rorappandClaude Opus 5.5
f90b7df31c
ci / test (push) Successful in 24s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
13 lines
355 B
Bash
13 lines
355 B
Bash
SLOT_ALIAS=tier1-general
|
|
VLLM_MODEL=Qwen/Qwen3-14B-AWQ
|
|
MAX_MODEL_LEN=32768
|
|
GPU_MEM_UTIL=0.90
|
|
MAX_LORAS=4
|
|
MAX_LORA_RANK=64
|
|
MODELS_DIR=/srv/models
|
|
ADAPTERS_DIR=/srv/adapters
|
|
HF_TOKEN=
|
|
# nur gpu-flex mit Profil "tts": GPU_MEM_UTIL für vLLM dann auf ~0.55 senken, damit Qwen3-TTS Platz hat
|
|
QWEN_TTS_IMAGE=
|
|
QWEN_TTS_MODEL=Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice
|