fn_registry/python/functions/ml/genconfig_to_sdcpp_args.md

name, kind, lang, domain, version, purity, signature, description, tags, params, output, uses_functions, uses_types, returns, returns_optional, error_type, imports, tested, tests, test_file_path, file_path

name	kind	lang	domain	version	purity	signature	description	tags	params	output	uses_functions	uses_types	returns	returns_optional	error_type	imports	tested	tests	test_file_path	file_path
genconfig_to_sdcpp_args	function	py	ml	1.0.0	pure	def genconfig_to_sdcpp_args(cfg: GenerationConfig) -> list[str]	Convierte un GenerationConfig a lista de args CLI para stable-diffusion.cpp (sd-cli). Mapea sampler via _SAMPLER_MAP, aplana LoRAs como pares --lora path:weight. Sin I/O ni dependencias externas.	ml sdcpp stable-diffusion-cpp cli converter pure	name desc cfg Instancia de GenerationConfig con los parametros de generacion validados.	Lista de strings con los argumentos CLI en orden. Listo para subprocess.run(['sd'] + args, ...).		generation_config_py_ml		false			true	sampler euler_a se mapea a euler_a en el flag --sampling-method sampler dpm++2m se mapea a dpmpp2m lora con path y weight se agrega como --lora path:weight multiples loras generan multiples pares --lora negative_prompt None produce string vacio en --negative-prompt model.path tiene prioridad sobre model.name en -m args contiene --prompt --seed --steps --cfg-scale --sampling-method -W -H -m	python/functions/ml/tests/test_genconfig_to_sdcpp_args.py	python/functions/ml/genconfig_to_sdcpp_args.py

Ejemplo

from ml.genconfig_to_sdcpp_args import genconfig_to_sdcpp_args
from ml.generation_config import GenerationConfig
from ml.model_ref import ModelRef
from ml.lora_ref import LoraRef

cfg = GenerationConfig(
    prompt="a cat",
    seed=1,
    steps=20,
    cfg_scale=7.0,
    sampler="dpm++2m",
    width=512,
    height=512,
    model=ModelRef(name="v1-5", model_type="sd15", path="/models/v1-5.ckpt"),
    loras=[LoraRef(path="/loras/detail.safetensors", weight=0.8)],
)

args = genconfig_to_sdcpp_args(cfg)
# ["--prompt", "a cat", "--negative-prompt", "", "--seed", "1",
#  "--steps", "20", "--cfg-scale", "7.0", "--sampling-method", "dpmpp2m",
#  "-W", "512", "-H", "512", "-m", "/models/v1-5.ckpt",
#  "--lora", "/loras/detail.safetensors:0.8"]

Notas

Mapa de samplers (_SAMPLER_MAP):

• euler → euler
• euler_a → euler_a
• dpm++2m → dpmpp2m
• dpm++2m_v2 → dpmpp2mv2
• heun → heun
• dpm2 → dpm2
• lcm → lcm

Si cfg.model.path es None, se usa cfg.model.name (nombre de hub o path relativo segun configuracion del entorno sdcpp). Los LoRAs sin path se omiten silenciosamente.