Initial: 3 tarballs (episodes_g 1.2GB + USDs 489MB + collected 266MB) + RESUME orchestrator + README

README.md

@@ -0,0 +1,241 @@
+---
+license: cc-by-4.0
+language:
+- en
+pretty_name: baseline_3 v4 — Sim Collection Assets (RESUME on A6000)
+tags:
+- robotics
+- manipulation
+- grasping
+- diffusion-policy
+- franka
+- dexycb
+- collection-assets
+size_categories:
+- 1B<n<10B
+---
+
+# baseline_3 v4 — sim collection RESUME package
+
+Everything needed for a partner machine to **resume** the baseline_3 v4 sim
+collection that ran on a dev box (RTX 5090) until the orchestrator was killed
+to free GPU. Resume-aware orchestrator detects what's already done and only
+collects the remaining objects.
+
+**Source repo (code lives here, NOT in this dataset)**:
+[`stzabl-png/UCB_Project @ gate3-curobo-ik`](https://github.com/stzabl-png/UCB_Project/tree/gate3-curobo-ik)
+— specifically `sim/run_grasp_sim_baseline3_v4.py` and
+`scripts/baseline_3_v4/*.sh`. Both shell scripts in this dataset are convenience
+copies of the repo versions — the canonical ones live in the repo.
+
+---
+
+## What's in this package
+
+| file | size | purpose |
+|------|------|---------|
+| `episodes_g.tar.gz` | 1.2 GB | 900 retargeted DexYCB hdf5 (collector INPUT). Includes ycb_dex_01–20 minus the no-data ones; specifically the **newly retargeted dex_11 (pitcher) + dex_14 (mug)** missing from previous runs. |
+| `obj_usd_cad_ycb.tar.gz` | 489 MB | YCB CAD USDs (17 files, Z-up tagged) + textures. Includes the **new dex_11/14/19 USDs** generated this round. |
+| `episodes_b3_v4_15obj_3yaw_2026-05-25.tar.gz` | 266 MB | Already-collected sim hdf5 for 6 objects (tuna 06, tomato 04, pudding 07, gelatin 08, marker 18, potted_meat 09). |
+| `v4_full15_results_2026-05-25.txt` | 1.5 KB | Results log with 6 DONE_RETRY entries → orchestrator uses this to skip done objects. |
+| `v4_full15_queue_resume.sh` | 4.6 KB | RESUME-aware orchestrator (2-parallel). Copy of repo `scripts/baseline_3_v4/`. |
+| `v4_chunked_with_retry.sh` | 4.6 KB | Per-object chunk-5+retry wrapper. Copy of repo `scripts/baseline_3_v4/`. |
+
+---
+
+## Before you start — what to install
+
+### 1. Conda env: `env_isaaclab` (heavy)
+
+```bash
+conda create -n env_isaaclab python=3.11 -y
+conda activate env_isaaclab
+
+# 1.1 IsaacSim 5.1 (NOT 4.x — the v4 collector uses 5.1 APIs)
+pip install --extra-index-url=https://pypi.nvidia.com isaacsim==5.1.0
+# Also accept the EULA when prompted on first run.
+
+# 1.2 cuRobo 0.8 (NOT 0.7 — the planner uses MotionPlanner / SceneCfg.create API)
+pip install curobo==0.8.0
+
+# 1.3 Other deps used by the collector + retry wrapper
+pip install h5py scipy 'numpy<2' termcolor pyyaml requests usd-core
+```
+
+Verify the install:
+```bash
+python -c "from isaacsim import SimulationApp; print('IsaacSim OK')"
+python -c "import curobo; print('cuRobo OK')"
+```
+
+### 2. Repo (gate3-curobo-ik branch)
+
+```bash
+git clone -b gate3-curobo-ik git@github.com:stzabl-png/UCB_Project.git
+cd UCB_Project
+# DP3 is already vendored under third_party/3D-Diffusion-Policy/3D-Diffusion-Policy
+# but you do NOT need to install it for sim collection (only for training).
