search

usnistgov / ccu_validation_scoring

Other
5 stars 0 forks source link

score-nd float rounding error #12

Open Jacob-Le opened 8 months ago

Jacob-Le commented 8 months ago

When attempting to apply score-nd to a particular file using version 1.3.2, I get the following error (parameters and steps to reproduce to follow):

File "/Users/e35003/Documents/gitclones/ccu_maksimus/ccu_validation_scoring/CCU_validation_scoring/score_submission.py", line 142, in score_nd_submission_dir_cli score_tad(ref, merged_hyp, "norm", thresholds, args.output_dir, mapping_df, float(args.time_span_scale_collar), float(args.text_span_scale_collar), args.align_hacks) File "/Users/e35003/Documents/gitclones/ccu_maksimus/ccu_validation_scoring/CCU_validation_scoring/score_norm_emotion.py", line 778, in score_tad sumup_tad_class_level_scores(pr_iou_scores, iou_thresholds, output_dir, class_type) File "/Users/e35003/Documents/gitclones/ccu_maksimus/ccu_validation_scoring/CCU_validation_scoring/score_norm_emotion.py", line 697, in sumup_tad_class_level_scores agg_table['value'] = agg_table['value'].round(3) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/e35003/Documents/virtual_environments/MAKSIMUS/lib/python3.11/site-packages/pandas/core/series.py", line 2688, in round result = self._values.round(decimals) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ TypeError: loop of ufunc does not support argument 0 of type float which has no callable rint method

Python version: 3.11 Pandas version: 2.1.3 Numpy version: 1.26.2 Scipy version: 1.11.4

Parameters used:

CCU_scoring score-nd -s <path_to_submission> -ref <path_to_ldc2023e21_version_4.0> -i <path_to_scoring_index> -o <output_path> -t intersection:gt:0 -aR 30 -xR 300 -vR class -aC 15 -xC 150

Scoring index file:

file_id N01006WA4

Isolated submission file that causes the error (file has been included as a csv, simply need to change the extension to .tab N01006WA4.csv):

file_id norm start end status llr N01006WA4 102 104.84 123.4 adhere 0.4697356 N01006WA4 502 10.89 17.49 adhere 0.2709458 N01006WA4 108 24.07 68.22 adhere 0.4048885 N01006WA4 102 135.44 144.52 adhere 0.5015393 N01006WA4 109 165.0 180.0 adhere 0.5656774

Example of a submission file where the error does not occur N01005FR3.csv:

file_id norm start end status llr N01005FR3 109 60.0 90.0 adhere 0.653696 N01005FR3 109 225.0 255.0 adhere 0.5354934 N01005FR3 109 300.0 315.0 adhere 0.6340112 N01005FR3 106 0.0 30.0 adhere 0.9711443 N01005FR3 103 1.88 7.46 adhere 0.2993998 N01005FR3 101 85.09 88.93 adhere 0.0919553 N01005FR3 109 165.0 180.0 violate 0.7992395 N01005FR3 109 375.0 390.0 adhere 0.6774845

Jacob-Le commented 8 months ago

For reference, in an email conversation the issue was that the pandas version for this release needed to be 1.5.3, but was instead 2.1.3. I was advised that the issue would be solved in a future release