528 skills · sorted by relevance
Launch distributed Megatron-LM training on a SLURM cluster with guidance on job scripts, environment variables, GPU settings, and containers. Monitor jobs and investigate failures across worker processes to identify the initial cause.
Edit an existing Isaac for Healthcare simulation scene, including object placement, cameras, and task success rules. Review changes in a running simulator, then save the confirmed scene to its source files when explicitly requested.
Collects a read-only health snapshot of a Jetson device, covering its model, memory, GPU, temperatures, power mode, storage, services, and processes. Helps investigate slowdowns, overheating, and memory pressure using observed device data.
Manage nvfancontrol fan profiles for Jetson and Tegra boards. Add, remove, or edit temperature-based fan curves and boot defaults in board support files, while tracking changes and preserving configuration differences between related carrier boards.
Configure how Dynamo routes requests to workers, then check model discovery and a chat response. Supports local and Kubernetes setups, including routing based on cached model data, with clear commands, results, and limitations.
Check installed DOCA versions across package metadata, application files, runtime tools, and BlueField system information. Use the results to investigate build and runtime mismatches, check feature availability, and understand NGC container version tags.
Run an Isaac for Healthcare workflow from robot recordings through dataset preparation, policy training, and checkpoint validation. Uses the maintained pipeline driver, checks the planned commands first, and verifies artifacts and logs at each stage.
Check whether a Dynamo deployment is ready for disaggregated serving, where input processing and token generation run separately. Read-only probes inspect transport settings, GPU networking capabilities, and NIXL transfer-test readiness, then report gaps and next steps.
Host-side DOCA UROM library (doca_urom_*): offload an HPC / UCX / MPI application's remote memory operations (puts, gets, atomics, collectives) from the host CPU onto a BlueField DPU. Trigger on "MPI all-reduce burning host CPU", "push UCX traffic onto the BlueField", "first doca_urom call returns NOT_PERMITTED", or "host library and DPU service look out of sync", even without naming DOCA UROM. An offload needs both halves: this host library and the UROM Service on the DPU (doca-urom-svc), which owns deploying that container. Covers creating UROM Service contexts (doca_urom_service_*) and Worker contexts (doca_urom_worker_*) that run plugins on the DPU, discovering plugins via doca_urom_service_get_plugins_list, progressing completions, and DOCA_ERROR_* from doca_urom_* calls. Route elsewhere for UROM Service deployment on the DPU side, MPI / UCX collective algorithm design, and RDMA / RoCE / IB substrate bring-up.
The English summary is being prepared. Showing the original description.
Create a minimal blank Workflow scaffold with a Scene containing ground and light plus an idle run mode. Use for fast new Workflow scaffolding.
The English summary is being prepared. Showing the original description.
Grade recorded workflow episodes against a written success criterion using a vision model, which interprets sampled camera images. Summarize pass and fail counts, or optionally extract successful episodes into a separate HDF5 dataset.
Load, configure, run, observe, and stop custom congestion control algorithms on BlueField using doca_spcx_cc. The skill guides comparisons with doca-pcc, evaluation under controlled network contention, and diagnosis of loading errors or unchanged traffic behavior.
Expand workflow HDF5 demonstrations with action jitter, optionally scoped to node segments. Use for synthetic variants; do not use to collect data, alter state directly, or generate new images.
The English summary is being prepared. Showing the original description.
Open a completed LeRobot dataset in a browser to inspect camera videos, state and action timelines, and episode details. The skill manages a local server and reports observations together with the command to stop it.
Find and run Holoscan Sensor Bridge example applications that match a connected development kit, board, software version, and sensors. Monitor timed runs, investigate failures, suggest code or environment changes, and summarize the session.
Assess a Linux host and choose an installation method for Holoscan SDK 5.x. Compare containers, Debian packages, Python wheels, and source builds against available releases, CUDA 13 compatibility, and the components needed.
Create or explain fixed-format HSB FPGA_top.sv wrappers from validated HOLOLINK_def.svh files. Do not use for def generation or validation.
The English summary is being prepared. Showing the original description.
Diagnose failed or unhealthy Dynamo deployments. Use when pods, model-cache jobs, PVCs, workers, frontend/router health, endpoints, or benchmark jobs fail; use recipe-runner/router-starter before this for normal bring-up.
The English summary is being prepared. Showing the original description.
Not for writing library code (name that library skill). Use this skill whenever another DOCA skill says "prefer the structured tool per doca-structured-tools-contract", or when the user wants a one-shot answer that consolidates info multiple manual commands would produce — DOCA env / version / devices / capabilities / validate / host vs DPU state. Trigger even when the user does not explicitly mention "structured tool" or "doca-env --json" — typical implicit phrasings include "is there one command that tells me everything about my DOCA install", "what version is X capability available since", "every PF/VF/SF visible on this BlueField with PCIe address", "will this pipe pass validate before commit", "diff host vs DPU state", or "why does the agent give a one-line answer on host A and five commands on host B". Refuse and route elsewhere for general DOCA orientation, specific library API how-to, or install-from-scratch guidance — those belong to the per-library skill, doca-public-knowledge-map, or doca-setup.
The English summary is being prepared. Showing the original description.
Record human-operated robot demonstrations in workflow HDF5 files, a structured data format. The skill guides supported input-device selection, foreground recording, and checks of robot motion, camera observations, saved episodes, and task outcomes.
Find suitable Earth2Studio sources for weather and climate variables at specified times, verify variable support, and generate a Python script that retrieves historical or forecast data as an xarray DataArray, an array with labeled dimensions.
Assess whether a DOCA application needs low-level communication controls beyond its higher-level APIs. For those cases, guide queue configuration, libibverbs migration, and error diagnosis using installed headers, version checks, and device capability queries.
Fine-tune GR00T or openpi robot policies using compatible LeRobot data. Check dataset requirements, preview the training configuration, run training, and verify that a saved checkpoint, a file containing the trained model, loads through the inference backend.
Generate, validate, compare, or explain HSB HOLOLINK_def.svh macros. Do not use for FPGA_top.sv wrappers or packetizer-only derivation. Generation runs bundled Python scripts locally through shell commands and writes validated .svh output files after user-confirmed paths.
The English summary is being prepared. Showing the original description.