+```
+
+### 3. GPU
+
+- Tested on A6000 48 GB and RTX 5090 32 GB.
+- 2-parallel collection uses ~10 GB sustained, ~18 GB peak (cuRobo subprocess spikes).
+- Single-parallel (`OBJS` loop run sequentially) if VRAM is tight.
+
+---
+
+## Download + extract (run from `UCB_Project/`)
+
+```bash
+pip install huggingface_hub
+mkdir -p Baseline1/data output
+
+# Pull tarballs
+huggingface-cli download UCBProject/baseline_3_v4_collection_assets \
+    --repo-type dataset \
+    --local-dir /tmp/v4_pkg
+
+# Extract into matching layout (uses --strip-components=0 so paths preserved as-is)
+tar xzf /tmp/v4_pkg/episodes_g.tar.gz                               # → Baseline1/data/episodes_g/
+tar xzf /tmp/v4_pkg/obj_usd_cad_ycb.tar.gz                          # → output/obj_usd_cad/ycb/
+tar xzf /tmp/v4_pkg/episodes_b3_v4_15obj_3yaw_2026-05-25.tar.gz     # → Baseline1/data/episodes_b3_v4_15obj_3yaw_2026-05-25/
+
+# Place the results log + scripts where orchestrator expects
+cp /tmp/v4_pkg/v4_full15_results_2026-05-25.txt /tmp/
+# Optional — only if your repo doesn't already have these
+# (gate3-curobo-ik DOES have them under scripts/baseline_3_v4/, prefer those):
+# cp /tmp/v4_pkg/v4_*.sh /tmp/
+```
+
+Verify layout:
+```bash
+ls Baseline1/data/episodes_g/*.hdf5 | wc -l              # expect 900
+ls output/obj_usd_cad/ycb/*.usd | wc -l                  # expect ~17
+ls Baseline1/data/episodes_b3_v4_15obj_3yaw_2026-05-25/*.hdf5 | wc -l   # expect ~153
+grep DONE_RETRY /tmp/v4_full15_results_2026-05-25.txt | wc -l    # expect 6
+```
+
+---
+
+## Run the resume
+
+**This is the only command you need:**
+
+```bash
+cd UCB_Project   # repo root
+
+# Override the env vars to match YOUR layout (replace paths with yours)
+PROJ=$HOME/UCB_Project \
+PY=$HOME/miniconda3/envs/env_isaaclab/bin/python \
+OUT=Baseline1/data/episodes_b3_v4_15obj_3yaw_2026-05-25 \
+RESULTS=/tmp/v4_full15_results_2026-05-25.txt \
+PAR=2 \
+  bash scripts/baseline_3_v4/v4_full15_queue_resume.sh
+```
+
+The orchestrator will:
+1. Read `RESULTS` and find 6 `DONE_RETRY` entries (06, 04, 07, 08, 18, 09) → skip them
+2. Launch the remaining 9 objects (12, 05, 03, 15, 02, 11, 14, 20, *and 19 which has 0 src ep and will silently emit nothing*) at PAR parallel
+3. Each object: chunk-5 + auto-retry on PhysX-pollution sanity fail
+4. New `DONE_RETRY` lines append to `RESULTS` as objects finish
+5. Final per-object summary appended at the end
+
+### Choosing `PAR` on A6000
+
+A6000 has 48 GB VRAM but **half** the CUDA cores of RTX 5090 (where this run
+was validated at PAR=2). cuRobo subprocesses (called per chunk by the
+collector) are compute-bound, not memory-bound; historically PAR=4 caused
+65 % `plan_fail` on RTX 5090 from cuRobo GPU contention. The A6000's fewer
+cores likely makes contention **worse**, not better, even with more VRAM.
