Workflow Templates (OSMO)
Canonical OSMO workflow templates for RL, imitation learning, VLA, SiL, and HiL workloads. Template names in this page are based on current YAML files and exclude stale legacy naming.
Template Inventory
| Template | Purpose | Source YAML path | Typical submit path |
|---|---|---|---|
train.yaml | Isaac Lab RL training with object-storage code delivery | training/rl/workflows/osmo/train.yaml | training/rl/scripts/submit-osmo-training.sh |
train-dataset.yaml | Isaac Lab RL training with dataset folder injection | training/rl/workflows/osmo/train-dataset.yaml | training/rl/scripts/submit-osmo-dataset-training.sh |
lerobot-train.yaml | LeRobot ACT or Diffusion training workflow | training/il/workflows/osmo/lerobot-train.yaml | training/il/scripts/submit-osmo-lerobot-training.sh |
groot-train.yaml | NVIDIA GR00T VLA fine-tuning workflow | training/vla/workflows/osmo/groot-train.yaml | training/vla/scripts/submit-osmo-lerobot-vla-fine-tuning.sh |
eval.yaml | Isaac Lab checkpoint evaluation workflow | evaluation/sil/workflows/osmo/eval.yaml | evaluation/sil/scripts/submit-osmo-eval.sh |
lerobot-eval.yaml | LeRobot policy evaluation workflow | evaluation/sil/workflows/osmo/lerobot-eval.yaml | evaluation/sil/scripts/submit-osmo-lerobot-eval.sh |
cpu-smoke.yaml | CPU-only HiL scheduling proof | evaluation/hil/workflows/osmo/cpu-smoke.yaml | data-pipeline/setup/hil/03-run-cpu-smoke.sh |
hil-evaluation.yaml | No-command HiL safety evaluation | evaluation/hil/workflows/osmo/hil-evaluation.yaml | data-pipeline/setup/hil/04-run-no-command-check.sh |
train.yaml
| Field | Details |
|---|---|
| Purpose | OSMO RL training. Code is packaged, uploaded to object storage with osmo data upload, and injected into the pod through a url: task input. |
| Source YAML path | training/rl/workflows/osmo/train.yaml |
| Primary parameters and overrides | default-values.task (Isaac-Velocity-Rough-Anymal-C-v0), default-values.num_envs ("2048"), default-values.max_iterations (empty), default-values.checkpoint_mode (from-scratch), default-values.training_backend (skrl), default-values.gpu ("1"), default-values.cpu ("30"). |
| Typical submit path | training/rl/scripts/submit-osmo-training.sh |
| Usage notes | Use for RL training. The submission delivers code via object storage; script flags typically override task, resources, and checkpoint behavior. |
train-dataset.yaml
| Field | Details |
|---|---|
| Purpose | OSMO RL training that mounts training code from an uploaded dataset path. |
| Source YAML path | training/rl/workflows/osmo/train-dataset.yaml |
| Primary parameters and overrides | default-values.dataset_bucket (training), default-values.dataset_name (training-code), default-values.task (Isaac-Velocity-Rough-Anymal-C-v0), default-values.num_envs ("2048"), default-values.checkpoint_mode (from-scratch), default-values.training_backend (skrl). |
| Typical submit path | training/rl/scripts/submit-osmo-dataset-training.sh |
| Usage notes | Use when payload size or reuse favors dataset-based delivery. The script stages and uploads training sources before submission. |
lerobot-train.yaml
| Field | Details |
|---|---|
| Purpose | OSMO LeRobot training with optional Azure Blob dataset source and checkpoint registration. |
| Source YAML path | training/il/workflows/osmo/lerobot-train.yaml |
| Primary parameters and overrides | default-values.policy_type (act), default-values.dataset_repo_id (empty), default-values.dataset_root (/workspace/data), default-values.blob_urls ([]), default-values.training_steps ("100000"), default-values.batch_size ("32"), default-values.learning_rate ("1e-4"), default-values.save_freq ("5000"), default-values.num_gpus ("1"), default-values.mixed_precision (no), default-values.platform (gpu_platform), default-values.register_checkpoint (empty). |
| Typical submit path | training/il/scripts/submit-osmo-lerobot-training.sh |
| Usage notes | Supports HuggingFace and blob-backed datasets. Keep policy type and data source aligned with script flags to avoid mixed-source configuration. |
groot-train.yaml
| Field | Details |
|---|---|
