batch tuning via batch_ratio is off on 'mps' since vram is not reported and thus not calculated at all in pipeline_analyze.py. the ratio is calculated only for 'npu' and 'cuda',. adding the following, ...