+
+Recommendation:
+
+| `PAR` | Sustained VRAM | Peak VRAM | Wall time (9 obj) | Risk |
+|-------|----------------|-----------|-------------------|------|
+| 1     | ~4 GB          | ~7 GB     | ~9-12 h           | None |
+| **2** | ~8 GB          | ~14-18 GB | **~5-7 h**        | **Validated, default** |
+| 3     | ~12 GB         | ~21-27 GB | ~4-5 h (if no contention drop) | Unknown — VRAM fine, cuRobo contention untested on A6000 |
+| 4     | ~16 GB         | ~28-36 GB | ~3-4 h (theoretical) | High — 65 % plan_fail on 5090; expect same or worse |
+
+**Start with PAR=2.** After object #1 finishes, look at its `DONE_RETRY` line
+in `RESULTS` and compare `plan_fail` ratio to dev-box reference (35-40 % is
+normal). If plan_fail is comparable, you can bump `PAR=3` for the rest (the
+orchestrator re-reads PAR on each loop iteration, but bumping mid-run requires
+killing + restarting the orchestrator — it'll resume from where it stopped).
+
+---
+
+## What to watch for — *specific* gotchas verified for this setup
+
+These are not theoretical — every one was either a confirmed bug we fixed or an
+explicit non-default the partner needs to know about. Cross-checked against the
+actual code, not just docs.
+
+| # | Gotcha | Status / Action |
+|---|--------|-----------------|
+| 1 | **Object mass = 0.05 kg** is hardcoded in `sim/run_grasp_sim_baseline3_v4.py` L962. Real per-class mass (0.3–0.6 kg) caused PhysX corruption + DP3 eval SR = 0 % when we tried it earlier. **Do not override.** | ✅ Already baked in, no action |
+| 2 | **`--yaw-aug` is True by default**. Producing orig + yaw90 + yaw180 + yaw270 = 4 attempts per source ep. The retry wrapper assumes this (`ATTEMPTS_PER_EP=4`). If you change one without the other, the case-C retry math (`MARKERS / ATTEMPTS_PER_EP`) breaks. | ✅ Defaults aligned, no action |
+| 3 | **`OUT` env var must match the existing collected dir** (`episodes_b3_v4_15obj_3yaw_2026-05-25`), NOT a fresh date — otherwise resume can't see the 6 done objects' hdf5 and the safety check refuses. | ⚠️ Use the env var shown above |
+| 4 | **`RESULTS` env var must point to the shipped log**. If you let it default to `/tmp/v4_full15_results_<TODAY>.txt`, that file won't exist → orchestrator will treat all 15 objects as "not done" and re-collect the 6. | ⚠️ Use the env var shown above |
+| 5 | **`PROJ` and `PY` env vars** override the hardcoded `/home/accelerator/...` defaults baked into the scripts for the dev box. **Always set these on partner machine.** | ⚠️ Use the env vars shown above |
+| 6 | **`scripts/baseline_3_v4/` lives in repo** (gate3-curobo-ik). The orchestrator launches `bash scripts/baseline_3_v4/v4_chunked_with_retry.sh ...` with a relative path — so it must be run from repo root. `cd UCB_Project` before launching. | ⚠️ Run from repo root |
+| 7 | **PAR=2 default** (overridable via env var). 4-par historically broke on RTX 5090 (65 % plan_fail). On A6000, 3-par is plausibly OK due to extra VRAM, but cuRobo compute contention is the binding constraint and A6000 has 50 % fewer CUDA cores — so 3-par is **untested**, may degrade. See "Choosing PAR on A6000" above. | ⚠️ Default 2; bump to 3 only after watching first object's plan_fail rate |
+| 8 | **dex_19 (large_clamp) has 0 source ep**. DexYCB raw simply doesn't contain large_clamp grasp sessions. The orchestrator will launch ycb_dex_19, find 0 hdf5 in `episodes_g/`, and the collector will silently emit nothing. Not a bug. | ✅ Already documented in script |
+| 9 | **Logs from prior runs in `/tmp/v4_obj_NN_*.out`** would contaminate stats. The orchestrator deletes per-object logs only for objects it's about to (re)launch — so done objects keep their original logs. | ✅ Handled |
+| 10 | **First boot is slow**: each new IsaacSim process takes ~30 sec to load all extensions; you'll see thousands of `Failed to create change watch ... errno=28/No space left on device` errors that are **non-fatal** (inotify watch limit; ignore). | ⚠️ Don't panic |
+
+---
+
+## Object-by-object expectation (based on dev-box run)
+
+The first 6 (already done, will be skipped):
+
+| obj | yield (saved) | notes |
+|---|---|---|
+| 06 tuna | 25 ep | clean (0 PhysX warn) |
+| 04 tomato | 29 | normal |
+| 07 pudding | 31 | normal |
+| 08 gelatin | 22 | normal |
+| 09 potted_meat | 36 | normal |
+| 18 marker | 10 | low yield (thin object) |
+
+Expect the remaining 9 to behave similarly: ~20-35 ep per object, ~50-60 min each at 2-parallel. Total ~5-7 h for the batch.
