ctt train-calibration hygiene 2026-07-03T16:31:19Z: scripts/slurm/eval_ctt_generated_rollout.sbatch
Browse files
scripts/slurm/eval_ctt_generated_rollout.sbatch
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/bin/bash
|
| 2 |
+
#SBATCH --job-name=ctt_ms_rollout
|
| 3 |
+
#SBATCH --account=def-yalda
|
| 4 |
+
#SBATCH --nodes=1
|
| 5 |
+
#SBATCH --ntasks=1
|
| 6 |
+
#SBATCH --cpus-per-task=4
|
| 7 |
+
#SBATCH --mem=24G
|
| 8 |
+
#SBATCH --time=01:00:00
|
| 9 |
+
#SBATCH --array=0-0
|
| 10 |
+
#SBATCH --output=outputs/hpc/logs/%x_%A_%a.out
|
| 11 |
+
#SBATCH --error=outputs/hpc/logs/%x_%A_%a.err
|
| 12 |
+
|
| 13 |
+
set -euo pipefail
|
| 14 |
+
|
| 15 |
+
PROJECT_DIR="${PROJECT_DIR:-$SLURM_SUBMIT_DIR}"
|
| 16 |
+
CHECKPOINT="${CHECKPOINT:?Set CHECKPOINT to a CTT model.pt}"
|
| 17 |
+
SOURCE_INDEX="${SOURCE_INDEX:-data/cil_charts/train/index.json}"
|
| 18 |
+
TARGET_INDEX="${TARGET_INDEX:-data/cil_charts/val/index.json}"
|
| 19 |
+
OUT_DIR="${OUT_DIR:-runs/ctt_residual_rollout_smoke_seed${SLURM_ARRAY_TASK_ID:-0}}"
|
| 20 |
+
|
| 21 |
+
SCRATCH_ROOT="/scratch/$USER/dovla"
|
| 22 |
+
SIF="${SIF:-$SCRATCH_ROOT/containers/pytorch_2.7.1_cuda12.8.sif}"
|
| 23 |
+
PYTHON="${PYTHON:-$SCRATCH_ROOT/envs/maniskill/bin/python}"
|
| 24 |
+
NATIVE_LIBS="$SCRATCH_ROOT/native_libs/lib"
|
| 25 |
+
CPU_RENDER_LIBS="$SCRATCH_ROOT/cpu_render_libs"
|
| 26 |
+
CA_BUNDLE="$SCRATCH_ROOT/ca-bundle.crt"
|
| 27 |
+
VULKAN_ICD="$CPU_RENDER_LIBS/share/vulkan/icd.d/lvp_icd.x86_64.json"
|
| 28 |
+
|
| 29 |
+
K="${K:-8}"
|
| 30 |
+
POOL_SIZE="${POOL_SIZE:-0}"
|
| 31 |
+
NEIGHBORS="${NEIGHBORS:-8}"
|
| 32 |
+
MAX_TARGET_CHARTS="${MAX_TARGET_CHARTS:-8}"
|
| 33 |
+
GROUP_BATCH_SIZE="${GROUP_BATCH_SIZE:-1}"
|
| 34 |
+
DEVICE="${DEVICE:-cpu}"
|
| 35 |
+
SIM_BACKEND="${SIM_BACKEND:-physx_cpu}"
|
| 36 |
+
RENDER_BACKEND="${RENDER_BACKEND:-cpu}"
|
| 37 |
+
RESTORE_TOLERANCE="${RESTORE_TOLERANCE:-1e-5}"
|
| 38 |
+
DELTA_SCALE="${DELTA_SCALE:-1.0}"
|
| 39 |
+
BOOTSTRAP_SAMPLES="${BOOTSTRAP_SAMPLES:-200}"
|
| 40 |
+
INCLUDE_TARGETS_WITHOUT_POSITIVES="${INCLUDE_TARGETS_WITHOUT_POSITIVES:-0}"
|
| 41 |
+
EXCLUDE_SELF_SOURCE="${EXCLUDE_SELF_SOURCE:-0}"
|
| 42 |
+
SKIP_METRICS="${SKIP_METRICS:-0}"
|
| 43 |
+
|
| 44 |
+
