Problem in calculating step success rate?

[jianwenzh]

I am looking into the eval code and find one place looks weird. See below my comment inline.
When counting whether a 'step' success, suppose we need to only count if if all 'action/type', 'grounding', 'text' are correct at the same time, instead of count in either one of them correct. In the current code logic, the 'grounding' and 'text' buckets are mixed together to count success rate, either of them correct, will counted as step success, although the denominator also counted. This part is weird.

full_type=0
full_step=0
full_gr=0
full_type_hit=0
full_step_hit=0
full_gr_hit=0
for key in [k for k in score_dict.keys() if not k.endswith("full")]:
    if key.endswith("grounding"):
        full_step_hit+=score_dict[key] # full_step should only count if all 'type/action', 'grounding' and 'text' correct at the same time, instead of count in both 'grounding' and 'text'? 
        full_step+=score_dict[key+'_full']
        full_gr_hit+=score_dict[key]
        full_gr+=score_dict[key+'_full']
    elif key.endswith("text"):
        full_step_hit+=score_dict[key] # similarly
        full_step+=score_dict[key+'_full']
    else:
        full_type_hit+=score_dict[key]
        full_type+=score_dict[key+'_full']
    logger.info(f"Type {key} Length {score_dict[key+'_full']} : {(score_dict[key] / score_dict[key+'_full'])}")
logger.info(f"ALL Type : {(full_type_hit / full_type)}")
logger.info(f"ALL Step : {(full_step_hit / full_step)}")
logger.info(f"ALL GR : {(full_gr_hit / full_gr)}")