Remove sequential CPU offload from model loading and simplify return value in safe_model_load function
f725d52
LPX
commited on