+
+dex_19 will produce 0 ep (no source data). Don't be alarmed by `0 saved` for it.
+
+---
+
+## After the run
+
+A final summary block appears at the tail of `RESULTS`:
+
+```
+=== FINAL PER-OBJ SUMMARY (after resume) ===
+  ycb_dex_06: orig=10  yaw=15  total=25
+  ycb_dex_04: orig=12  yaw=17  total=29
+  ...
+  TOTAL: NNN trajectories
+```
+
+Send that summary back to the dev box, and we'll handle zarr build + DP3
+training. Do **NOT** delete the collected hdf5 dir until then — it's the
+training input.
+
+---
+
+## Troubleshooting
+
+- **`ImportError: ...isaacsim`** → wrong conda env activated. Confirm with `which python` then `python -c "from isaacsim import SimulationApp"`.
+- **GPU OOM during 2-parallel** → drop one object out of `OBJS` list temporarily; or run sequentially by editing the orchestrator's `while [[ $(jobs -rp | wc -l) -ge 2 ]]` line → `-ge 1`.
+- **Every object's `plan_fail` count is huge (>80 %)** → likely cuRobo install issue. Run a single object first and grep its log for cuRobo errors. The collector falls back to legacy synthesize+IK if cuRobo plan fails, but ideally cuRobo should work.
+- **Orchestrator launches all 15 objects (didn't skip)** → `RESULTS` env var didn't get exported properly. Confirm `echo "$RESULTS"` inside the same shell where you launch.
+- **Resume hdf5 saved=0 for everything** → `OUT` env var wrong or directory empty; safety check should have caught this but if you see this pattern, abort and verify paths.
+
+---
+
+## License & citation
+
+Sim collection assets: CC-BY-4.0.
+DexYCB source data subject to the original DexYCB license.
+YCB object meshes: Yale-CMU-Berkeley YCB Object & Model Set license.

episodes_b3_v4_15obj_3yaw_2026-05-25.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cffd0d1102541258b064cbc105bf8c98e2bab2cc74dcc4e7bde0d1fb0727b14e
+size 277981293

episodes_g.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c05ff938f1efd6692031b706842ec818b9060b72e493d7efe2f8544074768f68
+size 1285206754

obj_usd_cad_ycb.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c64d81851206b3b33c84f335413d8720e49b1873ab9f9cfd4064cea66d00d89
+size 512676522

v4_chunked_with_retry.sh

@@ -0,0 +1,106 @@
+#!/usr/bin/env bash
+# Per-object chunked-retry wrapper for baseline_3 v4 collector.
+#
+# Each ep with --yaw-aug ON produces 4 attempts (orig + yaw90/180/270),
+# each printing exactly ONE terminal marker ("GRASPED + LIFTED" /
+# "object Z … not lifted" / "plan sequence failed" / "abort ep").
