[AutoNLP] refactor verbosity for more effect log control

[sijunhe]

PR types

PR changes

Description

• use log_to_driver in ray.init for more effective log control
• refactor verbosity to be at the constructor instead of at train
• add CsvLogger to avoid logging with Csv, json and tensorboardx by default

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Merging #4844 (86c0604) into develop (3ae3ea5) will decrease coverage by 0.17%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop    #4844      +/-   ##
===========================================
- Coverage    44.52%   44.36%   -0.17%     
===========================================
  Files          445      446       +1     
  Lines        64042    64328     +286     
===========================================
+ Hits         28515    28537      +22     
- Misses       35527    35791     +264     

Impacted Files	Coverage Δ
...dlenlp/experimental/autonlp/text_classification.py	98.29% <ø> (-0.01%)	⬇️
...addlenlp/experimental/autonlp/auto_trainer_base.py	90.24% <100.00%> (+0.56%)	⬆️
paddlenlp/data/data_collator.py	19.32% <0.00%> (-14.89%)	⬇️
paddlenlp/utils/log.py	67.21% <0.00%> (-3.28%)	⬇️
paddlenlp/trainer/trainer_utils.py	40.67% <0.00%> (-1.34%)	⬇️
paddlenlp/transformers/clip/modeling.py	57.92% <0.00%> (-0.96%)	⬇️
paddlenlp/transformers/mt5/modeling.py	84.67% <0.00%> (ø)
paddlenlp/transformers/bert/tokenizer.py	95.53% <0.00%> (ø)
paddlenlp/transformers/mt5/converter.py	0.00% <0.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[lugimzzz]

lgtm