60 lines
1.9 KiB
Bash
60 lines
1.9 KiB
Bash
# Generated as /etc/mike-ai/stack.env by install.sh. Never commit real values.
|
|
AI_BIND_ADDRESS=10.77.0.2
|
|
MODEL_DIR=/data/models
|
|
ROUTER_API_KEY=GENERATED_BY_INSTALLER
|
|
CONTROLLER_TOKEN=GENERATED_BY_INSTALLER
|
|
FLUX_MODEL_DIR=/data/models/FLUX.2-klein-4B
|
|
IMAGE_GPU_DEVICES=1
|
|
PIPER_TTS_VERSION=1.6.0
|
|
PIPER_VOICE=de_DE-thorsten-high
|
|
XTTS_IMAGE=ghcr.io/coqui-ai/xtts-streaming-server:latest-cuda121@sha256:f7fb3b1f9d4bc88af94da1b5959d8002f1e0b003c97557164034eb8a29f01b90
|
|
XTTS_CACHE_DIR=/data/models/xtts-v2-cache
|
|
XTTS_GPU_DEVICE=GPU-4834d9d7-5b61-3004-1fb3-4ae49d482d4b
|
|
AI_DNS=192.168.1.1
|
|
DEFAULT_REASONING_EFFORT=medium
|
|
|
|
FAST_MODEL_FILE=qwen-mix/Qwen3.8-27B-IQ4-MIX.gguf
|
|
MEDIUM_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
|
|
LARGE_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
|
|
ULTRA_MODEL_FILE=qwen-pure/qwen3.8-27b-IQ4_XS-pure.gguf
|
|
UNCENSORED_MODEL_FILE=qwen3.8-27b-abliterated/Qwen3.8-27B-ABLITERATED-Q4_K_M.gguf
|
|
UNCENSORED_PROJECTOR_FILE=qwen3.8-27b-abliterated/mmproj-Qwen3.8-27B-ABLITERATED-F16.gguf
|
|
EXPERIMENTAL_MODEL_FILE=qwen/Qwen3.8-27B-UD-IQ4_XS.gguf
|
|
VISION_PROJECTOR_FILE=qwen/mmproj-BF16.gguf
|
|
|
|
FAST_CONTEXT=76800
|
|
FAST_BATCH_SIZE=64
|
|
FAST_UBATCH_SIZE=32
|
|
MEDIUM_CONTEXT=160000
|
|
MEDIUM_BATCH_SIZE=2048
|
|
MEDIUM_UBATCH_SIZE=128
|
|
LARGE_CONTEXT=192000
|
|
LARGE_BATCH_SIZE=2048
|
|
LARGE_UBATCH_SIZE=128
|
|
ULTRA_CONTEXT=262144
|
|
ULTRA_BATCH_SIZE=2048
|
|
ULTRA_UBATCH_SIZE=128
|
|
UNCENSORED_CONTEXT=80000
|
|
UNCENSORED_BATCH_SIZE=2048
|
|
UNCENSORED_UBATCH_SIZE=128
|
|
EXPERIMENTAL_CONTEXT=76800
|
|
FAST_GPU_DEVICES=0,1
|
|
MEDIUM_GPU_DEVICES=0,1
|
|
MEDIUM_TENSOR_SPLIT=85,15
|
|
LARGE_GPU_DEVICES=0,1
|
|
LARGE_TENSOR_SPLIT=86,14
|
|
ULTRA_GPU_DEVICES=0,1
|
|
ULTRA_TENSOR_SPLIT=80,20
|
|
UNCENSORED_GPU_DEVICES=0,1
|
|
UNCENSORED_TENSOR_SPLIT=90,10
|
|
UNCENSORED_MTP_MAX=2
|
|
EXPERIMENTAL_GPU_DEVICES=0
|
|
LLAMA_THREADS=6
|
|
LLAMA_THREADS_BATCH=6
|
|
FAST_PARALLEL_SLOTS=1
|
|
MEDIUM_PARALLEL_SLOTS=1
|
|
LARGE_PARALLEL_SLOTS=1
|
|
ULTRA_PARALLEL_SLOTS=1
|
|
UNCENSORED_PARALLEL_SLOTS=1
|
|
EXPERIMENTAL_PARALLEL_SLOTS=1
|