Closed TrevorHansen closed 11 months ago
This PR reduces, on my machine, the runtime of the recently merged sharing-aware extractor (#10) from 8.8 seconds to 3.3 seconds.
Loaded 422 jsons. extractors: ['faster-greedy-dag', 'greedy-dag'] data/tensat/nasneta.json 17954094932524.336 15994232636654.406 data/tensat/nasrnn.json 1158.6070294405508 930.8220652824966 data/egg/math_simplify_add.json 3 7 data/egg/math_diff_ln.json 3 4 cumulative time for faster-greedy-dag: 3314ms cumulative time for greedy-dag: 47435ms faster-greedy-dag / greedy-dag geo mean tree: 0.9962 dag: 1.0002 micros: 0.8128 quantiles tree: 0.4286, 1.0000, 1.0000, 1.0000, 1.2447 dag: 0.9957, 1.0000, 1.0000, 1.0000, 1.0225 micros: 0.0284, 0.6061, 0.8866, 1.2203, 2.2857
Was previously:
Loaded 422 jsons. extractors: ['faster-greedy-dag', 'greedy-dag'] data/tensat/nasneta_acyclic.json 16023614866670.953 16023517195541.84 data/tensat/nasrnn.json 1158.6070294405508 930.8220652824966 data/egg/math_simplify_add.json 3 7 Cummulative time for faster-greedy-dag: 8839ms Cummulative time for greedy-dag: 47155ms faster-greedy-dag / greedy-dag geo mean tree: 0.9970 dag: 1.0002 micros: 0.9750 quantiles tree: 0.4286, 1.0000, 1.0000, 1.0000, 1.2447 dag: 0.9957, 1.0000, 1.0000, 1.0000, 1.0225 micros: 0.0910, 0.5888, 1.0630, 1.5546, 3.0305
It also fixes some spelling mistakes I introduced.
Looks great! Thanks!!
This PR reduces, on my machine, the runtime of the recently merged sharing-aware extractor (#10) from 8.8 seconds to 3.3 seconds.
Was previously:
It also fixes some spelling mistakes I introduced.