anhtld commited on
Commit
5985d58
·
verified ·
1 Parent(s): ba8b348

ctt measured rollout full validation 2026-07-03T01:27:41Z

Browse files
workspace/scripts/slurm/eval_ctt_generated_rollout.sbatch ADDED
@@ -0,0 +1,85 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ SKIP_METRICS="${SKIP_METRICS:-0}"
42
+
43
+ module load StdEnv/2023 apptainer/1.4.5
44
+ cd "$PROJECT_DIR"
45
+ mkdir -p outputs/hpc/logs "$OUT_DIR"
46
+
47
+ RUNTIME_DIR="/tmp/$USER/ctt-ms-rollout-$SLURM_JOB_ID-${SLURM_ARRAY_TASK_ID:-0}"
48
+ CACHE_DIR="/tmp/$USER/ctt-ms-rollout-mesa-$SLURM_JOB_ID-${SLURM_ARRAY_TASK_ID:-0}"
49
+ mkdir -p "$RUNTIME_DIR" "$CACHE_DIR"
50
+ chmod 700 "$RUNTIME_DIR"
51
+
52
+ export OMP_NUM_THREADS=1
53
+ export OPENBLAS_NUM_THREADS=1
54
+ export MKL_NUM_THREADS=1
55
+ export DOVLA_TORCH_THREADS=1
56
+
57
+ EXTRA_ARGS=()
58
+ if [[ "$INCLUDE_TARGETS_WITHOUT_POSITIVES" == "1" ]]; then
59
+ EXTRA_ARGS+=(--include-targets-without-positives)
60
+ fi
61
+ if [[ "$SKIP_METRICS" == "1" ]]; then
62
+ EXTRA_ARGS+=(--skip-metrics)
63
+ fi
64
+
65
+ apptainer exec \
66
+ --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" \
67
+ -B "$PROJECT_DIR:$PROJECT_DIR" \
68
+ -B "/scratch/$USER:/scratch/$USER" \
69
+ "$SIF" "$PYTHON" scripts/eval_ctt_generated_rollout.py \
70
+ --checkpoint "$CHECKPOINT" \
71
+ --source-index "$SOURCE_INDEX" \
72
+ --target-index "$TARGET_INDEX" \
73
+ --out-dir "$OUT_DIR" \
74
+ --k "$K" \
75
+ --pool-size "$POOL_SIZE" \
76
+ --neighbors "$NEIGHBORS" \
77
+ --max-target-charts "$MAX_TARGET_CHARTS" \
78
+ --group-batch-size "$GROUP_BATCH_SIZE" \
79
+ --device "$DEVICE" \
80
+ --sim-backend "$SIM_BACKEND" \
81
+ --render-backend "$RENDER_BACKEND" \
82
+ --restore-tolerance "$RESTORE_TOLERANCE" \
83
+ --delta-scale "$DELTA_SCALE" \
84
+ --bootstrap-samples "$BOOTSTRAP_SAMPLES" \
85
+ "${EXTRA_ARGS[@]}"