Datasets:

zhicao
/

openpi

#!/bin/bash
# Parallel reprocessing script for updating existing DROID .npz files
# Distributes files across 5 machines, each with 8 GPUs (40 GPUs total)
#
# Usage:
#   bash run_reprocessing_8gpu.sh <MACHINE_ID> [--dry-run]
#
# MACHINE_ID: 0, 1, 2, 3, or 4

# Parse arguments
if [ $# -eq 0 ]; then
    echo "Error: Missing machine ID argument"
    echo "Usage: bash run_reprocessing_8gpu.sh <MACHINE_ID> [--dry-run]"
    echo "  MACHINE_ID: 0, 1, 2, 3, or 4"
    echo ""
    echo "Distribution (4,976 total files, ~995 per machine):"
    echo "  Machine 0: files 0 to 995"
    echo "  Machine 1: files 995 to 1990"
    echo "  Machine 2: files 1990 to 2985"
    echo "  Machine 3: files 2985 to 3980"
    echo "  Machine 4: files 3980 to 4976"
    exit 1
fi

MACHINE_ID=$1
shift

# Validate machine ID
if [ $MACHINE_ID -lt 0 ] || [ $MACHINE_ID -gt 4 ]; then
    echo "Error: MACHINE_ID must be 0, 1, 2, 3, or 4"
    exit 1
fi

# Parse dry-run flag
DRY_RUN_FLAG=""
if [ "$1" == "--dry-run" ]; then
    DRY_RUN_FLAG="--dry-run"
    echo "DRY RUN MODE - Will check what needs reprocessing without saving"
fi

# Configuration
DATA_DIR="/mnt/kevin/data/droid_processed_1000pts"
TOTAL_FILES=4976
TOTAL_MACHINES=5
NUM_GPUS=8
FILES_PER_MACHINE=$(( TOTAL_FILES / TOTAL_MACHINES ))

# Calculate file range for this machine
MACHINE_START=$(( MACHINE_ID * FILES_PER_MACHINE ))

# Last machine gets remaining files
if [ $MACHINE_ID -eq 4 ]; then
    MACHINE_END=$TOTAL_FILES
else
    MACHINE_END=$(( (MACHINE_ID + 1) * FILES_PER_MACHINE ))
fi

FILES_PER_GPU=$(( (MACHINE_END - MACHINE_START) / NUM_GPUS ))

echo "========================================"
echo "DROID Reprocessing - Machine $MACHINE_ID"
echo "========================================"
echo "  Total files: $TOTAL_FILES"
echo "  Total machines: $TOTAL_MACHINES"
echo "  File range: $MACHINE_START to $MACHINE_END"
echo "  GPUs: $NUM_GPUS"
echo "  Files per GPU: ~$FILES_PER_GPU"
echo "  Data directory: $DATA_DIR"
if [ -n "$DRY_RUN_FLAG" ]; then
    echo "  Mode: DRY RUN"
else
    echo "  Mode: REPROCESS"
fi
echo "========================================"

# Create log directory
mkdir -p $DATA_DIR/reprocessing_logs

# Launch 8 parallel processes
for GPU_ID in {0..7}; do
    START_IDX=$(( MACHINE_START + GPU_ID * FILES_PER_GPU ))
    END_IDX=$(( MACHINE_START + (GPU_ID + 1) * FILES_PER_GPU ))

    # Last GPU gets remaining files
    if [ $GPU_ID -eq 7 ]; then
        END_IDX=$MACHINE_END
    fi

    echo "Launching GPU $GPU_ID: files $START_IDX to $END_IDX"

    CUDA_VISIBLE_DEVICES=$GPU_ID \
    /mnt/kevin/envs/miniconda3/envs/atm_ati_vdm_droid/bin/python \
        scripts/reprocess_existing_npz.py \
        --data-dir $DATA_DIR \
        --start-idx $START_IDX \
        --end-idx $END_IDX \
        --gpu-id $GPU_ID \
        $DRY_RUN_FLAG \
        > $DATA_DIR/reprocessing_logs/m${MACHINE_ID}_gpu${GPU_ID}.log 2>&1 &
done

echo ""
echo "All 8 GPU processes launched on Machine $MACHINE_ID!"
echo ""
echo "Monitor progress:"
echo "  tail -f $DATA_DIR/reprocessing_logs/m${MACHINE_ID}_gpu0.log"
echo ""
echo "Check completion:"
echo "  grep 'Reprocessing Complete' $DATA_DIR/reprocessing_logs/m${MACHINE_ID}_*.log | wc -l"
echo ""
echo "Stop all:"
echo "  pkill -f reprocess_existing_npz"
echo ""
echo "Machine breakdown:"
echo "  Machine 0: files 0 → 995 (995 files)"
echo "  Machine 1: files 995 → 1990 (995 files)"
echo "  Machine 2: files 1990 → 2985 (995 files)"
echo "  Machine 3: files 2985 → 3980 (995 files)"
echo "  Machine 4: files 3980 → 4976 (996 files)"
echo ""