+#
+# Handles 3 chunk-end cases:
+#   A. chunk fully complete                       → no retry
+#   B. sanity FAIL mid-chunk (EARLY-EXIT printed) → parse "remaining N eps", retry the tail
+#   C. process crash with no EARLY-EXIT marker    → count terminal markers vs LIMIT, retry diff
+#
+# Usage:
+#   bash v4_chunked_with_retry.sh OBJ_CODE CHUNK_SIZE OUT_DIR
+#
+# Env vars (all overridable):
+#   PY            python interpreter with env_isaaclab (IsaacSim 5.1 + cuRobo 0.8)
+#   PROJ          UCB_Project root
+#   RESULTS_FILE  per-run results log (DONE_RETRY etc. appended here)
+#   N_TOTAL       total source eps in episodes_g/ per object (default 50)
+#   NO_YAW_AUG=1  disable yaw aug → 1 attempt/ep instead of 4
+set -u
+
+# --- Configurable paths (override via env) -----------------------------------
+PY=${PY:-/home/accelerator/miniforge3/envs/env_isaaclab/bin/python}
+PROJ=${PROJ:-/home/accelerator/UCB_Project}
+RESULTS=${RESULTS_FILE:-/tmp/v4_results.txt}
+N_TOTAL=${N_TOTAL:-50}
+# -----------------------------------------------------------------------------
+
+NN=$1; CHUNK=${2:-5}; OUT=${3:-Baseline1/data/episodes_b3_v4_2par_c5}
+cd "$PROJ"; mkdir -p "$OUT"
+
+# v4 collector default --yaw-aug=True → 4 attempts/ep (orig + 3 yaws).
+# Pass NO_YAW_AUG=1 to use 1 attempt/ep instead.
+if [[ "${NO_YAW_AUG:-0}" == "1" ]]; then
+    ATTEMPTS_PER_EP=1
+    YAW_FLAG="--no-yaw-aug"
+else
+    ATTEMPTS_PER_EP=4
+    YAW_FLAG=""   # default ON, no flag needed
+fi
+
+run_chunk() {
+    local NN=$1 START=$2 LIMIT=$3 TAG=$4
+    local LOG=/tmp/v4_obj_${NN}_${TAG}.out
+    "$PY" -u sim/run_grasp_sim_baseline3_v4.py \
+        --object "ycb_dex_${NN}" --headless \
+        --start "$START" --limit "$LIMIT" \
+        --out-dir "$OUT" \
+        $YAW_FLAG \
+        > "$LOG" 2>&1
+
+    # case B: sanity FAIL → parse "remaining N eps" emitted by collector
+    if grep -q "EARLY-EXIT" "$LOG"; then
+        REMAINING=$(grep -oE "remaining [0-9]+ eps" "$LOG" | head -1 | grep -oE "[0-9]+")
+        if [[ -n "$REMAINING" && "$REMAINING" -gt 0 ]]; then
+            echo "$REMAINING"
+            return
+        fi
+    fi
+
+    # case C: process died without sanity (OOM/crash/kernel-kill).
+    # Each (ep, yaw attempt) prints ONE terminal marker. Compute MISSING by:
+    #   completed_eps = floor(markers / ATTEMPTS_PER_EP)
+    #   MISSING       = LIMIT - completed_eps
+    MARKERS=$(grep -cE "GRASPED \+ LIFTED|object Z .* not lifted|plan sequence failed|abort ep" "$LOG")
+    COMPLETED_EPS=$(( MARKERS / ATTEMPTS_PER_EP ))
+    if [[ "$COMPLETED_EPS" -lt "$LIMIT" ]]; then
+        MISSING=$(( LIMIT - COMPLETED_EPS ))
+        echo "$MISSING"
+        return
+    fi
+
+    # case A: clean finish
+    echo "0"
+}
+
+for ((S=0; S<N_TOTAL; S+=CHUNK)); do
+    L=$((N_TOTAL-S<CHUNK ? N_TOTAL-S : CHUNK))
+    SKIPPED=$(run_chunk "$NN" "$S" "$L" "c${S}")
+    if [[ "$SKIPPED" -gt 0 ]]; then
+        SKIP_START=$((S + L - SKIPPED))
+        SKIP_END=$((S + L))
+        for ((R=SKIP_START; R<SKIP_END; R++)); do
+            echo "  [retry] $NN ep $R (skipped by sanity/crash in chunk $S)" >&2
+            run_chunk "$NN" "$R" "1" "r${R}" > /dev/null
+        done
+    fi
+done
+
+# Aggregate stats across ALL logs for this object (chunks + retries) for the
+# DONE_RETRY line. Note: orig + yaw variants both counted by glob below.
