| 1 | #!/bin/bash |
| 2 | |
| 3 | MODEL_NAME="deepseek-ai/DeepSeek-V2-Lite" |
| 4 | LOCAL_MODEL_PATH="/models/models--deepseek-ai--DeepSeek-V2-Lite/snapshots/604d5664dddd88a0433dbae533b7fe9472482de0" |
| 5 | HOST="localhost" |
| 6 | PORT=8006 |
| 7 | NUM_PROMPTS=20 |
| 8 | REQUEST_RATE=5 |
| 9 | |
| 10 | # Parse command line arguments |
| 11 | while [[ $# -gt 0 ]]; do |
| 12 | case $1 in |
| 13 | --model) |
| 14 | MODEL_NAME="$2" |
| 15 | shift 2 |
| 16 | ;; |
| 17 | --local-model) |
| 18 | MODEL_NAME=$LOCAL_MODEL_PATH |
| 19 | shift |
| 20 | ;; |
| 21 | --host) |
| 22 | HOST="$2" |
| 23 | shift 2 |
| 24 | ;; |
| 25 | --port) |
| 26 | PORT="$2" |
| 27 | shift 2 |
| 28 | ;; |
| 29 | --num-prompts) |
| 30 | NUM_PROMPTS="$2" |
| 31 | shift 2 |
| 32 | ;; |
| 33 | --request-rate) |
| 34 | REQUEST_RATE="$2" |
| 35 | shift 2 |
| 36 | ;; |
| 37 | -h|--help) |
| 38 | echo "Usage: $0 [OPTIONS]" |
| 39 | echo "Options:" |
| 40 | echo " --model MODEL_NAME Set model name or path (default: deepseek-ai/DeepSeek-V2-Lite)" |
| 41 | echo " --local-model Use local model path (convenience option)" |
| 42 | exit 0 |
| 43 | ;; |
| 44 | *) |
| 45 | echo "Unknown option: $1" |
| 46 | echo "Use -h or --help for usage information" |
| 47 | exit 1 |
| 48 | ;; |
| 49 | esac |
| 50 | done |
| 51 | |
| 52 | vllm bench serve \ |
| 53 | --model "$MODEL_NAME" \ |
| 54 | --host "$HOST" \ |
| 55 | --port "$PORT" \ |
| 56 | --num-prompts "$NUM_PROMPTS" \ |
| 57 | --request-rate "$REQUEST_RATE" |