AGH-CEAI / AGH-CEAI/aegis_gym

[BUG] Load config from ClearML models

Aperta
#120 1 commento 0 reazioni 0 assegnatari Vedi su GitHub
Lingua principale
Python
Stelle
1
Fork
0
Merge medio
3g 1h
PR unite (30g)
2

Descrizione

### Description

We have more use cases for loading config from ClearML. At this moment it works perfectly for training the teacher model. However:
1. For the BC scenario we need to load the RL config and update the current one.
2. For the evaluation scenario [RL], with provided task_id, the DEFAULT RL config should be taken from the ClearML Task.
3. For the evaluation scenario [BC] with task_id the default BC config should be taken from the ClearML.
4. For the evaluation scenario [RL] with provided model_id, the default RL config should be taken the ClearML Model (how?).
5. For the evaluation scenario [BC] with provided model_id, the default BC config should be taken the ClearML Model (how?).

The whole logic should be omitted with the `--enforce-local-config` flag.

### References

- #113

Guida per i contributori

Nessuna guida per i contributori indicizzata per questo repository

Valutazione

Questa issue non è ancora stata valutata.

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.