+SAVED_ORIG=$(ls "$OUT"/*ycb_dex_${NN}.hdf5 2>/dev/null | wc -l)
+SAVED_YAW=$(ls "$OUT"/*ycb_dex_${NN}_yaw*.hdf5 2>/dev/null | wc -l)
+SAVED=$(( SAVED_ORIG + SAVED_YAW ))
+
+G=$(cat /tmp/v4_obj_${NN}_*.out 2>/dev/null | grep -cE "GRASPED \+ LIFTED")
+NL=$(cat /tmp/v4_obj_${NN}_*.out 2>/dev/null | grep -cE "object Z .* not lifted")
+PF=$(cat /tmp/v4_obj_${NN}_*.out 2>/dev/null | grep -c "plan sequence failed")
+SAN=$(cat /tmp/v4_obj_${NN}_*.out 2>/dev/null | grep -c "sanity check FAIL")
+AB=$(cat /tmp/v4_obj_${NN}_*.out 2>/dev/null | grep -c "abort ep")
+SF=$(cat /tmp/v4_obj_${NN}_*.out 2>/dev/null | grep -c "object settle FAIL")
+WARN=$(cat /tmp/v4_obj_${NN}_*.out 2>/dev/null | grep -ciE "Invalid PhysX")
+RTRY=$(ls /tmp/v4_obj_${NN}_r*.out 2>/dev/null | wc -l)
+echo "[$(date +%H:%M:%S)] DONE_RETRY ycb_dex_${NN}  saved=${SAVED} (orig=${SAVED_ORIG} yaw=${SAVED_YAW})  GRASPED=${G}  not_lifted=${NL}  plan_fail=${PF}  sanity=${SAN}  abort=${AB}  settle_fail=${SF}  PhysX_warns=${WARN}  retries=${RTRY}" >> "$RESULTS"

v4_full15_queue_resume.sh

@@ -0,0 +1,126 @@
+#!/usr/bin/env bash
+# Resume-capable 2-parallel orchestrator for baseline_3 v4 sim collection.
+# Skips any object already DONE_RETRY in the results log → safe to re-run on
+# a partner machine after rsync-ing partial results.
+#
+# Required env vars (override these for partner setup):
+#   PROJ          UCB_Project root          (default: /home/accelerator/UCB_Project)
+#   PY            env_isaaclab python       (default: /home/accelerator/miniforge3/envs/env_isaaclab/bin/python)
+#   OUT           output dir for hdf5       (default: Baseline1/data/episodes_b3_v4_15obj_3yaw_<TODAY>)
+#   RESULTS       results log path          (default: /tmp/v4_full15_results_<TODAY>.txt)
+#   PAR           concurrent IsaacSim procs (default: 2). On A6000 (48 GB) you can
+#                 try 3 if VRAM headroom is comfortable — but cuRobo subprocess
+#                 contention historically caused 65 % plan_fail at 4-par on RTX 5090.
+#                 A6000 has 50 % fewer CUDA cores so cuRobo contention may be WORSE
+#                 there even with more VRAM. Default 2 is safe; bump to 3 only after
+#                 watching the first 1-2 objects' plan_fail rate in RESULTS.