	#!/bin/bash
	# Parallel reprocessing script for updating existing DROID .npz files
	# Distributes files across 5 machines, each with 8 GPUs (40 GPUs total)
	#
	# Usage:
	# bash run_reprocessing_8gpu.sh <MACHINE_ID> [--dry-run]
	#
	# MACHINE_ID: 0, 1, 2, 3, or 4

	# Parse arguments
	if [ $# -eq 0 ]; then
	echo "Error: Missing machine ID argument"
	echo "Usage: bash run_reprocessing_8gpu.sh <MACHINE_ID> [--dry-run]"
	echo " MACHINE_ID: 0, 1, 2, 3, or 4"
	echo ""
	echo "Distribution (4,976 total files, ~995 per machine):"
	echo " Machine 0: files 0 to 995"
	echo " Machine 1: files 995 to 1990"
	echo " Machine 2: files 1990 to 2985"
	echo " Machine 3: files 2985 to 3980"
	echo " Machine 4: files 3980 to 4976"
	exit 1
	fi

	MACHINE_ID=$1
	shift

	# Validate machine ID
	if [ $MACHINE_ID -lt 0 ] \|\| [ $MACHINE_ID -gt 4 ]; then
	echo "Error: MACHINE_ID must be 0, 1, 2, 3, or 4"
	exit 1
	fi

	# Parse dry-run flag
	DRY_RUN_FLAG=""
	if [ "$1" == "--dry-run" ]; then
	DRY_RUN_FLAG="--dry-run"
	echo "DRY RUN MODE - Will check what needs reprocessing without saving"
	fi

	# Configuration
	DATA_DIR="/mnt/kevin/data/droid_processed_1000pts"
	TOTAL_FILES=4976
	TOTAL_MACHINES=5
	NUM_GPUS=8
	FILES_PER_MACHINE=$(( TOTAL_FILES / TOTAL_MACHINES ))

	# Calculate file range for this machine
	MACHINE_START=$(( MACHINE_ID * FILES_PER_MACHINE ))

	# Last machine gets remaining files
	if [ $MACHINE_ID -eq 4 ]; then
	MACHINE_END=$TOTAL_FILES
	else
	MACHINE_END=$(( (MACHINE_ID + 1) * FILES_PER_MACHINE ))
	fi

	FILES_PER_GPU=$(( (MACHINE_END - MACHINE_START) / NUM_GPUS ))

	echo "========================================"
	echo "DROID Reprocessing - Machine $MACHINE_ID"
	echo "========================================"
	echo " Total files: $TOTAL_FILES"
	echo " Total machines: $TOTAL_MACHINES"
	echo " File range: $MACHINE_START to $MACHINE_END"
	echo " GPUs: $NUM_GPUS"
	echo " Files per GPU: ~$FILES_PER_GPU"
	echo " Data directory: $DATA_DIR"
	if [ -n "$DRY_RUN_FLAG" ]; then
	echo " Mode: DRY RUN"
	else
	echo " Mode: REPROCESS"
	fi
	echo "========================================"

	# Create log directory
	mkdir -p $DATA_DIR/reprocessing_logs

	# Launch 8 parallel processes
	for GPU_ID in {0..7}; do
	START_IDX=$(( MACHINE_START + GPU_ID * FILES_PER_GPU ))
	END_IDX=$(( MACHINE_START + (GPU_ID + 1) * FILES_PER_GPU ))

	# Last GPU gets remaining files
	if [ $GPU_ID -eq 7 ]; then
	END_IDX=$MACHINE_END
	fi

	echo "Launching GPU $GPU_ID: files $START_IDX to $END_IDX"

	CUDA_VISIBLE_DEVICES=$GPU_ID \
	/mnt/kevin/envs/miniconda3/envs/atm_ati_vdm_droid/bin/python \
	scripts/reprocess_existing_npz.py \
	--data-dir $DATA_DIR \
	--start-idx $START_IDX \
	--end-idx $END_IDX \
	--gpu-id $GPU_ID \
	$DRY_RUN_FLAG \
	> $DATA_DIR/reprocessing_logs/m${MACHINE_ID}_gpu${GPU_ID}.log 2>&1 &
	done

	echo ""
	echo "All 8 GPU processes launched on Machine $MACHINE_ID!"
	echo ""
	echo "Monitor progress:"
	echo " tail -f $DATA_DIR/reprocessing_logs/m${MACHINE_ID}_gpu0.log"
	echo ""
	echo "Check completion:"
	echo " grep 'Reprocessing Complete' $DATA_DIR/reprocessing_logs/m${MACHINE_ID}_*.log \| wc -l"
	echo ""
	echo "Stop all:"
	echo " pkill -f reprocess_existing_npz"
	echo ""
	echo "Machine breakdown:"
	echo " Machine 0: files 0 → 995 (995 files)"
	echo " Machine 1: files 995 → 1990 (995 files)"
	echo " Machine 2: files 1990 → 2985 (995 files)"
	echo " Machine 3: files 2985 → 3980 (995 files)"
	echo " Machine 4: files 3980 → 4976 (996 files)"
	echo ""