search

thiagopbueno / model-aware-policy-optimization

MAPO: Model-Aware Policy Optimization algorithm
GNU General Public License v3.0
1 stars 0 forks source link

Feature: add defaults for debugging with episode traces #85

Closed 0xangelo closed 5 years ago

0xangelo commented 5 years ago

I tested it by setting "ouput": "logdir". This saves episode traces under each experimente directory. For example, here's the directory for a tune run with several trials:

(mapo) ➜  MAPO git:(master) ls data/MAPO/
MAPO_Navigation-v0_0_2019-08-22_07-54-195yfufcxt MAPO_Navigation-v0_5_2019-08-22_07-58-586nfnpp56
MAPO_Navigation-v0_1_2019-08-22_07-54-194tq69afn MAPO_Navigation-v0_6_2019-08-22_07-59-01c8s44kly
MAPO_Navigation-v0_2_2019-08-22_07-54-19o3j1hkew MAPO_Navigation-v0_7_2019-08-22_07-59-04wnhg2apo
MAPO_Navigation-v0_3_2019-08-22_07-54-19qty3cogb experiment_state-2019-08-22_07-54-19.json
MAPO_Navigation-v0_4_2019-08-22_07-58-57iyt0eu_r

Under the first directory, we have the following:

(mapo) ➜  MAPO git:(master) ls data/MAPO/MAPO_Navigation-v0_0_2019-08-22_07-54-195yfufcxt
checkpoint_20                                            params.pkl
events.out.tfevents.1566471283.Angelos-MacBook-Pro.local progress.csv
output-2019-08-22_07-54-30_worker-0_0.json               result.json
params.json

In output-DATE_worker_X_0.json we have the episode traces for the corresponding worker. We can check which configuration was running by looking into the params.json file in the same directory.

Is that enough for our purposes?

thiagopbueno commented 5 years ago

That is perfect! Thanks.