| set -Eeuo pipefail | |
| exec "/home/harness_user_1/dgx_ai_factory/venvs/devstral_v64_isolated_base_v6/bin/python" -m vllm.entrypoints.cli.main serve "/home/harness_user_1/dgx_ai_factory/models/devstral_small2_24b_v64_overlay_v6" --host 192.0.2.11 --port 8031 --served-model-name devstral-small2-24b --max-model-len 16384 --gpu-memory-utilization 0.65 --max-num-seqs 1 --limit-mm-per-prompt '{"image":0}' --generation-config vllm --uvicorn-log-level warning --no-enable-log-requests | |