23 lines
553 B
Bash
23 lines
553 B
Bash
|
#!/bin/bash
|
||
|
|
# Launch vLLM-Omni server for Fish Speech S2 Pro
|
||
|
|
#
|
||
|
|
# Usage:
|
||
|
|
# ./run_server.sh
|
||
|
|
# CUDA_VISIBLE_DEVICES=0 ./run_server.sh
|
||
|
|
|
||
|
|
set -e
|
||
|
|
|
||
|
|
MODEL="${MODEL:-fishaudio/s2-pro}"
|
||
|
|
PORT="${PORT:-8091}"
|
||
|
|
|
||
|
|
echo "Starting Fish Speech S2 Pro server with model: $MODEL"
|
||
|
|
|
||
|
|
FLASHINFER_DISABLE_VERSION_CHECK=1 \
|
||
|
|
vllm-omni serve "$MODEL" \
|
||
|
|
--stage-configs-path vllm_omni/model_executor/stage_configs/fish_speech_s2_pro.yaml \
|
||
|
|
--host 0.0.0.0 \
|
||
|
|
--port "$PORT" \
|
||
|
|
--gpu-memory-utilization 0.9 \
|
||
|
|
--trust-remote-code \
|
||
|
|
--enforce-eager \
|
||
|
|
--omni
|