i4h-workflow-dataset-transfer

name	i4h-workflow-dataset-transfer
version	0.6.0
description	Augment HDF5 camera streams with NVIDIA Cosmos Transfer for visual diversity. Use when the user asks to apply Cosmos, domain-randomize videos, or make sim look photoreal. Requires Docker and GPU.
license	Apache-2.0
metadata	{"author":"Isaac for Healthcare Team <[e-mail]>","tags":["isaac-for-healthcare","i4h","dataset","cosmos","domain-randomization"]}

i4h Workflow — Cosmos Transfer

Purpose

Augment HDF5 camera streams with NVIDIA Cosmos Transfer for visual diversity. Use when the user asks to apply Cosmos, domain-randomize videos, or make sim look photoreal.

Base Code

These steps drive the i4h-workflows base code (the workflows/agentic/ tree). To reuse an existing checkout, set I4H_WORKFLOWS to its path (no clone happens). Otherwise this resolves the current repo, or clones to ~/i4h-workflows — pick that default without prompting. Run every command below from the resolved root:

# Resolve the i4h-workflows base code (provides workflows/agentic/).
ROOT="${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}"
if [ ! -d "$ROOT/workflows/agentic" ]; then
  ROOT="${I4H_WORKFLOWS:-$HOME/i4h-workflows}"
  [ -d "$ROOT/workflows/agentic" ] || git clone https://github.com/isaac-for-healthcare/i4h-workflows "$ROOT"
fi
export I4H_WORKFLOWS="$ROOT"; cd "$ROOT"

Basics

• Env config (source of truth): workflows/agentic/config/environments/<env>.yaml — the camera streams (zenoh.camera_names) Cosmos augments.
• Cosmos modifies videos only; robot states and actions are unchanged.
• Requires Docker with NVIDIA GPU support and accepted Cosmos model licenses.
• For action/state augmentation use [[i4h-workflow-dataset-mimic]] instead.

Run

Component scripts live under workflows/agentic/cosmos/. Cosmos CLI flags change more often than the core workflow — consult --help for the current options.

REPO_ROOT="${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}"; [ -d "$REPO_ROOT/workflows/agentic" ] || REPO_ROOT="$HOME/i4h-workflows"
ENV_ID=scissor_pick_and_place
RUNS_ROOT="${REPO_ROOT}/workflows/agentic/runs"

# Point HDF5_PATH at a real recording to augment (absolute path). Recordings come from teleop,
# mimic, or validate (data/verify.hdf5 under each runs/eval_* dir). List candidates newest-first:
#   find "${RUNS_ROOT}" -name '*.hdf5' -printf '%TY-%Tm-%Td %TH:%TM  %p\n' | sort -r | head
HDF5_PATH="${HDF5_PATH:-}"
if [ ! -f "${HDF5_PATH}" ]; then
  echo "cosmos: set HDF5_PATH to an existing .hdf5 (got '${HDF5_PATH:-<unset>}'). Candidates:" >&2
  find "${RUNS_ROOT}" -name '*.hdf5' -printf '%TY-%Tm-%Td %TH:%TM  %p\n' 2>/dev/null | sort -r | head
  exit 1
fi

RUN_DIR="${RUNS_ROOT}/cosmos_${ENV_ID}_$(date +%Y%m%d_%H%M%S)"
mkdir -p "${RUN_DIR}/data" "${RUN_DIR}/logs"
ln -sfn "${RUN_DIR}" "${RUNS_ROOT}/.latest"
OUT="${RUN_DIR}/data/demo_cosmos.hdf5"

"${REPO_ROOT}/workflows/agentic/cosmos/run.sh" --help

Verify

• Output HDF5 preserves the original state/action groups.
• Camera streams are present and playable.
• Replay or convert + viz the output before declaring done ([[i4h-workflow-dataset-replay]], [[i4h-lerobot-viz]]).

Prerequisites

• Workflow set up via [[i4h-workflow-setup]] (the .venv must exist).
• Docker with NVIDIA GPU support and accepted Cosmos model licenses.
• An existing HDF5 recording to augment (set HDF5_PATH to an absolute path; the Run block lists candidates if it's unset or wrong).
• Component scripts under workflows/agentic/cosmos/; check run.sh --help for current flags.

Limitations

• Cosmos modifies videos only; robot states and actions are unchanged.
• Requires Docker with NVIDIA GPU support and accepted Cosmos model licenses.
• For action/state augmentation use [[i4h-workflow-dataset-mimic]] instead.
• Cosmos CLI flags change more often than the core workflow; consult --help for the current options.

Troubleshooting

• Error: .venv not found / module import fails - Cause: workflow not set up. Fix: run [[i4h-workflow-setup]] first.
• Error: Docker / GPU not available - Cause: Docker lacks NVIDIA GPU support. Fix: ensure Docker with NVIDIA GPU support is installed and running.
• Error: model license not accepted - Cause: Cosmos model licenses not accepted. Fix: accept the required Cosmos model licenses.
• Error: input HDF5 not found - Cause: wrong or missing --hdf5-path. Fix: point HDF5_PATH at an existing recording.

Final Response

Report input HDF5, output HDF5, camera streams modified, Docker or model-license blockers.