8a7e7695f4
- Remove redundant config examples (distributed/, standalone/, examples/) - Delete dev-local.yaml variants (use dev.yaml with env vars) - Delete prod.yaml (use multi-user.yaml or homelab-secure.yaml) - Clean up worker configs: remove docker.yaml, homelab-sandbox.yaml - Update remaining configs with current best practices - Simplify config schema and documentation
89 lines
1.9 KiB
YAML
89 lines
1.9 KiB
YAML
worker_id: "docker-worker"
|
|
base_path: "/tmp/fetchml-jobs"
|
|
entrypoint: "train.py"
|
|
|
|
redis_url: "redis://redis:6379/0"
|
|
|
|
local_mode: true
|
|
|
|
max_workers: 1
|
|
poll_interval_seconds: 2
|
|
|
|
auto_fetch_data: false
|
|
|
|
data_manager_path: "./data_manager"
|
|
dataset_cache_ttl: "30m"
|
|
|
|
data_dir: "/data/active"
|
|
|
|
snapshot_store:
|
|
enabled: true
|
|
endpoint: "blizzard.jfraeys.com"
|
|
secure: true
|
|
bucket: "fetchml-snapshots"
|
|
prefix: "snapshots"
|
|
timeout: "5m"
|
|
max_retries: 3
|
|
|
|
podman_image: "python:3.9-slim"
|
|
container_workspace: "/workspace"
|
|
container_results: "/results"
|
|
gpu_vendor: "nvidia"
|
|
gpu_visible_devices: [0]
|
|
gpu_devices: ["/dev/nvidia0"]
|
|
|
|
|
|
resources:
|
|
max_workers: 1
|
|
desired_rps_per_worker: 2
|
|
podman_cpus: "2"
|
|
podman_memory: "4Gi"
|
|
|
|
metrics:
|
|
enabled: true
|
|
listen_addr: ":9100"
|
|
metrics_flush_interval: "500ms"
|
|
|
|
task_lease_duration: "30m"
|
|
heartbeat_interval: "1m"
|
|
max_retries: 3
|
|
graceful_timeout: "5m"
|
|
|
|
# Plugin Configuration
|
|
plugins:
|
|
# Jupyter Notebook/Lab Service
|
|
jupyter:
|
|
enabled: true
|
|
image: "quay.io/jupyter/base-notebook:latest"
|
|
default_port: 8888
|
|
# Security settings
|
|
security:
|
|
trusted_channels:
|
|
- "conda-forge"
|
|
- "defaults"
|
|
- "pytorch"
|
|
blocked_packages:
|
|
- "requests"
|
|
- "urllib3"
|
|
- "httpx"
|
|
require_password: true
|
|
# Resource limits (enforced by scheduler quota system)
|
|
max_gpu_per_instance: 1
|
|
max_memory_per_instance: "8Gi"
|
|
|
|
# vLLM Inference Service
|
|
vllm:
|
|
enabled: true
|
|
image: "vllm/vllm-openai:latest"
|
|
default_port: 8000
|
|
# Model cache location
|
|
model_cache: "/models"
|
|
# Supported quantization methods: awq, gptq, fp8, squeezellm
|
|
default_quantization: "" # empty = no quantization
|
|
# Resource limits
|
|
max_gpu_per_instance: 4
|
|
max_model_len: 4096
|
|
# Environment variables passed to container
|
|
env:
|
|
- "HF_HOME=/models"
|
|
- "VLLM_WORKER_MULTIPROC_METHOD=spawn"
|