module load StdEnv/2023 apptainer/1.4.5
|
| 45 |
+
cd "$PROJECT_DIR"
|
| 46 |
+
mkdir -p outputs/hpc/logs "$OUT_DIR"
|
| 47 |
+
|
| 48 |
+
RUNTIME_DIR="/tmp/$USER/ctt-ms-rollout-$SLURM_JOB_ID-${SLURM_ARRAY_TASK_ID:-0}"
|
| 49 |
+
CACHE_DIR="/tmp/$USER/ctt-ms-rollout-mesa-$SLURM_JOB_ID-${SLURM_ARRAY_TASK_ID:-0}"
|
| 50 |
+
mkdir -p "$RUNTIME_DIR" "$CACHE_DIR"
|
| 51 |
+
chmod 700 "$RUNTIME_DIR"
|
| 52 |
+
|
| 53 |
+
export OMP_NUM_THREADS=1
|
| 54 |
+
export OPENBLAS_NUM_THREADS=1
|
| 55 |
+
export MKL_NUM_THREADS=1
|
| 56 |
+
export DOVLA_TORCH_THREADS=1
|
| 57 |
+
|
| 58 |
+
EXTRA_ARGS=()
|
| 59 |
+
if [[ "$INCLUDE_TARGETS_WITHOUT_POSITIVES" == "1" ]]; then
|
| 60 |
+
EXTRA_ARGS+=(--include-targets-without-positives)
|
| 61 |
+
fi
|
| 62 |
+
if [[ "$EXCLUDE_SELF_SOURCE" == "1" ]]; then
|
| 63 |
+
EXTRA_ARGS+=(--exclude-self-source)
|
| 64 |
+
fi
|
| 65 |
+
if [[ "$SKIP_METRICS" == "1" ]]; then
|
| 66 |
+
EXTRA_ARGS+=(--skip-metrics)
|
| 67 |
+
fi
|
| 68 |
+
|
| 69 |
+
apptainer exec \
|
| 70 |
+
--env "LD_LIBRARY_PATH=$CPU_RENDER_LIBS/lib:$NATIVE_LIBS,VK_ICD_FILENAMES=$VULKAN_ICD,VK_DRIVER_FILES=$VULKAN_ICD,XDG_RUNTIME_DIR=$RUNTIME_DIR,MESA_SHADER_CACHE_DIR=$CACHE_DIR,LIBGL_ALWAYS_SOFTWARE=1,LP_NUM_THREADS=1,SSL_CERT_FILE=$CA_BUNDLE,REQUESTS_CA_BUNDLE=$CA_BUNDLE,OMP_NUM_THREADS=1,OPENBLAS_NUM_THREADS=1,MKL_NUM_THREADS=1,DOVLA_TORCH_THREADS=1,MPLBACKEND=Agg,PYTHONDONTWRITEBYTECODE=1" \
|
| 71 |
+
-B "$PROJECT_DIR:$PROJECT_DIR" \
|
| 72 |
+
-B "/scratch/$USER:/scratch/$USER" \
|
| 73 |
+
"$SIF" "$PYTHON" scripts/eval_ctt_generated_rollout.py \
|
| 74 |
+
--checkpoint "$CHECKPOINT" \
|
| 75 |
+
--source-index "$SOURCE_INDEX" \
|
| 76 |
+
--target-index "$TARGET_INDEX" \
|
| 77 |
+
--out-dir "$OUT_DIR" \
|
| 78 |
+
--k "$K" \
|
| 79 |
+
--pool-size "$POOL_SIZE" \
|
| 80 |
+
--neighbors "$NEIGHBORS" \
|
| 81 |
+
--max-target-charts "$MAX_TARGET_CHARTS" \
|
| 82 |
+
--group-batch-size "$GROUP_BATCH_SIZE" \
|
| 83 |
+
--device "$DEVICE" \
|
| 84 |
+
--sim-backend "$SIM_BACKEND" \
|
| 85 |
+
--render-backend "$RENDER_BACKEND" \
|
| 86 |
+
--restore-tolerance "$RESTORE_TOLERANCE" \
|
| 87 |
+
--delta-scale "$DELTA_SCALE" \
|
| 88 |
+
--bootstrap-samples "$BOOTSTRAP_SAMPLES" \
|
| 89 |
+
"${EXTRA_ARGS[@]}"
|