| Purpose | NVIDIA GR00T N1.5 or N1.7 fine-tuning from a LeRobot-format dataset in Azure Blob Storage. |
| Source YAML path | training/vla/workflows/osmo/groot-train.yaml |
| Primary parameters and overrides | default-values.workflow_name (groot-train), default-values.gpu (1), default-values.platform (gpu_platform), default-values.dataset_path (/data/dataset), default-values.batch_size (4), default-values.max_steps (500), default-values.save_steps (100), default-values.resume (false), default-values.azure_upload (false). |
| Typical submit path | training/vla/scripts/submit-osmo-lerobot-vla-fine-tuning.sh |
| Usage notes | Requires a blob URL and data configuration. The wrapper selects version-specific GR00T source and model revisions and injects the runtime assets. |
eval.yaml
| Field | Details |
|---|---|
| Purpose | OSMO Isaac Lab checkpoint evaluation for policy export and rollout scoring. |
| Source YAML path | evaluation/sil/workflows/osmo/eval.yaml |
| Primary parameters and overrides | default-values.task (Isaac-Ant-v0), default-values.num_envs ("4"), default-values.max_steps ("500"), default-values.video_length ("200"), default-values.checkpoint_uri (empty), default-values.inference_format (both). |
| Typical submit path | evaluation/sil/scripts/submit-osmo-eval.sh |
| Usage notes | Requires checkpoint URI at submission. Use inference_format to control ONNX/JIT export behavior for downstream use. |
lerobot-eval.yaml
| Field | Details |
|---|---|
| Purpose | OSMO LeRobot evaluation for HuggingFace or AzureML model sources, with optional registration. |
| Source YAML path | evaluation/sil/workflows/osmo/lerobot-eval.yaml |
| Primary parameters and overrides | default-values.policy_repo_id (empty), default-values.policy_revision (empty), default-values.policy_type (act), default-values.dataset_repo_id (empty), default-values.dataset_revision (empty), default-values.eval_episodes ("10"), default-values.eval_batch_size ("10"), default-values.record_video ("false"), default-values.mlflow_enable ("false"), default-values.builtin_policy ("false"), default-values.register_model (empty), default-values.blob_storage_container (datasets). |
| Typical submit path | evaluation/sil/scripts/submit-osmo-lerobot-eval.sh |
| Usage notes | This is the canonical LeRobot OSMO evaluation template. |
cpu-smoke.yaml
| Field | Details |
|---|---|
| Purpose | CPU-only HiL scheduling proof for an existing connected edge pool. |
| Source YAML path | evaluation/hil/workflows/osmo/cpu-smoke.yaml |
| Primary parameters and overrides | default-values.workflow_name (hil-cpu-smoke), default-values.backend_name (hil-edge), default-values.pool_name (hil-edge), resources.default.gpu (0), resources.default.platform (hil_cpu). |
| Typical submit path | data-pipeline/setup/hil/03-run-cpu-smoke.sh |
| Usage notes | Submit through the wrapper with a validated HiL connection receipt. The workflow verifies non-root CPU execution without GPU devices. |
hil-evaluation.yaml
| Field | Details |
|---|---|
| Purpose | CPU-only no-command safety evaluation using deterministic UR10E observations. |
| Source YAML path | evaluation/hil/workflows/osmo/hil-evaluation.yaml |
| Primary parameters and overrides | default-values.workflow_name (ur10e-no-command), default-values.backend_name (hil-edge), default-values.pool_name (hil-edge), resources.default.gpu (0), resources.default.platform (hil_cpu). |
| Typical submit path | data-pipeline/setup/hil/04-run-no-command-check.sh |
| Usage notes | Submit through the wrapper with a validated HiL connection receipt. The workload proposes actions but contains no robot command transport and requires zero applied actions. |
Usage Notes
| Topic | Guidance |
|---|---|
| Source of truth | Use YAML files under training/ and evaluation/ as the canonical inventory. |
| Submission flow | Submit through the companion scripts listed above to resolve defaults from CLI, env vars, and Terraform outputs. |
| Runtime packaging | RL workflows deliver code via object storage (url: input) or dataset injection; choose based on reuse needs. |
| Related reference | See Reference index for adjacent script and artifact guides. |