+#
+# Resume usage (partner side, after copying our data into matching paths):
+#   PROJ=$HOME/UCB_Project \
+#   PY=$HOME/miniconda3/envs/env_isaaclab/bin/python \
+#   OUT=Baseline1/data/episodes_b3_v4_15obj_3yaw_2026-05-25 \
+#   RESULTS=/tmp/v4_full15_results_2026-05-25.txt \
+#   PAR=2 \
+#     bash scripts/baseline_3_v4/v4_full15_queue_resume.sh
+set -u
+
+PROJ=${PROJ:-/home/accelerator/UCB_Project}
+PY=${PY:-/home/accelerator/miniforge3/envs/env_isaaclab/bin/python}
+PAR=${PAR:-2}
+cd "$PROJ"
+
+TODAY=$(date +%Y-%m-%d)
+OUT=${OUT:-Baseline1/data/episodes_b3_v4_15obj_3yaw_${TODAY}}
+mkdir -p "$OUT"
+
+RESULTS=${RESULTS:-/tmp/v4_full15_results_${TODAY}.txt}
+
+# RESUME safety: if OUT already has data, refuse unless RESULTS log is present too
+# (so we know what's already done). Without RESULTS we'd re-collect overwriting hdf5.
+if [[ -n "$(ls -A "$OUT" 2>/dev/null)" && ! -f "$RESULTS" ]]; then
+    echo "ERROR: OUT=$OUT is non-empty but RESULTS=$RESULTS missing." >&2
+    echo "       Either ship the matching RESULTS log alongside OUT, or empty OUT first." >&2
+    exit 1
+fi
+
+# Touch RESULTS (don't truncate — we need the prior DONE_RETRY entries).
+touch "$RESULTS"
+
+# Exported for the per-object wrapper (consumed via $RESULTS_FILE there).
+export RESULTS_FILE="$RESULTS"
+export PY PROJ
+
+# Object list — 15 entries; partner can comment out specific ones to skip.
+OBJS=(
+  06   # tuna             — collected on dev box
+  04   # tomato           — collected
+  07   # pudding          — collected
+  08   # gelatin          — collected
+  18   # marker           — collected
+  09   # potted_meat      — collected
+  12   # bleach
+  05   # mustard
+  03   # sugar
+  15   # drill
+  02   # cracker_box
+  11   # pitcher_base
+  14   # mug
+  19   # large_clamp      — NOTE: no DexYCB raw → src ep dir has 0 ep → collector silently emits nothing
+  20   # extra_large_clamp
+)
+
+GIT_SHA=$(git rev-parse --short HEAD 2>/dev/null || echo "unknown")
+
+# Idempotent MANIFEST refresh.
+cat > "$OUT/MANIFEST.md" <<EOF
+# baseline_3 v4 collection run — last update $(date "+%Y-%m-%d %H:%M:%S %Z")
+
+- git: $GIT_SHA
+- collector: sim/run_grasp_sim_baseline3_v4.py
+- wrapper:   scripts/baseline_3_v4/v4_chunked_with_retry.sh (chunk-5, retry)
+- orchestrator: scripts/baseline_3_v4/v4_full15_queue_resume.sh (RESUME-aware, 2-parallel)
+- yaw config: 3 yaws/source (orig + 90 + 180 + 270 = 4 attempts/ep)
+- object mass: 0.05 kg (hardcoded in collector, see Baseline1/RETRAIN_V4_FULL12.md history)
+- objects (${#OBJS[@]}): $(IFS=,; echo "${OBJS[*]}")
+- results log: $RESULTS
+
+File naming inside OUT:
+  <source_ep>__ycb_dex_NN.hdf5          original yaw
+  <source_ep>__ycb_dex_NN_yaw{90,180,270}.hdf5   augmented
+EOF
+
+echo "[$(date +%H:%M:%S)] RESUME_START 15 obj 2-parallel chunk-5 yaw=all3  OUT=$OUT  GIT=$GIT_SHA" >> "$RESULTS"
+
+launched=0; skipped=0
+for NN in "${OBJS[@]}"; do
+    if grep -q "DONE_RETRY ycb_dex_${NN} " "$RESULTS" 2>/dev/null; then
+        echo "[$(date +%H:%M:%S)] SKIP ycb_dex_${NN} (already DONE_RETRY)" >> "$RESULTS"
+        skipped=$((skipped + 1))
+        continue
+    fi
+
+    # Clean stale per-object logs so wrapper's stats aren't contaminated.
+    rm -f /tmp/v4_obj_${NN}_*.out /tmp/v4_orchestrator_${NN}.out
+
+    # Concurrency cap (default 2; override via PAR=N).
