Single RF loading at training and dl1->dl2 #985
Conversation
|
|
||
| file_reg_energy = path_models + "/reg_energy.sav" | ||
| joblib.dump(reg_energy, file_reg_energy, compress=3) | ||
| if free_model_memory: |
There was a problem hiding this comment.
There is not really a reason to put this behind a flag, is there? The variable is not used after dumping.
There was a problem hiding this comment.
For some reason the function returns the models (probably useful if not saved). I personally don't know if there are any use case. If not we could remove save_models and free_model_memory
There was a problem hiding this comment.
It is at least used in test currently
Codecov Report
@@ Coverage Diff @@
## master #985 +/- ##
==========================================
+ Coverage 85.69% 85.88% +0.18%
==========================================
Files 78 78
Lines 6572 6645 +73
==========================================
+ Hits 5632 5707 +75
+ Misses 940 938 -2
Continue to review full report at Codecov.
|
…iate betwwen RF file location and already loaded RF
| @@ -165,6 +165,7 @@ def test_build_models(simulated_dl1_file, rf_models): | |||
| infile, | |||
| infile, | |||
| save_models=False, | |||
| free_model_memory=False, | |||
|
If you can just add a test for the new option, otherwise lgtm. |
|
Thanks @vuillaut. I am not allowed to merge myself so if anyone with merging right can do it would be nice. Else I will ask on slack later. |
To solve #979 and reduce the memory usage during training and usage of random forest.