Spaces:
Runtime error
Runtime error
da03
commited on
Commit
Β·
3085e5d
1
Parent(s):
9c3b17f
- start_system.sh +45 -2
start_system.sh
CHANGED
|
@@ -2,8 +2,37 @@
|
|
| 2 |
|
| 3 |
# Multi-GPU Neural OS Startup Script
|
| 4 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
# Default values
|
| 6 |
-
NUM_GPUS
|
| 7 |
DISPATCHER_PORT=8000
|
| 8 |
|
| 9 |
# Parse command line arguments
|
|
@@ -19,8 +48,15 @@ while [[ $# -gt 0 ]]; do
|
|
| 19 |
;;
|
| 20 |
-h|--help)
|
| 21 |
echo "Usage: $0 [--num-gpus N] [--port PORT]"
|
| 22 |
-
echo " --num-gpus N Number of GPU workers to start (default:
|
| 23 |
echo " --port PORT Dispatcher port (default: 8000)"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
exit 0
|
| 25 |
;;
|
| 26 |
*)
|
|
@@ -58,6 +94,13 @@ trap cleanup SIGINT SIGTERM
|
|
| 58 |
|
| 59 |
echo "π Starting Multi-GPU Neural OS System"
|
| 60 |
echo "========================================"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
echo "π Number of GPUs: $NUM_GPUS"
|
| 62 |
echo "π Dispatcher port: $DISPATCHER_PORT"
|
| 63 |
echo "π» Worker ports: $(seq -s', ' 8001 $((8000 + NUM_GPUS)))"
|
|
|
|
| 2 |
|
| 3 |
# Multi-GPU Neural OS Startup Script
|
| 4 |
|
| 5 |
+
# Function to detect number of GPUs automatically
|
| 6 |
+
detect_gpu_count() {
|
| 7 |
+
if command -v nvidia-smi >/dev/null 2>&1; then
|
| 8 |
+
# Use nvidia-smi to count GPUs
|
| 9 |
+
local gpu_count=$(nvidia-smi -L 2>/dev/null | wc -l)
|
| 10 |
+
if [ "$gpu_count" -gt 0 ]; then
|
| 11 |
+
echo "$gpu_count"
|
| 12 |
+
return 0
|
| 13 |
+
fi
|
| 14 |
+
fi
|
| 15 |
+
|
| 16 |
+
# If nvidia-smi fails, try alternative methods
|
| 17 |
+
if [ -d "/proc/driver/nvidia/gpus" ]; then
|
| 18 |
+
local gpu_count=$(ls -d /proc/driver/nvidia/gpus/*/information 2>/dev/null | wc -l)
|
| 19 |
+
if [ "$gpu_count" -gt 0 ]; then
|
| 20 |
+
echo "$gpu_count"
|
| 21 |
+
return 0
|
| 22 |
+
fi
|
| 23 |
+
fi
|
| 24 |
+
|
| 25 |
+
# Default fallback
|
| 26 |
+
echo "1"
|
| 27 |
+
return 1
|
| 28 |
+
}
|
| 29 |
+
|
| 30 |
+
# Detect GPU count automatically
|
| 31 |
+
DETECTED_GPUS=$(detect_gpu_count)
|
| 32 |
+
GPU_DETECTION_SUCCESS=$?
|
| 33 |
+
|
| 34 |
# Default values
|
| 35 |
+
NUM_GPUS=$DETECTED_GPUS
|
| 36 |
DISPATCHER_PORT=8000
|
| 37 |
|
| 38 |
# Parse command line arguments
|
|
|
|
| 48 |
;;
|
| 49 |
-h|--help)
|
| 50 |
echo "Usage: $0 [--num-gpus N] [--port PORT]"
|
| 51 |
+
echo " --num-gpus N Number of GPU workers to start (default: auto-detected)"
|
| 52 |
echo " --port PORT Dispatcher port (default: 8000)"
|
| 53 |
+
echo ""
|
| 54 |
+
echo "GPU Detection:"
|
| 55 |
+
echo " Automatically detects available GPUs using nvidia-smi"
|
| 56 |
+
echo " Currently detected: $DETECTED_GPUS GPU(s)"
|
| 57 |
+
if [ $GPU_DETECTION_SUCCESS -ne 0 ]; then
|
| 58 |
+
echo " β οΈ GPU detection failed - using fallback of 1 GPU"
|
| 59 |
+
fi
|
| 60 |
exit 0
|
| 61 |
;;
|
| 62 |
*)
|
|
|
|
| 94 |
|
| 95 |
echo "π Starting Multi-GPU Neural OS System"
|
| 96 |
echo "========================================"
|
| 97 |
+
echo "π GPU Detection: $DETECTED_GPUS GPU(s) detected"
|
| 98 |
+
if [ $GPU_DETECTION_SUCCESS -ne 0 ]; then
|
| 99 |
+
echo "β οΈ GPU detection failed - using fallback count"
|
| 100 |
+
elif command -v nvidia-smi >/dev/null 2>&1; then
|
| 101 |
+
echo "π Detected GPUs:"
|
| 102 |
+
nvidia-smi -L 2>/dev/null | sed 's/^/ /'
|
| 103 |
+
fi
|
| 104 |
echo "π Number of GPUs: $NUM_GPUS"
|
| 105 |
echo "π Dispatcher port: $DISPATCHER_PORT"
|
| 106 |
echo "π» Worker ports: $(seq -s', ' 8001 $((8000 + NUM_GPUS)))"
|