+    while [[ $(jobs -rp | wc -l) -ge $PAR ]]; do sleep 10; done
+    echo "[$(date +%H:%M:%S)] LAUNCH ycb_dex_${NN}" >> "$RESULTS"
+    bash scripts/baseline_3_v4/v4_chunked_with_retry.sh "$NN" 5 "$OUT" \
+        > /tmp/v4_orchestrator_${NN}.out 2>&1 &
+    launched=$((launched + 1))
+done
+wait
+
+echo "[$(date +%H:%M:%S)] QUEUE_RESUME_DONE  launched=${launched}  skipped=${skipped}" >> "$RESULTS"
+
+{
+echo ""
+echo "=== FINAL PER-OBJ SUMMARY (after resume) ==="
+for NN in "${OBJS[@]}"; do
+    O=$(ls "$OUT"/*ycb_dex_${NN}.hdf5 2>/dev/null | wc -l)
+    Y=$(ls "$OUT"/*ycb_dex_${NN}_yaw*.hdf5 2>/dev/null | wc -l)
+    echo "  ycb_dex_${NN}: orig=${O}  yaw=${Y}  total=$((O+Y))"
+done
+T=$(ls "$OUT"/*.hdf5 2>/dev/null | wc -l)
+echo "  TOTAL: $T trajectories"
+} >> "$RESULTS"

v4_full15_results_2026-05-25.txt

@@ -0,0 +1,21 @@
+[15:28:11] QUEUE_START 15 obj 2-parallel chunk-5 yaw=all3 (orig+90+180+270)
+  OUT=Baseline1/data/episodes_b3_v4_15obj_3yaw_2026-05-25
+  GIT_SHA=9da1164
+[15:28:11] LAUNCH ycb_dex_06
+[15:28:11] LAUNCH ycb_dex_04
+[16:31:40] DONE_RETRY ycb_dex_06  saved=25 (orig=10 yaw=15)  GRASPED=25  not_lifted=107  plan_fail=68  sanity=0  abort=0  settle_fail=0  PhysX_warns=0  retries=0
+[16:31:41] LAUNCH ycb_dex_07
+[16:31:48] DONE_RETRY ycb_dex_04  saved=29 (orig=12 yaw=17)  GRASPED=29  not_lifted=101  plan_fail=63  sanity=4  abort=4  settle_fail=0  PhysX_warns=880  retries=7
+[16:31:51] LAUNCH ycb_dex_08
+[17:34:57] DONE_RETRY ycb_dex_08  saved=22 (orig=10 yaw=12)  GRASPED=22  not_lifted=112  plan_fail=60  sanity=3  abort=2  settle_fail=0  PhysX_warns=2310  retries=2
+[17:35:02] LAUNCH ycb_dex_18
+[17:35:37] DONE_RETRY ycb_dex_07  saved=31 (orig=9 yaw=22)  GRASPED=31  not_lifted=98  plan_fail=64  sanity=4  abort=6  settle_fail=0  PhysX_warns=1001  retries=6
+[17:35:42] LAUNCH ycb_dex_09
+[18:29:14] DONE_RETRY ycb_dex_18  saved=10 (orig=3 yaw=7)  GRASPED=10  not_lifted=91  plan_fail=95  sanity=2  abort=2  settle_fail=0  PhysX_warns=506  retries=3
+[18:29:23] LAUNCH ycb_dex_12
+[18:38:34] DONE_RETRY ycb_dex_09  saved=36 (orig=12 yaw=24)  GRASPED=37  not_lifted=94  plan_fail=62  sanity=5  abort=4  settle_fail=0  PhysX_warns=2783  retries=7
+
+[19:33:11] === RESUME after pgrp-SIGHUP at ~18:52 ===
+[19:33:11] RESUME_START 9 obj: 12 05 03 15 02 11 14 19 20
+[19:33:11] LAUNCH ycb_dex_12
+[19:33:11] LAUNCH ycb_dex_05