Lyon
95 lines
3.0 KiB
YAML
95 lines
3.0 KiB
YAML
services:
|
|
model-init:
|
|
image: ${VOICE_IMAGE}
|
|
build: &voice-build
|
|
context: .
|
|
dockerfile: Containerfile
|
|
args:
|
|
BASE_IMAGE: ${BASE_IMAGE}
|
|
SOURCE_REPOSITORY: ${SOURCE_REPOSITORY}
|
|
SOURCE_REF: ${SOURCE_REF}
|
|
SOURCE_SUBMODULES: ${SOURCE_SUBMODULES}
|
|
REFERENCE_WAV_PATH: ${REFERENCE_WAV_PATH}
|
|
REFERENCE_WAV_SHA256: ${REFERENCE_WAV_SHA256}
|
|
PIP_VERSION: ${PIP_VERSION}
|
|
PIP_INDEX_URL: ${PIP_INDEX_URL}
|
|
SETUPTOOLS_INSTALL_SPEC: ${SETUPTOOLS_INSTALL_SPEC}
|
|
TORCH_VERSION_IDENTITY: ${TORCH_VERSION_IDENTITY}
|
|
TORCH_LIBRARY_PATH: ${TORCH_LIBRARY_PATH}
|
|
TORCHAUDIO_VERSION_IDENTITY: ${TORCHAUDIO_VERSION_IDENTITY}
|
|
OPENAI_WHISPER_INSTALL_SPEC: ${OPENAI_WHISPER_INSTALL_SPEC}
|
|
ONNXRUNTIME_GPU_INSTALL_SPEC: ${ONNXRUNTIME_GPU_INSTALL_SPEC}
|
|
ONNXRUNTIME_GPU_INDEX_URL: ${ONNXRUNTIME_GPU_INDEX_URL}
|
|
restart: "no"
|
|
environment:
|
|
MODEL_REPOSITORY: ${MODEL_REPOSITORY}
|
|
MODEL_BRANCH: ${MODEL_BRANCH}
|
|
MODEL_REF: ${MODEL_REF}
|
|
MODEL_REQUIRE_BRANCH_HEAD_MATCH: ${MODEL_REQUIRE_BRANCH_HEAD_MATCH}
|
|
command: [/app/model-init.sh]
|
|
volumes:
|
|
- ${VOICE_CACHE_DIR}/model:/model
|
|
logging: &voice-logs
|
|
driver: json-file
|
|
options:
|
|
max-size: 20m
|
|
max-file: "3"
|
|
|
|
voice:
|
|
image: ${VOICE_IMAGE}
|
|
build: *voice-build
|
|
restart: unless-stopped
|
|
gpus: all
|
|
shm_size: 4gb
|
|
deploy:
|
|
replicas: ${VOICE_REPLICAS}
|
|
depends_on:
|
|
model-init:
|
|
condition: service_completed_successfully
|
|
ports:
|
|
- ${VOICE_BIND_ADDRESS}:${VOICE_PORT}:${VOICE_PORT}
|
|
environment:
|
|
MODEL_ID: ${MODEL_ID}
|
|
MODEL_DIR: /model
|
|
MODEL_REF: ${MODEL_REF}
|
|
SOURCE_REF: ${SOURCE_REF}
|
|
PROMPT_WAV: /opt/CosyVoice/${REFERENCE_WAV_PATH}
|
|
PROMPT_TEXT: ${REFERENCE_PROMPT_TEXT}
|
|
REFERENCE_VOICE_ID: ${REFERENCE_VOICE_ID}
|
|
VOICE_DEVICE: ${VOICE_DEVICE}
|
|
VOICE_FP16: ${VOICE_FP16}
|
|
VOICE_LOAD_TRT: ${VOICE_LOAD_TRT}
|
|
VOICE_LOAD_VLLM: ${VOICE_LOAD_VLLM}
|
|
VOICE_WORKERS: ${VOICE_WORKERS}
|
|
VOICE_CONCURRENCY: ${VOICE_CONCURRENCY}
|
|
VOICE_SPEED_MIN: ${VOICE_SPEED_MIN}
|
|
VOICE_SPEED_MAX: ${VOICE_SPEED_MAX}
|
|
VOICE_PORT: ${VOICE_PORT}
|
|
PYTORCH_CUDA_ALLOC_CONF: ${PYTORCH_CUDA_ALLOC_CONF}
|
|
CUDA_MODULE_LOADING: ${CUDA_MODULE_LOADING}
|
|
LD_LIBRARY_PATH: ${TORCH_LIBRARY_PATH}
|
|
volumes:
|
|
- ${VOICE_CACHE_DIR}/model:/model:ro
|
|
- ${VOICE_DATA_DIR}:/data
|
|
healthcheck:
|
|
test: [CMD, curl, --fail, --silent, "http://127.0.0.1:${VOICE_PORT}/health"]
|
|
interval: 30s
|
|
timeout: 10s
|
|
retries: 20
|
|
start_period: 10m
|
|
logging: *voice-logs
|
|
|
|
frpc:
|
|
image: ${FRPC_IMAGE}
|
|
restart: unless-stopped
|
|
depends_on:
|
|
voice:
|
|
condition: service_healthy
|
|
env_file:
|
|
- secret.env
|
|
command: [-c, /etc/frp/frpc.toml]
|
|
volumes:
|
|
- ./frpc.toml:/etc/frp/frpc.toml:ro
|
|
network_mode: host
|
|
logging: *voice-logs
|