add: refactoring, attributes to parse linear trees as well

[YYYasin19]

This is a new PR that incorporates some of the changes in #15
The other PR can stay open (as a draft maybe) as a reference though I'm not optimistic what its chances of success are, since lzma seems to be doing most of the optimization already.

---

changes here include

• some light refactoring
• start on supporting linear trees in lgbm
• (maybe) some refactoring on the benchmarks

[github-actions]

(benchmark 5088954335 / attempt 1)
Base results / Our results / Change

Model	Size	Dump Time	Load Time
sklearn rf 20M	20.8 MiB / 3.0 MiB / 6.87 x	0.02 s / 0.04 s / 2.17 x	0.02 s / 0.03 s / 1.77 x
sklearn rf 20M lzma	6.5 MiB / 2.0 MiB / 3.26 x	14.26 s / 1.41 s / 0.10 x	0.67 s / 0.22 s / 0.33 x
sklearn rf 200M	212.3 MiB / 30.6 MiB / 6.94 x	0.16 s / 0.38 s / 2.46 x	0.19 s / 0.34 s / 1.75 x
sklearn rf 200M lzma	47.4 MiB / 14.6 MiB / 3.24 x	117.01 s / 21.14 s / 0.18 x	5.19 s / 1.71 s / 0.33 x
sklearn rf 1G	1157.5 MiB / 166.8 MiB / 6.94 x	1.02 s / 1.87 s / 1.83 x	1.17 s / 1.60 s / 1.36 x
sklearn rf 1G lzma	258.1 MiB / 98.1 MiB / 2.63 x	616.50 s / 129.68 s / 0.21 x	28.34 s / 10.27 s / 0.36 x
sklearn gb 2M	2.2 MiB / 1.1 MiB / 2.08 x	0.04 s / 0.35 s / 8.48 x	0.04 s / 0.20 s / 4.48 x
sklearn gb 2M lzma	0.6 MiB / 0.2 MiB / 3.80 x	1.20 s / 0.54 s / 0.45 x	0.11 s / 0.17 s / 1.61 x
lgbm gbdt 2M	2.6 MiB / 1.0 MiB / 2.78 x	0.12 s / 0.31 s / 2.64 x	0.01 s / 0.18 s / 12.26 x
lgbm gbdt 2M lzma	0.9 MiB / 0.5 MiB / 1.90 x	1.98 s / 0.65 s / 0.33 x	0.09 s / 0.22 s / 2.35 x
lgbm gbdt 5M	5.3 MiB / 1.9 MiB / 2.81 x	0.23 s / 0.61 s / 2.67 x	0.03 s / 0.37 s / 12.00 x
lgbm gbdt 5M lzma	1.7 MiB / 0.8 MiB / 1.96 x	4.85 s / 1.38 s / 0.28 x	0.18 s / 0.45 s / 2.45 x
lgbm gbdt 20M	22.7 MiB / 7.6 MiB / 3.00 x	0.91 s / 2.48 s / 2.74 x	0.13 s / 1.50 s / 11.41 x
lgbm gbdt 20M lzma	6.3 MiB / 3.0 MiB / 2.09 x	25.14 s / 6.30 s / 0.25 x	0.73 s / 1.77 s / 2.43 x
lgbm gbdt 100M	101.1 MiB / 33.0 MiB / 3.06 x	3.96 s / 11.09 s / 2.80 x	0.62 s / 147.74 s / 239.18 x
lgbm gbdt 100M lzma	25.6 MiB / 10.6 MiB / 2.41 x	112.24 s / 30.10 s / 0.27 x	2.97 s / 142.08 s / 47.86 x
lgbm rf 10M	10.9 MiB / 3.2 MiB / 3.46 x	0.46 s / 0.83 s / 1.82 x	0.05 s / 0.67 s / 12.74 x
lgbm rf 10M lzma	0.7 MiB / 0.4 MiB / 1.85 x	2.47 s / 1.15 s / 0.47 x	0.14 s / 0.68 s / 4.72 x

[YYYasin19]

according to this file in the lightgbm code, the required features for linear trees are:

• leaf_const
• num_features
• leaf_features
• leaf_coeff

of which we should've covered all :)

[jonashaag]

Thanks!

[pavelzw]

Thanks @YYYasin19!