Only log relevant config

[Sinacam]

What does this PR do?

Currently, logs.json contains many completely unrelated config items, which is very misleading.
For example, logs from a linear run has

"val_size": 0.2,
"epochs": 10000,
"batch_size": 16,
"optimizer": "adam",
"learning_rate": 0.0001,
"momentum": 0.9,
"lr_scheduler": null,
"patience": 5,
"early_stopping_metric": "P@1",
"init_weight": "kaiming_uniform",
"loss_function": "binary_cross_entropy_with_logits",
"metric_threshold": 0.5,

These items looks reasonable enough that a user may think they are relevant to linear models, which is wrong.
This PR removes all config values that has never been read, which solves this problem.
There are potentially problems with search_params.py because config is serialized when being passed as an argument to other processes. I haven't checked this yet.
search_params.py behaves correctly with this PR as well.

Test CLI & API (bash tests/autotest.sh)

Test APIs used by main.py.

• Test Pass
  • (Copy and paste the last outputted line here.)
• Not Applicable (i.e., the PR does not include API changes.)

Check API Document

If any new APIs are added, please check if the description of the APIs is added to API document.

• API document is updated (linear, nn)
• Not Applicable (i.e., the PR does not include API changes.)

Test quickstart & API (bash tests/docs/test_changed_document.sh)

If any APIs in quickstarts or tutorials are modified, please run this test to check if the current examples can run correctly after the modified APIs are released.

[Gordon119]

Please help change configs at L337 to AttributeDict