debug deeprank for models with no feature value

[manonreau]

1) Compute grid_shape only if it is not provided as an input

2) Compute the feature_mean if self.clip_features == True feature_mean was computed when self.normalize_features == True, this is not required anymore.

3) Format logger message in the compute_norm function

4) Allow feature clipping in the _clip_feature function only if values exists for that feature

5) Add a condition not to transform features values when they correspond to an empty vector in the mapping process

6) set save_hit_rate as False by default since it call the IRMSD target that may not be used by the users

7) Added plots as optional in the test() function since the users, in principle, have no target information for the test set, excepted in benchmark conditions

[coveralls]

Pull Request Test Coverage Report for Build 1468853462

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 77.12%

---

Totals
Change from base Build 1094124503:	0.0%
Covered Lines:	1628
Relevant Lines:	2111

---

💛 - Coveralls

[manonreau]

I have left a few comments but overall it looks good :) Thanks !

I modified the NeuralNet.py class to skip the input data set in the plotting function if no target is provided. This way no benchmarking mode is required.

[CunliangGeng]

Hi @manonreau I pushed two commits (since it's not easy to comment all the details), take a look please :-)

[CunliangGeng]

BTW, please also close the related issues after merging the PR :-)