hyena layer for my U-Net," "swap attention with FFT convolution," "subquadratic alternative for my 3D segmentation network," or "long-context model with O(L log L) scaling" should all activate this skill.
    508 repo stars